VGG16_MRNET.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "VGG16_MRNET.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true,
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/ptnv-s/Knee-Abnormality-and-Common-Disorders/blob/colab/VGG16_MRNET.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Deep Learning Based Musculoskeletal Diagnosis, Abnormality and Common Disorders detection using Knee Magnetic Resonance Imaging"
      ],
      "metadata": {
        "id": "fosR27toc8kH"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "---\n",
        "\n",
        "Magnetic resonance (MR) imaging of the knee is the standard of care imaging modality to evaluate knee disorders, and more musculoskeletal MR examinations are performed on the knee than on any other region of the body.\n",
        "\n",
        "---\n",
        "\n",
        "We will be implementing deep learning baselines for detecting general abnormalities and specific diagnoses (anterior cruciate ligament [ACL] tears and meniscal tears) on knee MRI exams.\n",
        "\n",
        "---\n",
        "\n",
        "The most common indications for the knee MRI examinations in this study included acute and chronic pain, follow-up or preoperative evaluation, injury/trauma.\n",
        "\n",
        "---\n",
        "We have used the [MRNet dataset](https://stanfordmlgroup.github.io/competitions/mrnet/)  which consists of 1,370 knee MRI exams performed at Stanford University Medical Center. The dataset contains 1,104 (80.6%) abnormal exams, with 319 (23.3%) ACL tears and 508 (37.1%) meniscal tears. The labels were obtained through manual extraction from clinical reports.\n",
        "\n",
        "---\n",
        "\n",
        "Examinations were performed with GE scanners (GE Discovery, GE Healthcare, Waukesha, WI) with standard knee MRI coil and a routine non-contrast knee MRI protocol that included the following sequences: coronal T1 weighted, coronal T2 with fat saturation, sagittal proton density (PD) weighted, sagittal T2 with fat saturation, and axial PD weighted with fat saturation. A total of 775 (56.6%) examinations used a 3.0-T magnetic field the remaining used a 1.5-T magnetic field as reported [here](https://stanfordmlgroup.github.io/projects/mrnet/) .\n",
        "\n",
        "For more details refer to the paper related to this study [here](https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.1002699).\n",
        "\n",
        "---\n",
        "\n",
        "Given a MRI scan our deep learning model will be predicting 3 outcomes for knee MRI exams (anterior cruciate ligament [ACL] tears, meniscal tears, and general abnormalities)."
      ],
      "metadata": {
        "id": "dT4KaPSJgPWT"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "#### Importing Basic Libraries"
      ],
      "metadata": {
        "id": "gLgaQloxVQNB"
      }
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IgsRtqKALAR5"
      },
      "source": [
        "import os\n",
        "import numpy as np\n",
        "from matplotlib import pyplot as plt\n",
        "import math"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IABg1SuM9wfW"
      },
      "source": [
        "from keras.applications.inception_v3 import InceptionV3\n",
        "from keras.applications.vgg19 import VGG19\n",
        "from keras.preprocessing import image\n",
        "from keras.models import Model, Sequential\n",
        "from keras.layers import Dense, GlobalAveragePooling2D, Dropout, Flatten\n",
        "from keras import backend as K\n",
        "from keras.layers import Input"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "##### Defining Custom Metric for Evaluation"
      ],
      "metadata": {
        "id": "iSYNpSQHwhAj"
      }
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NmrsJ-BsjnyH"
      },
      "source": [
        "import keras.backend as K\n",
        "\n",
        "def f1_score(y_true, y_pred):\n",
        "    c1 = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)))\n",
        "    c2 = K.sum(K.round(K.clip(y_pred, 0, 1)))\n",
        "    c3 = K.sum(K.round(K.clip(y_true, 0, 1)))\n",
        "    if c3 == 0.0 :\n",
        "        return 0.0\n",
        "    precision = c1 / c2\n",
        "    recall = c1 / c3\n",
        "    f1_score = 2 * (precision * recall) / (precision + recall)\n",
        "    return f1_score\n",
        "  \n",
        "\n",
        "def f1(y_true, y_pred):\n",
        "    def recall(y_true, y_pred):\n",
        "        true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)))\n",
        "        possible_positives = K.sum(K.round(K.clip(y_true, 0, 1)))\n",
        "        recall = true_positives / (possible_positives + K.epsilon())\n",
        "        return recall\n",
        "\n",
        "    def precision(y_true, y_pred):\n",
        "        true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)))\n",
        "        predicted_positives = K.sum(K.round(K.clip(y_pred, 0, 1)))\n",
        "        precision = true_positives / (predicted_positives + K.epsilon())\n",
        "        return precision\n",
        "    precision = precision(y_true, y_pred)\n",
        "    recall = recall(y_true, y_pred)\n",
        "    return 2*((precision*recall)/(precision+recall+K.epsilon()))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "##Loading Files for loading Different Planes of Scans of Dataset\n",
        "As we know from EDA Notebook this dataset has 3 views of Knee since datset has MRI scans namely -\n",
        "\n",
        "*   Axial\n",
        "*   Coronal\n",
        "*   Sagittal"
      ],
      "metadata": {
        "id": "6zl48zFBZKST"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "metadata": {
        "id": "VyF5dHFy5Unq",
        "outputId": "29ffc5a1-17e6-44c7-c16b-8235c656c2b2",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "U4pEnRbJ8zbJ",
        "outputId": "2f3e65c6-dd24-4635-fc5e-a277298403bb",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "import csv\n",
        "with open(\"/content/drive/MyDrive/Medical/MRNet-v1.0/train-abnormal.csv\") as csv_file:\n",
        "    csv_reader = csv.reader(csv_file)\n",
        "    line_count = 0\n",
        "    trainAb_labels=[]\n",
        "    for row in csv_reader:\n",
        "           trainAb_labels.append(int(row[1]))\n",
        "print(trainAb_labels)\n",
        "trainAb_labels= np.array(trainAb_labels)\n",
        "print(len(trainAb_labels))\n",
        "\n",
        "with open(\"/content/drive/MyDrive/Medical/MRNet-v1.0/train-acl.csv\") as csv_file:\n",
        "    csv_reader = csv.reader(csv_file)\n",
        "    line_count = 0\n",
        "    trainACL_labels=[]\n",
        "    for row in csv_reader:\n",
        "           trainACL_labels.append(int(row[1]))\n",
        "print(trainAb_labels)\n",
        "trainACL_labels= np.array(trainACL_labels)\n",
        "print(len(trainACL_labels))\n",
        "\n",
        "\n",
        "with open(\"/content/drive/MyDrive/Medical/MRNet-v1.0/train-meniscus.csv\") as csv_file:\n",
        "    csv_reader = csv.reader(csv_file)\n",
        "    line_count = 0\n",
        "    trainMen_labels=[]\n",
        "    for row in csv_reader:\n",
        "           trainMen_labels.append(int(row[1]))\n",
        "print(trainMen_labels)\n",
        "trainAb_labels= np.array(trainMen_labels)\n",
        "print(len(trainMen_labels))\n",
        "\n",
        "with open(\"/content/drive/MyDrive/Medical/MRNet-v1.0/valid-abnormal.csv\") as csv_file:\n",
        "    csv_reader = csv.reader(csv_file)\n",
        "    line_count = 0\n",
        "    testAb_labels=[]\n",
        "    for row in csv_reader:\n",
        "           testAb_labels.append(int(row[1]))\n",
        "print(testAb_labels)\n",
        "testAb_labels= np.array(testAb_labels)\n",
        "print(len(testAb_labels))\n",
        "\n",
        "with open(\"/content/drive/MyDrive/Medical/MRNet-v1.0/valid-acl.csv\") as csv_file:\n",
        "    csv_reader = csv.reader(csv_file)\n",
        "    line_count = 0\n",
        "    testACL_labels=[]\n",
        "    for row in csv_reader:\n",
        "           testACL_labels.append(int(row[1]))\n",
        "print(testACL_labels)\n",
        "testACL_labels= np.array(testACL_labels)\n",
        "print(len(testACL_labels))\n",
        "\n",
        "\n",
        "with open(\"/content/drive/MyDrive/Medical/MRNet-v1.0/valid-meniscus.csv\") as csv_file:\n",
        "    csv_reader = csv.reader(csv_file)\n",
        "    line_count = 0\n",
        "    testMen_labels=[]\n",
        "    for row in csv_reader:\n",
        "           testMen_labels.append(int(row[1]))\n",
        "print(testMen_labels)\n",
        "testMen_labels= np.array(testMen_labels)\n",
        "print(len(testMen_labels))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 0, 1, 1, 0, 1, 1, 1, 0, 1, 0, 1, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 0, 1, 0, 1, 0, 0, 1, 1, 1, 0, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 0, 1, 0, 0, 1, 0, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1, 0, 1, 1, 0, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 0, 1, 0, 1, 1, 1, 0, 0, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 1, 1, 1, 0, 0, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 0, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 0, 1, 0, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 1, 0, 1, 0, 0, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 0, 1, 0, 0, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 0, 1, 0, 1, 1, 1, 0, 0, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 0, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 0, 0, 1, 0, 1, 1, 0, 1, 0, 0, 0, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 0, 1, 1, 1, 1, 0, 0, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1]\n",
            "1130\n",
            "[1 1 1 ... 0 1 1]\n",
            "1130\n",
            "[0, 1, 0, 1, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 1, 1, 1, 0, 1, 0, 0, 0, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 1, 0, 0, 1, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 1, 1, 1, 0, 0, 0, 1, 0, 0, 0, 1, 1, 1, 1, 1, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 1, 0, 1, 0, 1, 0, 0, 0, 0, 1, 0, 0, 1, 0, 1, 0, 1, 1, 0, 1, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 1, 1, 0, 1, 1, 0, 0, 1, 0, 1, 0, 0, 1, 0, 0, 1, 0, 1, 0, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 1, 0, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 1, 1, 0, 0, 0, 1, 0, 1, 0, 1, 1, 0, 1, 1, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 1, 0, 1, 1, 0, 1, 0, 1, 0, 1, 0, 0, 1, 0, 0, 1, 0, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 1, 1, 0, 0, 0, 0, 1, 0, 0, 0, 1, 1, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 1, 0, 0, 0, 1, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 1, 1, 0, 0, 0, 1, 0, 1, 1, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 1, 1, 0, 1, 1, 0, 1, 0, 1, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 1, 1, 0, 1, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 1, 0, 0, 0, 1, 0, 0, 0, 1, 1, 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 1, 1, 0, 0, 0, 0, 1, 1, 0, 1, 0, 0, 1, 1, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 1, 1, 0, 0, 1, 0, 0, 0, 1, 1, 0, 0, 1, 1, 1, 0, 0, 0, 0, 1, 1, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 1, 0, 1, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, 0, 1, 1, 0, 0, 0, 1, 0, 1, 1, 0, 1, 1, 0, 0, 0, 1, 1, 1, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1, 1, 1, 0, 0, 1, 0, 1, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 1, 1, 0, 1, 1, 1, 0, 1, 0, 0, 1, 1, 0, 1, 1, 0, 0, 0, 1, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 1, 1, 0, 0, 0, 0, 1, 1, 0, 1, 0, 0, 0, 0, 1, 1, 0, 1, 0, 1, 0, 0, 1, 0, 0, 0, 1, 1, 0, 0, 1, 1, 0, 0, 0, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 1, 1, 0, 1, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 1, 0, 0, 1, 0, 1, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 1, 1, 0, 0, 1, 0, 1, 1, 0, 1, 1, 1, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0]\n",
            "1130\n",
            "[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]\n",
            "120\n",
            "[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 0, 0, 0, 1, 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 1, 0]\n",
            "120\n",
            "[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 0, 1, 1, 0, 1, 1, 1, 1, 0, 1, 0, 1, 1, 1, 0, 1, 0, 1, 0, 0, 0, 1, 0, 1, 0, 0, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]\n",
            "120\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DzmL2nU9M9wt"
      },
      "source": [
        "train_Ab_pred = []\n",
        "test_Ab_pred = []\n",
        "train_Men_pred = []\n",
        "test_Men_pred = []\n",
        "train_ACL_pred = []\n",
        "test_ACL_pred = []"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "HpaCqxa5Lji9"
      },
      "source": [
        "##AxialPlane Scans"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AUmEt_uPNA7K"
      },
      "source": [
        "###Reading train & test data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xvMN_V8lKfA4",
        "outputId": "e84a0875-cbdb-492d-fa73-3fbe7d9bf0a3",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "\n",
        "axial_dir = os.listdir(\"/content/drive/MyDrive/Medical/MRNet-v1.0/train/axial/\")\n",
        "axial_dir = [f.lower() for f in axial_dir]   \n",
        "a=sorted(axial_dir)\n",
        "axial_list=np.array(a)\n",
        "axial_reloaded=[]\n",
        "\n",
        "print(axial_list.shape)\n",
        "for j in range(1,1131):\n",
        "    file= \"/content/drive/MyDrive/Medical/MRNet-v1.0/train/axial/\"+axial_list[j]    \n",
        "    axial_reloaded.append(np.load(file,allow_pickle='false'))   \n",
        "\n",
        "axial_reloaded=np.array(axial_reloaded)\n",
        "\n",
        "print(axial_reloaded.shape)\n",
        "s=axial_reloaded.shape\n",
        "print(axial_reloaded[0].shape[0])\n",
        "\n",
        "train_axial=[]\n",
        "\n",
        "print(len(axial_reloaded))\n",
        "\n",
        "for j in range(len(axial_reloaded)):\n",
        "    mid=math.floor((axial_reloaded[j].shape[0])/2)\n",
        "    train_axial.append([])    \n",
        "\n",
        "    for n in range(mid-1,mid+2):\n",
        "         train_axial[j].append(axial_reloaded[j][n])\n",
        "train_axial=np.array(train_axial)\n",
        "train_axial=np.reshape(train_axial,(1130,256,256,3))\n",
        "print(train_axial.shape)\n",
        "\n",
        "axial_dir = os.listdir(\"/content/drive/MyDrive/Medical/MRNet-v1.0/valid/axial/\")\n",
        "axial_dir = [f.lower() for f in axial_dir]   \n",
        "a=sorted(axial_dir)\n",
        "axialTest_list=np.array(a)\n",
        "axialTest_reloaded=[]\n",
        "\n",
        "print(axialTest_list.shape)\n",
        "\n",
        "for j in range(0,120):\n",
        "    file= \"/content/drive/MyDrive/Medical/MRNet-v1.0/valid/axial/\"+axialTest_list[j]\n",
        "    axialTest_reloaded.append(np.load(file,allow_pickle='false'))   \n",
        "\n",
        "axialTest_reloaded=np.array(axialTest_reloaded)\n",
        "\n",
        "print(axialTest_reloaded.shape)\n",
        "\n",
        "s=axialTest_reloaded.shape\n",
        "print(axialTest_reloaded[0].shape[0])\n",
        "\n",
        "test_axial=[]\n",
        "\n",
        "print(len(axialTest_reloaded))\n",
        "\n",
        "for j in range(len(axialTest_reloaded)):\n",
        "    mid=math.floor((axialTest_reloaded[j].shape[0])/2)\n",
        "    test_axial.append([])    \n",
        "\n",
        "    for n in range(mid-1,mid+2):\n",
        "         test_axial[j].append(axialTest_reloaded[j][n])\n",
        "test_axial=np.array(test_axial)\n",
        "test_axial=np.reshape(test_axial,(120,256,256,3))\n",
        "print(test_axial.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(1131,)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "JLywW23cJSWb"
      },
      "source": [
        "#### Finetuning VGG16 Model for Axial Plane Scanes checking for Meniscus Condition"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "FwBkOdWI_tjx",
        "outputId": "2a6e18f7-6f2f-4187-d4e1-977cbbcb5a7f",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "input_tensor = Input(shape=(256, 256,3))  \n",
        "base_model_ax_men = VGG16(input_tensor=input_tensor, weights='imagenet', include_top=False)\n",
        "print(len(base_model_ax_men.layers))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/vgg16/vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5\n",
            "58892288/58889256 [==============================] - 0s 0us/step\n",
            "58900480/58889256 [==============================] - 0s 0us/step\n",
            "19\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uzEJBpRkUYX7"
      },
      "source": [
        "x = base_model_ax_men.output\n",
        "x = GlobalAveragePooling2D()(x)\n",
        "x=Dropout(0.2)(x)\n",
        "x = Dense(1024, activation='relu')(x)\n",
        "x=Dropout(0.2)(x)\n",
        "x=Dropout(0.2)(x)\n",
        "predictions = Dense(1, activation='sigmoid')(x)\n",
        "model = Model(inputs=base_model_ax_men.input,outputs=predictions)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uPoHK2f3r1ul"
      },
      "source": [
        "for layer in model.layers[:15]:\n",
        "   layer.trainable = True\n",
        "for layer in model.layers[15:]:\n",
        "  layer.trainable = False"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "trainMen_labels = np.array(trainMen_labels)"
      ],
      "metadata": {
        "id": "YfZte1bG7wxv"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "from tensorflow.keras.callbacks import EarlyStopping\n",
        "from tensorflow.keras.optimizers import SGD"
      ],
      "metadata": {
        "id": "CjTKu3t0EK5W"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "btpx20hvfyuv",
        "outputId": "a73ab664-04fb-462b-bf6a-0c23f44179fb",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "early_stopping_monitor = EarlyStopping(patience=10)    \n",
        "model.compile(optimizer=SGD(lr=0.001,momentum=0.9), loss='binary_crossentropy',metrics=['accuracy', f1_score])\n",
        "history=model.fit(train_axial ,trainMen_labels, validation_split=0.05,shuffle=True ,epochs=100, callbacks=[early_stopping_monitor], batch_size = 16)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/keras/optimizer_v2/gradient_descent.py:102: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.\n",
            "  super(SGD, self).__init__(name, **kwargs)\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/100\n",
            "68/68 [==============================] - 59s 613ms/step - loss: 0.8385 - accuracy: 0.5862 - f1_score: nan - val_loss: 0.6708 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 2/100\n",
            "68/68 [==============================] - 38s 554ms/step - loss: 0.6576 - accuracy: 0.6291 - f1_score: nan - val_loss: 0.6608 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 3/100\n",
            "68/68 [==============================] - 38s 555ms/step - loss: 0.6471 - accuracy: 0.6459 - f1_score: nan - val_loss: 0.6453 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 4/100\n",
            "68/68 [==============================] - 38s 554ms/step - loss: 0.6380 - accuracy: 0.6486 - f1_score: nan - val_loss: 0.6354 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 5/100\n",
            "68/68 [==============================] - 38s 552ms/step - loss: 0.6272 - accuracy: 0.6524 - f1_score: nan - val_loss: 0.6293 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 6/100\n",
            "68/68 [==============================] - 37s 551ms/step - loss: 0.6239 - accuracy: 0.6589 - f1_score: nan - val_loss: 0.6319 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 7/100\n",
            "68/68 [==============================] - 37s 551ms/step - loss: 0.5994 - accuracy: 0.6570 - f1_score: nan - val_loss: 0.6104 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 8/100\n",
            "68/68 [==============================] - 37s 551ms/step - loss: 0.5712 - accuracy: 0.6719 - f1_score: nan - val_loss: 0.6232 - val_accuracy: 0.6316 - val_f1_score: nan\n",
            "Epoch 9/100\n",
            "68/68 [==============================] - 37s 551ms/step - loss: 0.5944 - accuracy: 0.6710 - f1_score: nan - val_loss: 0.6040 - val_accuracy: 0.6316 - val_f1_score: nan\n",
            "Epoch 10/100\n",
            "68/68 [==============================] - 38s 552ms/step - loss: 0.5731 - accuracy: 0.6747 - f1_score: nan - val_loss: 0.6042 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 11/100\n",
            "68/68 [==============================] - 37s 551ms/step - loss: 0.5732 - accuracy: 0.6878 - f1_score: nan - val_loss: 0.6199 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 12/100\n",
            "68/68 [==============================] - 38s 552ms/step - loss: 0.5577 - accuracy: 0.6841 - f1_score: nan - val_loss: 0.5964 - val_accuracy: 0.7193 - val_f1_score: 0.6107\n",
            "Epoch 13/100\n",
            "68/68 [==============================] - 37s 551ms/step - loss: 0.5485 - accuracy: 0.6952 - f1_score: nan - val_loss: 0.5965 - val_accuracy: 0.7193 - val_f1_score: 0.5179\n",
            "Epoch 14/100\n",
            "68/68 [==============================] - 38s 552ms/step - loss: 0.5397 - accuracy: 0.6850 - f1_score: nan - val_loss: 0.5976 - val_accuracy: 0.7544 - val_f1_score: 0.6250\n",
            "Epoch 15/100\n",
            "68/68 [==============================] - 38s 552ms/step - loss: 0.5316 - accuracy: 0.7167 - f1_score: nan - val_loss: 0.6142 - val_accuracy: 0.6667 - val_f1_score: 0.4837\n",
            "Epoch 16/100\n",
            "68/68 [==============================] - 38s 552ms/step - loss: 0.5338 - accuracy: 0.7055 - f1_score: nan - val_loss: 0.6063 - val_accuracy: 0.7018 - val_f1_score: 0.6173\n",
            "Epoch 17/100\n",
            "68/68 [==============================] - 38s 552ms/step - loss: 0.5356 - accuracy: 0.6925 - f1_score: nan - val_loss: 0.6011 - val_accuracy: 0.7544 - val_f1_score: 0.6478\n",
            "Epoch 18/100\n",
            "68/68 [==============================] - 38s 552ms/step - loss: 0.5188 - accuracy: 0.7111 - f1_score: nan - val_loss: 0.6291 - val_accuracy: 0.7368 - val_f1_score: 0.6199\n",
            "Epoch 19/100\n",
            "68/68 [==============================] - 38s 552ms/step - loss: 0.5066 - accuracy: 0.7158 - f1_score: nan - val_loss: 0.6377 - val_accuracy: 0.7368 - val_f1_score: 0.6278\n",
            "Epoch 20/100\n",
            "68/68 [==============================] - 38s 552ms/step - loss: 0.5007 - accuracy: 0.7213 - f1_score: nan - val_loss: 0.6575 - val_accuracy: 0.6667 - val_f1_score: 0.4686\n",
            "Epoch 21/100\n",
            "68/68 [==============================] - 38s 552ms/step - loss: 0.5038 - accuracy: 0.7297 - f1_score: nan - val_loss: 0.6142 - val_accuracy: 0.7544 - val_f1_score: 0.6429\n",
            "Epoch 22/100\n",
            "68/68 [==============================] - 38s 552ms/step - loss: 0.5006 - accuracy: 0.7279 - f1_score: nan - val_loss: 0.6138 - val_accuracy: 0.8070 - val_f1_score: 0.6877\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4LHGDU-xtbvJ"
      },
      "source": [
        "####Plotting accuracy of both training and validation sets per epoch."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eJh9LZBSrfij",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "2f1be9e7-40bd-439a-d707-4fa6499947c0"
      },
      "source": [
        "p = np.array(model.predict(train_axial))\n",
        "train_Men_pred.append(p)\n",
        "p = np.array(model.predict(test_axial))\n",
        "test_Men_pred.append(p)\n",
        "plt.plot(history.history['accuracy'])\n",
        "plt.plot(history.history['val_accuracy'])\n",
        "plt.plot(history.history['f1_score'])\n",
        "plt.title('model accuracy')\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'test', 'f1score'], loc='upper left')\n",
        "plt.grid() \n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "##### Helper function to clear Memory Utilization enabling it to Work in Colab with K80 GPU and standard RAM allowance."
      ],
      "metadata": {
        "id": "fZ5xHX9YuhtR"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import gc\n",
        "def clean_up(model):\n",
        "    K.clear_session()\n",
        "    del model\n",
        "    gc.collect() #garbage collection\n",
        "    print(\"...Finished clean up...\")"
      ],
      "metadata": {
        "id": "Y7lPq_uhdGAn"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "filename = '/content/drive/MyDrive/Medical/MRNET MODEL/modelvgg1.h5'\n",
        "model.save(filename)\n",
        "print('>Saved %s' % filename)\n",
        "clean_up(model=model)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "VWbftECScuC6",
        "outputId": "fb051ced-dab3-44ab-fcfc-a3360b90ac11"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            ">Saved /content/drive/MyDrive/Medical/MRNET MODEL/modelvgg1.h5\n",
            "...Finished clean up...\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wmZf2R0ezncx"
      },
      "source": [
        "#### Finetuning VGG16 Model for Axial Plane Scanes checking for ACL Condition"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "W91sxbOZzgVF"
      },
      "source": [
        "input_tensor = Input(shape=(256, 256,3))  \n",
        "base_model_ax_ACL = VGG16(input_tensor=input_tensor, weights='imagenet', include_top=False)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yEnh6bpQzw0s"
      },
      "source": [
        "x = base_model_ax_ACL.output\n",
        "x = GlobalAveragePooling2D()(x)\n",
        "x=Dropout(0.2)(x)\n",
        "x = Dense(1024, activation='relu')(x)\n",
        "x=Dropout(0.2)(x)\n",
        "x=Dropout(0.2)(x)\n",
        "predictions = Dense(1, activation='sigmoid')(x)\n",
        "model2 = Model(inputs=base_model_ax_ACL.input,outputs=predictions)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Qcy9VIrbz0uS"
      },
      "source": [
        "for layer in model2.layers[:15]:\n",
        "   layer.trainable = True\n",
        "for layer in model2.layers[15:]:\n",
        "  layer.trainable = False"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "K5bkMw8rz53v",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "a65ce335-1986-4522-c537-7b7c1f8eb0e0"
      },
      "source": [
        "early_stopping_monitor = EarlyStopping(patience=10)    \n",
        "model2.compile(optimizer=SGD(lr=0.001,momentum=0.9), loss='binary_crossentropy',metrics=['accuracy',f1_score])\n",
        "history2=model2.fit(train_axial ,trainACL_labels, validation_split=0.05,shuffle=True ,epochs=100, callbacks=[early_stopping_monitor], batch_size = 20)\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/100\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/keras/optimizer_v2/gradient_descent.py:102: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.\n",
            "  super(SGD, self).__init__(name, **kwargs)\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "54/54 [==============================] - 61s 990ms/step - loss: 0.6492 - accuracy: 0.7558 - f1_score: nan - val_loss: 0.4591 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 2/100\n",
            "54/54 [==============================] - 46s 845ms/step - loss: 0.5229 - accuracy: 0.7968 - f1_score: nan - val_loss: 0.4350 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 3/100\n",
            "54/54 [==============================] - 46s 844ms/step - loss: 0.4963 - accuracy: 0.8024 - f1_score: nan - val_loss: 0.4253 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 4/100\n",
            "54/54 [==============================] - 46s 844ms/step - loss: 0.4985 - accuracy: 0.8080 - f1_score: nan - val_loss: 0.4436 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 5/100\n",
            "54/54 [==============================] - 46s 843ms/step - loss: 0.4861 - accuracy: 0.8136 - f1_score: nan - val_loss: 0.3998 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 6/100\n",
            "54/54 [==============================] - 46s 844ms/step - loss: 0.4665 - accuracy: 0.8089 - f1_score: nan - val_loss: 0.4350 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 7/100\n",
            "54/54 [==============================] - 46s 844ms/step - loss: 0.4299 - accuracy: 0.8117 - f1_score: nan - val_loss: 0.3982 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 8/100\n",
            "54/54 [==============================] - 46s 845ms/step - loss: 0.4371 - accuracy: 0.8108 - f1_score: nan - val_loss: 0.3719 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 9/100\n",
            "54/54 [==============================] - 46s 847ms/step - loss: 0.4127 - accuracy: 0.8145 - f1_score: nan - val_loss: 0.3580 - val_accuracy: 0.8246 - val_f1_score: nan\n",
            "Epoch 10/100\n",
            "54/54 [==============================] - 46s 847ms/step - loss: 0.3765 - accuracy: 0.8211 - f1_score: nan - val_loss: 0.3507 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 11/100\n",
            "54/54 [==============================] - 46s 847ms/step - loss: 0.4022 - accuracy: 0.8239 - f1_score: nan - val_loss: 0.3670 - val_accuracy: 0.8246 - val_f1_score: nan\n",
            "Epoch 12/100\n",
            "54/54 [==============================] - 46s 846ms/step - loss: 0.4115 - accuracy: 0.8239 - f1_score: nan - val_loss: 0.3521 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 13/100\n",
            "54/54 [==============================] - 46s 847ms/step - loss: 0.3912 - accuracy: 0.8295 - f1_score: nan - val_loss: 0.4246 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 14/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.3899 - accuracy: 0.8304 - f1_score: nan - val_loss: 0.3342 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 15/100\n",
            "54/54 [==============================] - 46s 847ms/step - loss: 0.3750 - accuracy: 0.8229 - f1_score: nan - val_loss: 0.3400 - val_accuracy: 0.8246 - val_f1_score: nan\n",
            "Epoch 16/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.3718 - accuracy: 0.8201 - f1_score: nan - val_loss: 0.3540 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 17/100\n",
            "54/54 [==============================] - 46s 846ms/step - loss: 0.3480 - accuracy: 0.8425 - f1_score: nan - val_loss: 0.3681 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 18/100\n",
            "54/54 [==============================] - 46s 847ms/step - loss: 0.3548 - accuracy: 0.8369 - f1_score: nan - val_loss: 0.3354 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 19/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.3467 - accuracy: 0.8378 - f1_score: nan - val_loss: 0.3288 - val_accuracy: 0.8596 - val_f1_score: nan\n",
            "Epoch 20/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.3465 - accuracy: 0.8453 - f1_score: nan - val_loss: 0.3506 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 21/100\n",
            "54/54 [==============================] - 46s 847ms/step - loss: 0.3427 - accuracy: 0.8490 - f1_score: nan - val_loss: 0.3308 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 22/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.3231 - accuracy: 0.8546 - f1_score: nan - val_loss: 0.3640 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 23/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.3177 - accuracy: 0.8555 - f1_score: nan - val_loss: 0.3371 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 24/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.3160 - accuracy: 0.8611 - f1_score: nan - val_loss: 0.4652 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 25/100\n",
            "54/54 [==============================] - 46s 847ms/step - loss: 0.3183 - accuracy: 0.8621 - f1_score: nan - val_loss: 0.4684 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 26/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.3216 - accuracy: 0.8537 - f1_score: nan - val_loss: 0.3208 - val_accuracy: 0.8947 - val_f1_score: nan\n",
            "Epoch 27/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.3045 - accuracy: 0.8667 - f1_score: nan - val_loss: 0.3448 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 28/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.2986 - accuracy: 0.8630 - f1_score: nan - val_loss: 0.3186 - val_accuracy: 0.8772 - val_f1_score: 0.5635\n",
            "Epoch 29/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.2958 - accuracy: 0.8621 - f1_score: nan - val_loss: 0.3381 - val_accuracy: 0.8772 - val_f1_score: nan\n",
            "Epoch 30/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.2750 - accuracy: 0.8863 - f1_score: nan - val_loss: 0.3222 - val_accuracy: 0.8772 - val_f1_score: nan\n",
            "Epoch 31/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.2891 - accuracy: 0.8770 - f1_score: nan - val_loss: 0.3501 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 32/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.2808 - accuracy: 0.8742 - f1_score: nan - val_loss: 0.3132 - val_accuracy: 0.9123 - val_f1_score: 0.6333\n",
            "Epoch 33/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.2496 - accuracy: 0.8919 - f1_score: nan - val_loss: 0.3690 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 34/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.2658 - accuracy: 0.8807 - f1_score: nan - val_loss: 0.3836 - val_accuracy: 0.8947 - val_f1_score: nan\n",
            "Epoch 35/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.2333 - accuracy: 0.8947 - f1_score: nan - val_loss: 0.3096 - val_accuracy: 0.8596 - val_f1_score: 0.5857\n",
            "Epoch 36/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.2262 - accuracy: 0.9096 - f1_score: nan - val_loss: 0.2983 - val_accuracy: 0.9123 - val_f1_score: 0.6746\n",
            "Epoch 37/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.2255 - accuracy: 0.8975 - f1_score: nan - val_loss: 0.3780 - val_accuracy: 0.8947 - val_f1_score: nan\n",
            "Epoch 38/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.1973 - accuracy: 0.9199 - f1_score: nan - val_loss: 0.4208 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 39/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.2111 - accuracy: 0.9180 - f1_score: nan - val_loss: 0.3805 - val_accuracy: 0.8596 - val_f1_score: nan\n",
            "Epoch 40/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.2077 - accuracy: 0.9226 - f1_score: nan - val_loss: 0.3291 - val_accuracy: 0.8596 - val_f1_score: nan\n",
            "Epoch 41/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.1719 - accuracy: 0.9329 - f1_score: nan - val_loss: 0.5981 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 42/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.1746 - accuracy: 0.9366 - f1_score: nan - val_loss: 0.3681 - val_accuracy: 0.8947 - val_f1_score: 0.5667\n",
            "Epoch 43/100\n",
            "54/54 [==============================] - 46s 851ms/step - loss: 0.1806 - accuracy: 0.9217 - f1_score: nan - val_loss: 0.3424 - val_accuracy: 0.8772 - val_f1_score: 0.5778\n",
            "Epoch 44/100\n",
            "54/54 [==============================] - 46s 850ms/step - loss: 0.1486 - accuracy: 0.9432 - f1_score: nan - val_loss: 0.5047 - val_accuracy: 0.8772 - val_f1_score: nan\n",
            "Epoch 45/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.1497 - accuracy: 0.9469 - f1_score: nan - val_loss: 0.3887 - val_accuracy: 0.8947 - val_f1_score: 0.5857\n",
            "Epoch 46/100\n",
            "54/54 [==============================] - 46s 850ms/step - loss: 0.1489 - accuracy: 0.9338 - f1_score: nan - val_loss: 0.5396 - val_accuracy: 0.8596 - val_f1_score: nan\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0acHIajk0WVz",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 354
        },
        "outputId": "8a4eea99-b290-4348-b565-3156f16bf1bb"
      },
      "source": [
        "p = np.array(model2.predict(train_axial))\n",
        "train_ACL_pred.append(p)\n",
        "p = np.array(model2.predict(test_axial))\n",
        "test_ACL_pred.append(p)                    \n",
        "score2 = model2.evaluate(test_axial ,testACL_labels,verbose=1)       \n",
        "print('Train loss:',score2[0])\n",
        "print('Train accuracy:',score2[1])                                                                \n",
        "plt.figure()\n",
        "plt.grid()                                                                                                     \n",
        "plt.title('Training performance')                          \n",
        "plt.plot(history2.epoch, history2.history['loss'], label='train loss+error')  \n",
        "plt.plot(history2.epoch, history2.history['val_loss'], label='val_error')\n",
        "plt.plot(history2.epoch, history2.history['f1_score'], label='f1_score')\n",
        "plt.legend()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "4/4 [==============================] - 1s 359ms/step - loss: 1.0762 - accuracy: 0.7083 - f1_score: 0.3989\n",
            "Train loss: 1.0762392282485962\n",
            "Train accuracy: 0.7083333134651184\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7fe3be262150>"
            ]
          },
          "metadata": {},
          "execution_count": 23
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "filename = '/content/drive/MyDrive/Medical/MRNET MODEL/modelVGG162.h5'\n",
        "model2.save(filename)\n",
        "print('>Saved %s' % filename)\n",
        "clean_up(model=model2)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9ylqxFhbekCo",
        "outputId": "6ae28c96-d172-4e0d-b979-aa6ca2e05a7a"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            ">Saved /content/drive/MyDrive/Medical/MRNET MODEL/modelVGG162.h5\n",
            "...Finished clean up...\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "JIzTeIQJ08Ep"
      },
      "source": [
        "#### Finetuning VGG16 Model for Axial Plane Scanes checking for Abnormal Condition"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "L-L9iI9z0-wB"
      },
      "source": [
        "input_tensor = Input(shape=(256, 256,3))  \n",
        "base_model_ax_AB = VGG16(input_tensor=input_tensor, weights='imagenet', include_top=False)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "y2exvZb-19SL"
      },
      "source": [
        "x = base_model_ax_AB.output\n",
        "x = GlobalAveragePooling2D()(x)\n",
        "x=Dropout(0.2)(x)\n",
        "x = Dense(1024, activation='relu')(x)\n",
        "x=Dropout(0.2)(x)\n",
        "x=Dropout(0.2)(x)\n",
        "predictions = Dense(1, activation='sigmoid')(x)\n",
        "model3 = Model(inputs=base_model_ax_AB.input,outputs=predictions)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "a7mXwAtA2IL0"
      },
      "source": [
        "for layer in model3.layers[:15]:\n",
        "   layer.trainable = True\n",
        "for layer in model3.layers[15:]:\n",
        "  layer.trainable = False"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gfK3QM3Y2OBx",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "e920d46f-bd11-49d6-c3c7-0a1458c3ebcd"
      },
      "source": [
        "early_stopping_monitor = EarlyStopping(patience=10)    \n",
        "model3.compile(optimizer=SGD(lr=0.001,momentum=0.9), loss='binary_crossentropy',metrics=['accuracy',f1_score])\n",
        "history3=model3.fit(train_axial ,trainAb_labels, validation_split=0.05,shuffle=True ,epochs=100, callbacks=[early_stopping_monitor], batch_size = 20)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/100\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/keras/optimizer_v2/gradient_descent.py:102: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.\n",
            "  super(SGD, self).__init__(name, **kwargs)\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "54/54 [==============================] - 47s 851ms/step - loss: 0.7498 - accuracy: 0.6263 - f1_score: nan - val_loss: 0.6541 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 2/100\n",
            "54/54 [==============================] - 46s 845ms/step - loss: 0.6590 - accuracy: 0.6431 - f1_score: nan - val_loss: 0.6634 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 3/100\n",
            "54/54 [==============================] - 46s 846ms/step - loss: 0.6576 - accuracy: 0.6403 - f1_score: nan - val_loss: 0.6436 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 4/100\n",
            "54/54 [==============================] - 46s 845ms/step - loss: 0.6451 - accuracy: 0.6440 - f1_score: nan - val_loss: 0.6288 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 5/100\n",
            "54/54 [==============================] - 46s 846ms/step - loss: 0.6095 - accuracy: 0.6608 - f1_score: nan - val_loss: 0.6033 - val_accuracy: 0.6316 - val_f1_score: nan\n",
            "Epoch 6/100\n",
            "54/54 [==============================] - 46s 845ms/step - loss: 0.6121 - accuracy: 0.6570 - f1_score: nan - val_loss: 0.5930 - val_accuracy: 0.6491 - val_f1_score: nan\n",
            "Epoch 7/100\n",
            "54/54 [==============================] - 46s 845ms/step - loss: 0.5978 - accuracy: 0.6664 - f1_score: nan - val_loss: 0.5853 - val_accuracy: 0.6667 - val_f1_score: 0.4325\n",
            "Epoch 8/100\n",
            "54/54 [==============================] - 46s 846ms/step - loss: 0.6203 - accuracy: 0.6580 - f1_score: nan - val_loss: 0.5864 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 9/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.5893 - accuracy: 0.6692 - f1_score: nan - val_loss: 0.5854 - val_accuracy: 0.6667 - val_f1_score: 0.2551\n",
            "Epoch 10/100\n",
            "54/54 [==============================] - 46s 847ms/step - loss: 0.5641 - accuracy: 0.6775 - f1_score: nan - val_loss: 0.5558 - val_accuracy: 0.6667 - val_f1_score: 0.3148\n",
            "Epoch 11/100\n",
            "54/54 [==============================] - 46s 847ms/step - loss: 0.5570 - accuracy: 0.6822 - f1_score: nan - val_loss: 0.5553 - val_accuracy: 0.6667 - val_f1_score: 0.4037\n",
            "Epoch 12/100\n",
            "54/54 [==============================] - 46s 847ms/step - loss: 0.5620 - accuracy: 0.6813 - f1_score: nan - val_loss: 0.6248 - val_accuracy: 0.6491 - val_f1_score: nan\n",
            "Epoch 13/100\n",
            "54/54 [==============================] - 46s 847ms/step - loss: 0.5423 - accuracy: 0.6869 - f1_score: nan - val_loss: 0.5439 - val_accuracy: 0.7544 - val_f1_score: 0.6148\n",
            "Epoch 14/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.5273 - accuracy: 0.7176 - f1_score: nan - val_loss: 0.5525 - val_accuracy: 0.7193 - val_f1_score: 0.6698\n",
            "Epoch 15/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.5625 - accuracy: 0.6887 - f1_score: nan - val_loss: 0.5588 - val_accuracy: 0.6491 - val_f1_score: 0.4177\n",
            "Epoch 16/100\n",
            "54/54 [==============================] - 46s 847ms/step - loss: 0.5227 - accuracy: 0.7232 - f1_score: nan - val_loss: 0.5529 - val_accuracy: 0.7368 - val_f1_score: 0.6027\n",
            "Epoch 17/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.5152 - accuracy: 0.7279 - f1_score: nan - val_loss: 0.5548 - val_accuracy: 0.6842 - val_f1_score: 0.4984\n",
            "Epoch 18/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.5181 - accuracy: 0.7092 - f1_score: nan - val_loss: 0.5691 - val_accuracy: 0.6491 - val_f1_score: 0.3889\n",
            "Epoch 19/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.5343 - accuracy: 0.7074 - f1_score: nan - val_loss: 0.5406 - val_accuracy: 0.7719 - val_f1_score: 0.6897\n",
            "Epoch 20/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.4974 - accuracy: 0.7400 - f1_score: 0.5740 - val_loss: 0.5400 - val_accuracy: 0.7193 - val_f1_score: 0.5974\n",
            "Epoch 21/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.5027 - accuracy: 0.7232 - f1_score: nan - val_loss: 0.6178 - val_accuracy: 0.6842 - val_f1_score: 0.4464\n",
            "Epoch 22/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.5080 - accuracy: 0.7251 - f1_score: nan - val_loss: 0.5895 - val_accuracy: 0.6842 - val_f1_score: 0.5114\n",
            "Epoch 23/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.4967 - accuracy: 0.7418 - f1_score: nan - val_loss: 0.5520 - val_accuracy: 0.7368 - val_f1_score: 0.6951\n",
            "Epoch 24/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.4945 - accuracy: 0.7232 - f1_score: 0.5685 - val_loss: 0.5658 - val_accuracy: 0.7368 - val_f1_score: 0.6027\n",
            "Epoch 25/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.4876 - accuracy: 0.7521 - f1_score: nan - val_loss: 0.5332 - val_accuracy: 0.7544 - val_f1_score: 0.6687\n",
            "Epoch 26/100\n",
            "54/54 [==============================] - 46s 851ms/step - loss: 0.4654 - accuracy: 0.7605 - f1_score: 0.6324 - val_loss: 0.5952 - val_accuracy: 0.6667 - val_f1_score: 0.4568\n",
            "Epoch 27/100\n",
            "54/54 [==============================] - 46s 850ms/step - loss: 0.4636 - accuracy: 0.7390 - f1_score: 0.5814 - val_loss: 0.5429 - val_accuracy: 0.7193 - val_f1_score: 0.6217\n",
            "Epoch 28/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.4659 - accuracy: 0.7474 - f1_score: nan - val_loss: 0.5572 - val_accuracy: 0.7193 - val_f1_score: 0.6436\n",
            "Epoch 29/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.4742 - accuracy: 0.7521 - f1_score: nan - val_loss: 0.5934 - val_accuracy: 0.6491 - val_f1_score: 0.3613\n",
            "Epoch 30/100\n",
            "54/54 [==============================] - 46s 850ms/step - loss: 0.4539 - accuracy: 0.7670 - f1_score: 0.6372 - val_loss: 0.6181 - val_accuracy: 0.6842 - val_f1_score: 0.4464\n",
            "Epoch 31/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.4366 - accuracy: 0.7698 - f1_score: nan - val_loss: 0.5973 - val_accuracy: 0.7193 - val_f1_score: 0.6436\n",
            "Epoch 32/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.4472 - accuracy: 0.7726 - f1_score: nan - val_loss: 0.5536 - val_accuracy: 0.7368 - val_f1_score: 0.6893\n",
            "Epoch 33/100\n",
            "54/54 [==============================] - 46s 850ms/step - loss: 0.4312 - accuracy: 0.7856 - f1_score: nan - val_loss: 0.6441 - val_accuracy: 0.6667 - val_f1_score: 0.4568\n",
            "Epoch 34/100\n",
            "54/54 [==============================] - 46s 850ms/step - loss: 0.4512 - accuracy: 0.7642 - f1_score: 0.6166 - val_loss: 0.5731 - val_accuracy: 0.7193 - val_f1_score: 0.6217\n",
            "Epoch 35/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.4231 - accuracy: 0.7922 - f1_score: 0.6744 - val_loss: 0.5952 - val_accuracy: 0.7018 - val_f1_score: 0.5768\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8qVUA8jD2ZWb",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 354
        },
        "outputId": "df8aa70e-19b6-4f70-8028-bcf98d0f5ff1"
      },
      "source": [
        "p = np.array(model3.predict(train_axial))\n",
        "train_Ab_pred.append(p)\n",
        "p = np.array(model3.predict(test_axial))\n",
        "test_Ab_pred.append(p)               \n",
        "score3 = model3.evaluate(test_axial ,testAb_labels,verbose=1)       \n",
        "print('Test loss:',score3[0])\n",
        "print('Test accuracy:',score3[1])                                                           \n",
        "plt.figure()\n",
        "plt.grid()                                                                                                  \n",
        "plt.title('Training performance')                          \n",
        "plt.plot(history3.epoch, history3.history['loss'], label='train loss+error')  \n",
        "plt.plot(history3.epoch, history3.history['val_loss'], label='val_error') \n",
        "plt.plot(history3.epoch, history3.history['f1_score'], label='f1_score') \n",
        "plt.legend()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "4/4 [==============================] - 1s 359ms/step - loss: 0.7807 - accuracy: 0.6167 - f1_score: 0.6199\n",
            "Test loss: 0.7807239294052124\n",
            "Test accuracy: 0.6166666746139526\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7fe3be72d3d0>"
            ]
          },
          "metadata": {},
          "execution_count": 29
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "filename = '/content/drive/MyDrive/Medical/MRNET MODEL/VGGmodel3.h5'\n",
        "model3.save(filename)\n",
        "print('>Saved %s' % filename)\n",
        "clean_up(model=model3)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "TDjJ7rluf2BR",
        "outputId": "57fc60d5-152b-4604-d1f6-1cb32738dfed"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            ">Saved /content/drive/MyDrive/Medical/MRNET MODEL/VGGmodel3.h5\n",
            "...Finished clean up...\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "##### Deleting Files which wont be required further to reduce memory load"
      ],
      "metadata": {
        "id": "0vgZl5bxuw4a"
      }
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "K-U-NrNAC3Gl"
      },
      "source": [
        "del train_axial\n",
        "del test_axial\n",
        "del axial_dir\n",
        "del axial_reloaded\n",
        "del axialTest_reloaded\n",
        "del axial_list\n",
        "del axialTest_list\n",
        "del model\n",
        "del model2\n",
        "del model3\n",
        "del base_model_ax_men\n",
        "del base_model_ax_ACL\n",
        "del base_model_ax_AB\n",
        "del p"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gvtnuUab3sGt"
      },
      "source": [
        "##Coronal Plane Scans"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VJKP50dP-T7d"
      },
      "source": [
        "###Reading train & test data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sL-8TZ2J36tA",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "6028f26d-8cb9-4eae-c2c0-a0ef3cb6f46a"
      },
      "source": [
        "coronal_dir = os.listdir(\"/content/drive/MyDrive/Medical/MRNet-v1.0/train/coronal/\")\n",
        "coronal_dir = [f.lower() for f in coronal_dir]   \n",
        "c=sorted(coronal_dir)\n",
        "coronal_list=np.array(c)\n",
        "coronal_reloaded=[]\n",
        "print(coronal_list.shape)\n",
        "for j in range(1,1131):\n",
        "    file2= \"/content/drive/MyDrive/Medical/MRNet-v1.0/train/coronal/\"+coronal_list[j]\n",
        "    coronal_reloaded.append(np.load(file2,allow_pickle='false'))\n",
        "\n",
        "coronal_reloaded=np.array(coronal_reloaded)\n",
        "print(coronal_reloaded.shape)\n",
        "train_coronal=[]\n",
        " \n",
        "for j in range(len(coronal_reloaded)):\n",
        "    mid2=math.floor((coronal_reloaded[j].shape[0])/2)\n",
        "    train_coronal.append([])\n",
        "        \n",
        "    for n in range(mid2-1,mid2+2):\n",
        "         train_coronal[j].append(coronal_reloaded[j][n])\n",
        "\n",
        "train_coronal=np.array(train_coronal)\n",
        "train_coronal=np.reshape(train_coronal,(1130,256,256,3))\n",
        "print(train_coronal.shape)\n",
        "\n",
        "coronal_dir = os.listdir(\"/content/drive/MyDrive/Medical/MRNet-v1.0/valid/coronal/\")\n",
        "coronal_dir = [f.lower() for f in coronal_dir]   \n",
        "c=sorted(coronal_dir)\n",
        "coronalTest_list=np.array(c)\n",
        "coronalTest_reloaded=[]\n",
        "print(coronalTest_list.shape)\n",
        "for j in range(0,120):\n",
        "    file2= \"/content/drive/MyDrive/Medical/MRNet-v1.0/valid/coronal/\"+coronalTest_list[j]\n",
        "    coronalTest_reloaded.append(np.load(file2,allow_pickle='false'))\n",
        "\n",
        "coronalTest_reloaded=np.array(coronalTest_reloaded)\n",
        "print(coronalTest_reloaded.shape)\n",
        "test_coronal=[]\n",
        " \n",
        "for j in range(len(coronalTest_reloaded)):\n",
        "    mid2=math.floor((coronalTest_reloaded[j].shape[0])/2)\n",
        "    test_coronal.append([])\n",
        "        \n",
        "    for n in range(mid2-1,mid2+2):\n",
        "         test_coronal[j].append(coronalTest_reloaded[j][n])\n",
        "\n",
        "test_coronal=np.array(test_coronal)\n",
        "test_coronal=np.reshape(test_coronal,(120,256,256,3))\n",
        "print(test_coronal.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(1131,)\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:12: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray.\n",
            "  if sys.path[0] == '':\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(1130,)\n",
            "(1130, 256, 256, 3)\n",
            "(120,)\n",
            "(120,)\n",
            "(120, 256, 256, 3)\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:40: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "RM003h3Y46L-"
      },
      "source": [
        "#### Finetuning VGG16 Model for Coronal Plane Scanes checking for Meniscus Condition"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "N-4cC08a50ot"
      },
      "source": [
        "input_tensor = Input(shape=(256, 256,3))  \n",
        "base_model_Co_Men = VGG16(input_tensor=input_tensor, weights='imagenet', include_top=False)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HRaoDsnw57HB"
      },
      "source": [
        "x = base_model_Co_Men.output\n",
        "x = GlobalAveragePooling2D()(x)\n",
        "x=Dropout(0.2)(x)\n",
        "x = Dense(1024, activation='relu')(x)\n",
        "x=Dropout(0.2)(x)\n",
        "x=Dropout(0.2)(x)\n",
        "predictions = Dense(1, activation='sigmoid')(x)\n",
        "model4 = Model(inputs=base_model_Co_Men.input,outputs=predictions)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Xa1OIRRw5_q-"
      },
      "source": [
        "for layer in model4.layers[:15]:\n",
        "   layer.trainable = True\n",
        "for layer in model4.layers[15:]:\n",
        "  layer.trainable = False"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0sTn67eX6GyY",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "ccd9a65b-f379-411d-854d-c1794db21030"
      },
      "source": [
        "early_stopping_monitor = EarlyStopping(patience=5)    \n",
        "model4.compile(optimizer=SGD(lr=0.001,momentum=0.9), loss='binary_crossentropy',metrics=['accuracy',f1_score])\n",
        "history4=model4.fit(train_coronal ,trainMen_labels, validation_split=0.05,shuffle=True ,epochs=100, callbacks=[early_stopping_monitor], batch_size = 20)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/keras/optimizer_v2/gradient_descent.py:102: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.\n",
            "  super(SGD, self).__init__(name, **kwargs)\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/100\n",
            "54/54 [==============================] - 48s 855ms/step - loss: 0.9005 - accuracy: 0.5461 - f1_score: nan - val_loss: 0.6736 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 2/100\n",
            "54/54 [==============================] - 46s 843ms/step - loss: 0.6630 - accuracy: 0.6067 - f1_score: nan - val_loss: 0.6661 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 3/100\n",
            "54/54 [==============================] - 45s 842ms/step - loss: 0.6666 - accuracy: 0.6337 - f1_score: nan - val_loss: 0.6613 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 4/100\n",
            "54/54 [==============================] - 45s 842ms/step - loss: 0.6562 - accuracy: 0.6449 - f1_score: nan - val_loss: 0.6510 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 5/100\n",
            "54/54 [==============================] - 45s 842ms/step - loss: 0.6405 - accuracy: 0.6347 - f1_score: nan - val_loss: 0.6559 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 6/100\n",
            "54/54 [==============================] - 45s 841ms/step - loss: 0.6465 - accuracy: 0.6459 - f1_score: nan - val_loss: 0.6409 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 7/100\n",
            "54/54 [==============================] - 45s 840ms/step - loss: 0.6457 - accuracy: 0.6272 - f1_score: nan - val_loss: 0.6400 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 8/100\n",
            "54/54 [==============================] - 45s 842ms/step - loss: 0.6327 - accuracy: 0.6514 - f1_score: nan - val_loss: 0.6327 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 9/100\n",
            "54/54 [==============================] - 45s 841ms/step - loss: 0.6323 - accuracy: 0.6431 - f1_score: nan - val_loss: 0.6346 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 10/100\n",
            "54/54 [==============================] - 45s 842ms/step - loss: 0.6356 - accuracy: 0.6449 - f1_score: nan - val_loss: 0.6344 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 11/100\n",
            "54/54 [==============================] - 46s 846ms/step - loss: 0.6307 - accuracy: 0.6449 - f1_score: nan - val_loss: 0.6373 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 12/100\n",
            "54/54 [==============================] - 46s 847ms/step - loss: 0.6350 - accuracy: 0.6309 - f1_score: nan - val_loss: 0.6313 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 13/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.6296 - accuracy: 0.6533 - f1_score: nan - val_loss: 0.6291 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 14/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.6490 - accuracy: 0.6477 - f1_score: nan - val_loss: 0.6338 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 15/100\n",
            "54/54 [==============================] - 46s 847ms/step - loss: 0.6303 - accuracy: 0.6496 - f1_score: nan - val_loss: 0.6266 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 16/100\n",
            "54/54 [==============================] - 46s 847ms/step - loss: 0.6173 - accuracy: 0.6524 - f1_score: nan - val_loss: 0.6283 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 17/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.6252 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6286 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 18/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.6127 - accuracy: 0.6552 - f1_score: nan - val_loss: 0.6203 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 19/100\n",
            "54/54 [==============================] - 46s 847ms/step - loss: 0.6133 - accuracy: 0.6580 - f1_score: nan - val_loss: 0.6583 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 20/100\n",
            "54/54 [==============================] - 46s 847ms/step - loss: 0.6245 - accuracy: 0.6496 - f1_score: nan - val_loss: 0.6144 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 21/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.6114 - accuracy: 0.6533 - f1_score: nan - val_loss: 0.6255 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 22/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.6103 - accuracy: 0.6533 - f1_score: nan - val_loss: 0.6194 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 23/100\n",
            "54/54 [==============================] - 46s 850ms/step - loss: 0.6062 - accuracy: 0.6542 - f1_score: nan - val_loss: 0.6159 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 24/100\n",
            "54/54 [==============================] - 46s 851ms/step - loss: 0.5998 - accuracy: 0.6570 - f1_score: nan - val_loss: 0.6072 - val_accuracy: 0.5965 - val_f1_score: nan\n",
            "Epoch 25/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.5991 - accuracy: 0.6692 - f1_score: nan - val_loss: 0.6094 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 26/100\n",
            "54/54 [==============================] - 46s 846ms/step - loss: 0.5982 - accuracy: 0.6552 - f1_score: nan - val_loss: 0.6066 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 27/100\n",
            "54/54 [==============================] - 46s 844ms/step - loss: 0.5867 - accuracy: 0.6524 - f1_score: nan - val_loss: 0.6026 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 28/100\n",
            "54/54 [==============================] - 46s 843ms/step - loss: 0.5913 - accuracy: 0.6542 - f1_score: nan - val_loss: 0.6064 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 29/100\n",
            "54/54 [==============================] - 46s 844ms/step - loss: 0.5862 - accuracy: 0.6598 - f1_score: nan - val_loss: 0.5928 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 30/100\n",
            "54/54 [==============================] - 46s 845ms/step - loss: 0.5710 - accuracy: 0.6561 - f1_score: nan - val_loss: 0.6352 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 31/100\n",
            "54/54 [==============================] - 46s 845ms/step - loss: 0.6053 - accuracy: 0.6608 - f1_score: nan - val_loss: 0.6118 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 32/100\n",
            "54/54 [==============================] - 46s 844ms/step - loss: 0.5884 - accuracy: 0.6636 - f1_score: nan - val_loss: 0.6066 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 33/100\n",
            "54/54 [==============================] - 46s 844ms/step - loss: 0.5804 - accuracy: 0.6608 - f1_score: nan - val_loss: 0.5864 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 34/100\n",
            "54/54 [==============================] - 46s 844ms/step - loss: 0.5733 - accuracy: 0.6552 - f1_score: nan - val_loss: 0.6085 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 35/100\n",
            "54/54 [==============================] - 46s 844ms/step - loss: 0.5614 - accuracy: 0.6589 - f1_score: nan - val_loss: 0.5964 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 36/100\n",
            "54/54 [==============================] - 46s 844ms/step - loss: 0.5614 - accuracy: 0.6757 - f1_score: nan - val_loss: 0.6132 - val_accuracy: 0.6316 - val_f1_score: nan\n",
            "Epoch 37/100\n",
            "54/54 [==============================] - 46s 846ms/step - loss: 0.5595 - accuracy: 0.6608 - f1_score: nan - val_loss: 0.5941 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 38/100\n",
            "54/54 [==============================] - 46s 847ms/step - loss: 0.5627 - accuracy: 0.6673 - f1_score: nan - val_loss: 0.6335 - val_accuracy: 0.6140 - val_f1_score: nan\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EsGIEL8-jvLQ",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 354
        },
        "outputId": "1c1abbb4-055f-406b-8c23-9d94e32113d8"
      },
      "source": [
        "p = np.array(model4.predict(train_coronal))\n",
        "train_Men_pred.append(p)\n",
        "p = np.array(model4.predict(test_coronal))\n",
        "test_Men_pred.append(p)\n",
        "score4 = model4.evaluate(test_coronal ,testMen_labels,verbose=1)       \n",
        "print('Test loss:',score4[0])\n",
        "print('Test accuracy:',score4[1])                                              \n",
        "plt.figure()\n",
        "plt.grid()                                                                                                   \n",
        "plt.title('Training performance')                          \n",
        "plt.plot(history4.epoch, history4.history['loss'], label='train loss+error')  \n",
        "plt.plot(history4.epoch, history4.history['val_loss'], label='val_error')  \n",
        "plt.plot(history4.epoch, history4.history['f1_score'], label='f1_score')  \n",
        "plt.legend()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "4/4 [==============================] - 1s 359ms/step - loss: 0.6888 - accuracy: 0.5667 - f1_score: nan\n",
            "Test loss: 0.6887705326080322\n",
            "Test accuracy: 0.5666666626930237\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7fe3be34f2d0>"
            ]
          },
          "metadata": {},
          "execution_count": 37
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "filename = '/content/drive/MyDrive/Medical/MRNET MODEL/VGGmodel4.h5'\n",
        "model4.save(filename)\n",
        "print('>Saved %s' % filename)\n",
        "clean_up(model=model4)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "O2yQjTejgNn6",
        "outputId": "9fe02fc8-7b97-4202-925c-b59f858b4a77"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            ">Saved /content/drive/MyDrive/Medical/MRNET MODEL/VGGmodel4.h5\n",
            "...Finished clean up...\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ev8SCy58luHL"
      },
      "source": [
        "#### Finetuning VGG16 Model for Coronal Plane Scanes checking for ACL Condition"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NqYG7kydluHO"
      },
      "source": [
        "input_tensor = Input(shape=(256, 256,3))  \n",
        "base_model_Co_ACL = VGG16(input_tensor=input_tensor, weights='imagenet', include_top=False)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uhIvHBiIluHY"
      },
      "source": [
        "x = base_model_Co_ACL.output\n",
        "x = GlobalAveragePooling2D()(x)\n",
        "x=Dropout(0.2)(x)\n",
        "x = Dense(1024, activation='relu')(x)\n",
        "x=Dropout(0.2)(x)\n",
        "x=Dropout(0.2)(x)\n",
        "predictions = Dense(1, activation='sigmoid')(x)\n",
        "model5 = Model(inputs=base_model_Co_ACL.input,outputs=predictions)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5fPQriI-luHh"
      },
      "source": [
        "for layer in model5.layers[:15]:\n",
        "   layer.trainable = True\n",
        "for layer in model5.layers[15:]:\n",
        "  layer.trainable = False"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GjmEc-AkluHx",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "40ea4e26-d62b-45f6-a8d6-d779dd9059fb"
      },
      "source": [
        "early_stopping_monitor = EarlyStopping(patience=10)    \n",
        "model5.compile(optimizer=SGD(lr=0.001,momentum=0.9), loss='binary_crossentropy',metrics=['accuracy',f1_score])\n",
        "history5=model5.fit(train_coronal ,trainACL_labels, validation_split=0.05,shuffle=True ,epochs=100, callbacks=[early_stopping_monitor], batch_size = 20)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/100\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/keras/optimizer_v2/gradient_descent.py:102: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.\n",
            "  super(SGD, self).__init__(name, **kwargs)\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "54/54 [==============================] - 47s 854ms/step - loss: 0.6111 - accuracy: 0.7791 - f1_score: nan - val_loss: 0.4777 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 2/100\n",
            "54/54 [==============================] - 46s 845ms/step - loss: 0.5306 - accuracy: 0.8080 - f1_score: nan - val_loss: 0.4525 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 3/100\n",
            "54/54 [==============================] - 46s 845ms/step - loss: 0.5076 - accuracy: 0.8155 - f1_score: nan - val_loss: 0.4463 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 4/100\n",
            "54/54 [==============================] - 46s 845ms/step - loss: 0.4978 - accuracy: 0.8108 - f1_score: nan - val_loss: 0.4505 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 5/100\n",
            "54/54 [==============================] - 46s 845ms/step - loss: 0.4946 - accuracy: 0.8145 - f1_score: nan - val_loss: 0.4839 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 6/100\n",
            "54/54 [==============================] - 46s 845ms/step - loss: 0.4958 - accuracy: 0.8145 - f1_score: nan - val_loss: 0.4379 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 7/100\n",
            "54/54 [==============================] - 46s 847ms/step - loss: 0.4778 - accuracy: 0.8145 - f1_score: nan - val_loss: 0.4443 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 8/100\n",
            "54/54 [==============================] - 46s 847ms/step - loss: 0.4748 - accuracy: 0.8145 - f1_score: nan - val_loss: 0.4541 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 9/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.4844 - accuracy: 0.8145 - f1_score: nan - val_loss: 0.4341 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 10/100\n",
            "54/54 [==============================] - 46s 847ms/step - loss: 0.4747 - accuracy: 0.8145 - f1_score: nan - val_loss: 0.4398 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 11/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.4678 - accuracy: 0.8145 - f1_score: nan - val_loss: 0.4505 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 12/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.4595 - accuracy: 0.8145 - f1_score: nan - val_loss: 0.4296 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 13/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.4500 - accuracy: 0.8145 - f1_score: nan - val_loss: 0.4313 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 14/100\n",
            "54/54 [==============================] - 46s 850ms/step - loss: 0.4474 - accuracy: 0.8155 - f1_score: nan - val_loss: 0.4151 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 15/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.4385 - accuracy: 0.8145 - f1_score: nan - val_loss: 0.4272 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 16/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.4434 - accuracy: 0.8136 - f1_score: nan - val_loss: 0.4325 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 17/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.4345 - accuracy: 0.8145 - f1_score: nan - val_loss: 0.4260 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 18/100\n",
            "54/54 [==============================] - 46s 848ms/step - loss: 0.4246 - accuracy: 0.8136 - f1_score: nan - val_loss: 0.4600 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 19/100\n",
            "54/54 [==============================] - 46s 849ms/step - loss: 0.4239 - accuracy: 0.8145 - f1_score: nan - val_loss: 0.4303 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 20/100\n",
            "54/54 [==============================] - 46s 850ms/step - loss: 0.4211 - accuracy: 0.8145 - f1_score: nan - val_loss: 0.4185 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 21/100\n",
            "54/54 [==============================] - 46s 850ms/step - loss: 0.4105 - accuracy: 0.8145 - f1_score: nan - val_loss: 0.4399 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 22/100\n",
            "54/54 [==============================] - 46s 850ms/step - loss: 0.3987 - accuracy: 0.8145 - f1_score: nan - val_loss: 0.4548 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 23/100\n",
            "54/54 [==============================] - 46s 852ms/step - loss: 0.4069 - accuracy: 0.8145 - f1_score: nan - val_loss: 0.4392 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 24/100\n",
            "54/54 [==============================] - 46s 856ms/step - loss: 0.4314 - accuracy: 0.8136 - f1_score: nan - val_loss: 0.4400 - val_accuracy: 0.8421 - val_f1_score: nan\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "61nsbN5vluH8",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "95b2eb05-ecc9-4e48-aa70-f8e3701128ff"
      },
      "source": [
        "p = np.array(model5.predict(train_coronal))\n",
        "train_ACL_pred.append(p)\n",
        "p = np.array(model5.predict(test_coronal, verbose=1))\n",
        "test_ACL_pred.append(p)                     \n",
        "score5 = model5.evaluate(test_coronal ,testACL_labels,verbose=1)       \n",
        "print('Test loss:',score5[0])\n",
        "print('Test accuracy:',score5[1])                                                               \n",
        "plt.figure()\n",
        "plt.grid()                                                                                                        \n",
        "plt.title('Training performance')                          \n",
        "plt.plot(history5.epoch, history5.history['loss'], label='train loss+error')  \n",
        "plt.plot(history5.epoch, history5.history['val_loss'], label='val_error') \n",
        "plt.plot(history5.epoch, history5.history['f1_score'], label='f1_score') \n",
        "plt.legend()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "4/4 [==============================] - 1s 371ms/step\n",
            "4/4 [==============================] - 1s 360ms/step - loss: 0.6721 - accuracy: 0.5500 - f1_score: nan\n",
            "Test loss: 0.6721179485321045\n",
            "Test accuracy: 0.550000011920929\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7fe3bddb6d50>"
            ]
          },
          "metadata": {},
          "execution_count": 43
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "filename = '/content/drive/MyDrive/Medical/MRNET MODEL/VGGmodel5.h5'\n",
        "model5.save(filename)\n",
        "print('>Saved %s' % filename)\n",
        "clean_up(model=model5)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "5rNQtSgkjHtR",
        "outputId": "15debfae-0d52-4c4d-800f-bd8b27818265"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            ">Saved /content/drive/MyDrive/Medical/MRNET MODEL/VGGmodel5.h5\n",
            "...Finished clean up...\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "p4460kOMm_LA"
      },
      "source": [
        "#### Finetuning VGG16 Model for Coronal Plane Scanes checking for Abnormal Condition"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5mMGdm5dm_LC"
      },
      "source": [
        "input_tensor = Input(shape=(256, 256,3))  \n",
        "base_model_Co_Ab = VGG16(input_tensor=input_tensor, weights='imagenet', include_top=False)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wXUGXsfLm_LF"
      },
      "source": [
        "x = base_model_Co_Ab.output\n",
        "x = GlobalAveragePooling2D()(x)\n",
        "x=Dropout(0.2)(x)\n",
        "x = Dense(1024, activation='relu')(x)\n",
        "x=Dropout(0.2)(x)\n",
        "x=Dropout(0.2)(x)\n",
        "predictions = Dense(1, activation='sigmoid')(x)\n",
        "model6 = Model(inputs=base_model_Co_Ab.input,outputs=predictions)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "b9YffZeim_LH"
      },
      "source": [
        "for layer in model6.layers[:15]:\n",
        "   layer.trainable = True\n",
        "for layer in model6.layers[15:]:\n",
        "  layer.trainable = False"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "J_O8H3RXm_LK",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "9301c714-fc9d-48c2-ef1f-a9d5ed8a80f5"
      },
      "source": [
        "early_stopping_monitor = EarlyStopping(patience=10)    \n",
        "model6.compile(optimizer=SGD(lr=0.001,momentum=0.9), loss='binary_crossentropy',metrics=['accuracy',f1_score])\n",
        "history6=model6.fit(train_coronal ,trainAb_labels, validation_split=0.05,shuffle=True ,epochs=100, callbacks=[early_stopping_monitor], batch_size = 20)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/100\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/keras/optimizer_v2/gradient_descent.py:102: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.\n",
            "  super(SGD, self).__init__(name, **kwargs)\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "54/54 [==============================] - 48s 867ms/step - loss: 0.7002 - accuracy: 0.6309 - f1_score: nan - val_loss: 0.6740 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 2/100\n",
            "54/54 [==============================] - 46s 859ms/step - loss: 0.6505 - accuracy: 0.6384 - f1_score: nan - val_loss: 0.6678 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 3/100\n",
            "54/54 [==============================] - 46s 859ms/step - loss: 0.6413 - accuracy: 0.6319 - f1_score: nan - val_loss: 0.6501 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 4/100\n",
            "54/54 [==============================] - 46s 857ms/step - loss: 0.6420 - accuracy: 0.6449 - f1_score: nan - val_loss: 0.6492 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 5/100\n",
            "54/54 [==============================] - 46s 850ms/step - loss: 0.6135 - accuracy: 0.6486 - f1_score: nan - val_loss: 0.6433 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 6/100\n",
            "54/54 [==============================] - 46s 850ms/step - loss: 0.6111 - accuracy: 0.6449 - f1_score: nan - val_loss: 0.6664 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 7/100\n",
            "54/54 [==============================] - 46s 850ms/step - loss: 0.6259 - accuracy: 0.6524 - f1_score: nan - val_loss: 0.6764 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 8/100\n",
            "54/54 [==============================] - 46s 854ms/step - loss: 0.6233 - accuracy: 0.6459 - f1_score: nan - val_loss: 0.6329 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 9/100\n",
            "54/54 [==============================] - 46s 858ms/step - loss: 0.6042 - accuracy: 0.6468 - f1_score: nan - val_loss: 0.6322 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 10/100\n",
            "54/54 [==============================] - 46s 858ms/step - loss: 0.5919 - accuracy: 0.6542 - f1_score: nan - val_loss: 0.6269 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 11/100\n",
            "54/54 [==============================] - 46s 851ms/step - loss: 0.5739 - accuracy: 0.6468 - f1_score: nan - val_loss: 0.6273 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 12/100\n",
            "54/54 [==============================] - 46s 851ms/step - loss: 0.5742 - accuracy: 0.6496 - f1_score: nan - val_loss: 0.6093 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 13/100\n",
            "54/54 [==============================] - 46s 850ms/step - loss: 0.5752 - accuracy: 0.6524 - f1_score: nan - val_loss: 0.6074 - val_accuracy: 0.5965 - val_f1_score: nan\n",
            "Epoch 14/100\n",
            "54/54 [==============================] - 46s 850ms/step - loss: 0.6047 - accuracy: 0.6486 - f1_score: nan - val_loss: 0.6233 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 15/100\n",
            "54/54 [==============================] - 46s 850ms/step - loss: 0.5833 - accuracy: 0.6552 - f1_score: nan - val_loss: 0.6097 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 16/100\n",
            "54/54 [==============================] - 46s 850ms/step - loss: 0.5674 - accuracy: 0.6468 - f1_score: nan - val_loss: 0.6170 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 17/100\n",
            "54/54 [==============================] - 46s 850ms/step - loss: 0.5655 - accuracy: 0.6542 - f1_score: nan - val_loss: 0.6109 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 18/100\n",
            "54/54 [==============================] - 46s 851ms/step - loss: 0.5617 - accuracy: 0.6542 - f1_score: nan - val_loss: 0.6038 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 19/100\n",
            "54/54 [==============================] - 46s 851ms/step - loss: 0.5627 - accuracy: 0.6570 - f1_score: nan - val_loss: 0.5945 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 20/100\n",
            "54/54 [==============================] - 46s 852ms/step - loss: 0.5315 - accuracy: 0.6542 - f1_score: nan - val_loss: 0.6139 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 21/100\n",
            "54/54 [==============================] - 46s 851ms/step - loss: 0.5379 - accuracy: 0.6561 - f1_score: nan - val_loss: 0.5884 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 22/100\n",
            "54/54 [==============================] - 46s 852ms/step - loss: 0.5325 - accuracy: 0.6542 - f1_score: nan - val_loss: 0.5890 - val_accuracy: 0.5789 - val_f1_score: nan\n",
            "Epoch 23/100\n",
            "54/54 [==============================] - 46s 852ms/step - loss: 0.5191 - accuracy: 0.6617 - f1_score: nan - val_loss: 0.6303 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 24/100\n",
            "54/54 [==============================] - 46s 852ms/step - loss: 0.5154 - accuracy: 0.6729 - f1_score: nan - val_loss: 0.5754 - val_accuracy: 0.6667 - val_f1_score: 0.3106\n",
            "Epoch 25/100\n",
            "54/54 [==============================] - 46s 852ms/step - loss: 0.5023 - accuracy: 0.6710 - f1_score: nan - val_loss: 0.6026 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 26/100\n",
            "54/54 [==============================] - 46s 855ms/step - loss: 0.5191 - accuracy: 0.6831 - f1_score: nan - val_loss: 0.7519 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 27/100\n",
            "54/54 [==============================] - 46s 857ms/step - loss: 0.5232 - accuracy: 0.6719 - f1_score: nan - val_loss: 0.5946 - val_accuracy: 0.6491 - val_f1_score: 0.2458\n",
            "Epoch 28/100\n",
            "54/54 [==============================] - 46s 857ms/step - loss: 0.5157 - accuracy: 0.7018 - f1_score: nan - val_loss: 0.6019 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 29/100\n",
            "54/54 [==============================] - 46s 857ms/step - loss: 0.4949 - accuracy: 0.7074 - f1_score: nan - val_loss: 0.5743 - val_accuracy: 0.6842 - val_f1_score: 0.4063\n",
            "Epoch 30/100\n",
            "54/54 [==============================] - 46s 858ms/step - loss: 0.4746 - accuracy: 0.7185 - f1_score: nan - val_loss: 0.6129 - val_accuracy: 0.6842 - val_f1_score: 0.5628\n",
            "Epoch 31/100\n",
            "54/54 [==============================] - 46s 858ms/step - loss: 0.4717 - accuracy: 0.7213 - f1_score: nan - val_loss: 0.6309 - val_accuracy: 0.7193 - val_f1_score: 0.5383\n",
            "Epoch 32/100\n",
            "54/54 [==============================] - 46s 859ms/step - loss: 0.4634 - accuracy: 0.7288 - f1_score: nan - val_loss: 0.6116 - val_accuracy: 0.6316 - val_f1_score: 0.4125\n",
            "Epoch 33/100\n",
            "54/54 [==============================] - 46s 861ms/step - loss: 0.4644 - accuracy: 0.7428 - f1_score: nan - val_loss: 0.5933 - val_accuracy: 0.7368 - val_f1_score: 0.6453\n",
            "Epoch 34/100\n",
            "54/54 [==============================] - 47s 862ms/step - loss: 0.4568 - accuracy: 0.7353 - f1_score: nan - val_loss: 0.6546 - val_accuracy: 0.6842 - val_f1_score: 0.6480\n",
            "Epoch 35/100\n",
            "54/54 [==============================] - 47s 862ms/step - loss: 0.4582 - accuracy: 0.7549 - f1_score: nan - val_loss: 0.6237 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 36/100\n",
            "54/54 [==============================] - 47s 862ms/step - loss: 0.4379 - accuracy: 0.7493 - f1_score: nan - val_loss: 0.7700 - val_accuracy: 0.6316 - val_f1_score: 0.4027\n",
            "Epoch 37/100\n",
            "54/54 [==============================] - 47s 861ms/step - loss: 0.4625 - accuracy: 0.7446 - f1_score: nan - val_loss: 0.6846 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 38/100\n",
            "54/54 [==============================] - 46s 859ms/step - loss: 0.4592 - accuracy: 0.7679 - f1_score: nan - val_loss: 0.6222 - val_accuracy: 0.6316 - val_f1_score: 0.4397\n",
            "Epoch 39/100\n",
            "54/54 [==============================] - 46s 855ms/step - loss: 0.4166 - accuracy: 0.7661 - f1_score: 0.6020 - val_loss: 0.6038 - val_accuracy: 0.6667 - val_f1_score: 0.6248\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6Rn0iI-Rm_LQ",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "3c832e64-8357-43bb-9068-5b68c8a8b249"
      },
      "source": [
        "p = np.array(model6.predict(train_coronal))\n",
        "train_Ab_pred.append(p)\n",
        "p = np.array(model6.predict(test_coronal))\n",
        "test_Ab_pred.append(p)              \n",
        "score6 = model6.evaluate(test_coronal ,testAb_labels,verbose=1)       \n",
        "print('Test loss:',score6[0])\n",
        "print('Test accuracy:',score6[1])                                                          \n",
        "plt.figure()\n",
        "plt.grid()                                                                                                          \n",
        "plt.title('Training performance')                          \n",
        "plt.plot(history6.epoch, history6.history['loss'], label='train loss+error')  \n",
        "plt.plot(history6.epoch, history6.history['val_loss'], label='val_error')   \n",
        "plt.plot(history6.epoch, history6.history['f1_score'], label='f1_score')   \n",
        "plt.legend()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "4/4 [==============================] - 1s 358ms/step - loss: 0.8803 - accuracy: 0.6250 - f1_score: 0.6324\n",
            "Test loss: 0.8803246021270752\n",
            "Test accuracy: 0.625\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7fe3bc644610>"
            ]
          },
          "metadata": {},
          "execution_count": 49
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "filename = '/content/drive/MyDrive/Medical/MRNET MODEL/VGGmodel6.h5'\n",
        "model6.save(filename)\n",
        "print('>Saved %s' % filename)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "y5hC24kdBJHp",
        "outputId": "1e2f41ad-33a1-4335-ec94-7733eb5b8019"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            ">Saved /content/drive/MyDrive/Medical/MRNET MODEL/VGGmodel6.h5\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "clean_up(model=model6)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "iUK2S_nCHJgc",
        "outputId": "73175d4e-21ef-481f-c348-24419f0c462c"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "...Finished clean up...\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QPaXpEFeEHRK"
      },
      "source": [
        "del train_coronal\n",
        "del test_coronal\n",
        "del coronal_dir\n",
        "del coronal_reloaded\n",
        "del coronalTest_reloaded\n",
        "del coronal_list\n",
        "del coronalTest_list\n",
        "del model4\n",
        "del model5\n",
        "del model6\n",
        "del base_model_Co_Men\n",
        "del base_model_Co_ACL\n",
        "del base_model_Co_Ab\n",
        "del p"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "eh7IUVUIxzFM"
      },
      "source": [
        "##Sagittal Planes Scans"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "f-j4PHNJEMRA"
      },
      "source": [
        "###Reading train & test data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Uqm-RJB1xzFQ",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "d142d36e-655c-49e6-af22-2aaf88f4382d"
      },
      "source": [
        "sagittal_dir = os.listdir(\"/content/drive/MyDrive/Medical/MRNet-v1.0/train/sagittal/\")\n",
        "sagittal_dir = [f.lower() for f in sagittal_dir]   \n",
        "c=sorted(sagittal_dir)\n",
        "sagittal_list=np.array(c)\n",
        "sagittal_reloaded=[]\n",
        "print(sagittal_list.shape)\n",
        "for j in range(1,1131):\n",
        "    file2= \"/content/drive/MyDrive/Medical/MRNet-v1.0/train/sagittal/\"+sagittal_list[j]\n",
        "    sagittal_reloaded.append(np.load(file2,allow_pickle='false'))\n",
        "sagittal_reloaded=np.array(sagittal_reloaded)\n",
        "print(sagittal_reloaded.shape)\n",
        "train_sagittal=[]\n",
        " \n",
        "for j in range(len(sagittal_reloaded)):\n",
        "    mid2=math.floor((sagittal_reloaded[j].shape[0])/2)\n",
        "    train_sagittal.append([])\n",
        "       \n",
        "    for n in range(mid2-1,mid2+2):\n",
        "         train_sagittal[j].append(sagittal_reloaded[j][n])\n",
        "\n",
        "train_sagittal=np.array(train_sagittal)\n",
        "train_sagittal=np.reshape(train_sagittal,(1130,256,256,3))\n",
        "print(train_sagittal.shape)\n",
        "sagittal_dir = os.listdir(\"/content/drive/MyDrive/Medical/MRNet-v1.0/valid/sagittal/\")\n",
        "sagittal_dir = [f.lower() for f in sagittal_dir]   \n",
        "c=sorted(sagittal_dir)\n",
        "sagittalTest_list=np.array(c)\n",
        "sagittalTest_reloaded=[]\n",
        "print(sagittalTest_list.shape)\n",
        "for j in range(0,120):\n",
        "    file2= \"/content/drive/MyDrive/Medical/MRNet-v1.0/valid/sagittal/\"+sagittalTest_list[j]\n",
        "    sagittalTest_reloaded.append(np.load(file2,allow_pickle='false'))\n",
        "\n",
        "sagittalTest_reloaded=np.array(sagittalTest_reloaded)\n",
        "print(sagittalTest_reloaded.shape)\n",
        "test_sagittal=[]\n",
        " \n",
        "for j in range(len(sagittalTest_reloaded)):\n",
        "    mid2=math.floor((sagittalTest_reloaded[j].shape[0])/2)\n",
        "    test_sagittal.append([])\n",
        "       \n",
        "    for n in range(mid2-1,mid2+2):\n",
        "         test_sagittal[j].append(sagittalTest_reloaded[j][n])\n",
        "\n",
        "test_sagittal=np.array(test_sagittal)\n",
        "test_sagittal=np.reshape(test_sagittal,(120,256,256,3))\n",
        "print(test_sagittal.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(1131,)\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:12: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray.\n",
            "  if sys.path[0] == '':\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(1130,)\n",
            "(1130, 256, 256, 3)\n",
            "(120,)\n",
            "(120,)\n",
            "(120, 256, 256, 3)\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:40: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Also reading axial scans for future use"
      ],
      "metadata": {
        "id": "c3fhgtN2vfKf"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "axial_dir = os.listdir(\"/content/drive/MyDrive/Medical/MRNet-v1.0/train/axial/\")\n",
        "axial_dir = [f.lower() for f in axial_dir]   \n",
        "a=sorted(axial_dir)\n",
        "axial_list=np.array(a)\n",
        "axial_reloaded=[]\n",
        "print(axial_list.shape)\n",
        "for j in range(1,1131):\n",
        "    file= \"/content/drive/MyDrive/Medical/MRNet-v1.0/train/axial/\"+axial_list[j]    \n",
        "    axial_reloaded.append(np.load(file,allow_pickle='false'))   \n",
        "\n",
        "axial_reloaded=np.array(axial_reloaded)\n",
        "print(axial_reloaded.shape)\n",
        "s=axial_reloaded.shape\n",
        "print(axial_reloaded[0].shape[0])\n",
        "train_axial=[]\n",
        "print(len(axial_reloaded))\n",
        "for j in range(len(axial_reloaded)):\n",
        "    mid=math.floor((axial_reloaded[j].shape[0])/2)\n",
        "    train_axial.append([])    \n",
        "\n",
        "    for n in range(mid-1,mid+2):\n",
        "         train_axial[j].append(axial_reloaded[j][n])\n",
        "train_axial=np.array(train_axial)\n",
        "train_axial=np.reshape(train_axial,(1130,256,256,3))\n",
        "print(train_axial.shape)\n",
        "axial_dir = os.listdir(\"/content/drive/MyDrive/Medical/MRNet-v1.0/valid/axial/\")\n",
        "axial_dir = [f.lower() for f in axial_dir]   \n",
        "a=sorted(axial_dir)\n",
        "axialTest_list=np.array(a)\n",
        "axialTest_reloaded=[]\n",
        "\n",
        "print(axialTest_list.shape)\n",
        "\n",
        "for j in range(0,120):\n",
        "    file= \"/content/drive/MyDrive/Medical/MRNet-v1.0/valid/axial/\"+axialTest_list[j]\n",
        "    axialTest_reloaded.append(np.load(file,allow_pickle='false'))   \n",
        "\n",
        "axialTest_reloaded=np.array(axialTest_reloaded)\n",
        "print(axialTest_reloaded.shape)\n",
        "s=axialTest_reloaded.shape\n",
        "print(axialTest_reloaded[0].shape[0])\n",
        "test_axial=[]\n",
        "print(len(axialTest_reloaded))\n",
        "for j in range(len(axialTest_reloaded)):\n",
        "    mid=math.floor((axialTest_reloaded[j].shape[0])/2)\n",
        "    test_axial.append([])    \n",
        "\n",
        "    for n in range(mid-1,mid+2):\n",
        "         test_axial[j].append(axialTest_reloaded[j][n])\n",
        "test_axial=np.array(test_axial)\n",
        "test_axial=np.reshape(test_axial,(120,256,256,3))\n",
        "print(test_axial.shape)"
      ],
      "metadata": {
        "id": "uRxuqxZHJf9G",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "9e994ffc-9adb-48aa-fdb5-6660c93f2edb"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(1131,)\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:13: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray.\n",
            "  del sys.path[0]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(1130,)\n",
            "44\n",
            "1130\n",
            "(1130, 256, 256, 3)\n",
            "(120,)\n",
            "(120,)\n",
            "25\n",
            "120\n",
            "(120, 256, 256, 3)\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:47: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZhRgNysBy9HJ"
      },
      "source": [
        "#### Finetuning VGG16 Model for Sagittal Plane Scanes checking for Mensical Condition"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2c_Dtv7Jy9HP",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "1f4b5e4a-0cb3-4191-a07f-3157c57873f2"
      },
      "source": [
        "input_tensor = Input(shape=(256, 256,3))  \n",
        "base_model_Sa_Men = VGG16(input_tensor=input_tensor, weights='imagenet', include_top=False)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/vgg16/vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5\n",
            "58892288/58889256 [==============================] - 1s 0us/step\n",
            "58900480/58889256 [==============================] - 1s 0us/step\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Cde65r48y9Hd"
      },
      "source": [
        "x = base_model_Sa_Men.output\n",
        "x = GlobalAveragePooling2D()(x)\n",
        "x=Dropout(0.2)(x)\n",
        "x = Dense(1024, activation='relu')(x)\n",
        "x=Dropout(0.2)(x)\n",
        "x=Dropout(0.2)(x)\n",
        "predictions = Dense(1, activation='sigmoid')(x)\n",
        "model7 = Model(inputs=base_model_Sa_Men.input,outputs=predictions)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1fBMXXA4y9Hj"
      },
      "source": [
        "for layer in model7.layers[:15]:\n",
        "   layer.trainable = True\n",
        "for layer in model7.layers[15:]:\n",
        "   layer.trainable = True"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "trainMen_labels = np.array(trainMen_labels)"
      ],
      "metadata": {
        "id": "Ml-ISqmzKQHN"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vzHwBUqyy9Hm",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "aaecb827-a7e0-41a7-c885-c20120100e19"
      },
      "source": [
        "early_stopping_monitor = EarlyStopping(patience=10)    \n",
        "model7.compile(optimizer=SGD(lr=0.001,momentum=0.9), loss='binary_crossentropy',metrics=['accuracy',f1_score])\n",
        "history7=model7.fit(train_sagittal ,trainMen_labels, validation_split=0.05,shuffle=True ,epochs=100, callbacks=[early_stopping_monitor], batch_size = 20)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/keras/optimizer_v2/gradient_descent.py:102: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.\n",
            "  super(SGD, self).__init__(name, **kwargs)\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/100\n",
            "54/54 [==============================] - 49s 872ms/step - loss: 0.8814 - accuracy: 0.6253 - f1_score: nan - val_loss: 0.6869 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 2/100\n",
            "54/54 [==============================] - 45s 842ms/step - loss: 0.6757 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6755 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 3/100\n",
            "54/54 [==============================] - 45s 841ms/step - loss: 0.6629 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6697 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 4/100\n",
            "54/54 [==============================] - 45s 840ms/step - loss: 0.6555 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6675 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 5/100\n",
            "54/54 [==============================] - 45s 839ms/step - loss: 0.6514 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6669 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 6/100\n",
            "54/54 [==============================] - 45s 840ms/step - loss: 0.6500 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6670 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 7/100\n",
            "54/54 [==============================] - 45s 840ms/step - loss: 0.6483 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6673 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 8/100\n",
            "54/54 [==============================] - 45s 840ms/step - loss: 0.6481 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6679 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 9/100\n",
            "54/54 [==============================] - 45s 840ms/step - loss: 0.6474 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6683 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 10/100\n",
            "54/54 [==============================] - 45s 840ms/step - loss: 0.6473 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6686 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 11/100\n",
            "54/54 [==============================] - 45s 840ms/step - loss: 0.6473 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6690 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 12/100\n",
            "54/54 [==============================] - 45s 840ms/step - loss: 0.6470 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6691 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 13/100\n",
            "54/54 [==============================] - 45s 840ms/step - loss: 0.6473 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6693 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 14/100\n",
            "54/54 [==============================] - 45s 840ms/step - loss: 0.6469 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6696 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 15/100\n",
            "54/54 [==============================] - 45s 840ms/step - loss: 0.6470 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6696 - val_accuracy: 0.6140 - val_f1_score: nan\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "B_Ql5qSTy9Hp",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "214c4d33-03d1-4024-b2b4-233dfc8bbe7d"
      },
      "source": [
        "p = np.array(model7.predict(train_sagittal))\n",
        "train_Men_pred.append(p)\n",
        "p = np.array(model7.predict(test_sagittal))\n",
        "test_Men_pred.append(p)            \n",
        "score7 = model7.evaluate(test_sagittal ,testMen_labels,verbose=1)       \n",
        "print('Test loss:',score7[0])\n",
        "print('Test accuracy:',score7[1])                                                         \n",
        "plt.figure()\n",
        "plt.grid()                                                                                                \n",
        "plt.title('Training performance')                          \n",
        "plt.plot(history7.epoch, history7.history['loss'], label='train loss+error')  \n",
        "plt.plot(history7.epoch, history7.history['val_loss'], label='val_error')\n",
        "plt.plot(history7.epoch, history7.history['f1_score'], label='f1_score')\n",
        "plt.legend()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "4/4 [==============================] - 1s 356ms/step - loss: 0.6987 - accuracy: 0.5667 - f1_score: nan\n",
            "Test loss: 0.6986700892448425\n",
            "Test accuracy: 0.5666666626930237\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7fe3bd3e3490>"
            ]
          },
          "metadata": {},
          "execution_count": 60
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "filename = '/content/drive/MyDrive/Medical/MRNET MODEL/VGGmodel7.h5'\n",
        "model7.save(filename)\n",
        "print('>Saved %s' % filename)\n",
        "clean_up(model=model7)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_UhIWFPTBPFz",
        "outputId": "e9e1de2c-f1f9-4b5f-ca73-a5f2c50d2d85"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            ">Saved /content/drive/MyDrive/Medical/MRNET MODEL/VGGmodel7.h5\n",
            "...Finished clean up...\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "JWF36DXt0lj1"
      },
      "source": [
        "#### Finetuning VGG16 Model for Sagittal Plane Scanes checking for ACL Condition"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "u6ItTeBD0lj4"
      },
      "source": [
        "input_tensor = Input(shape=(256, 256,3))  \n",
        "base_model_Sa_ACL = VGG16(input_tensor=input_tensor, weights='imagenet', include_top=False)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2Xe_fFX40lkE"
      },
      "source": [
        "x = base_model_Sa_ACL.output\n",
        "x = GlobalAveragePooling2D()(x)\n",
        "x=Dropout(0.2)(x)\n",
        "x = Dense(1024, activation='relu')(x)\n",
        "x=Dropout(0.2)(x)\n",
        "x=Dropout(0.2)(x)\n",
        "predictions = Dense(1, activation='sigmoid')(x)\n",
        "model8 = Model(inputs=base_model_Sa_ACL.input,outputs=predictions)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nrvR0Zko0lkK"
      },
      "source": [
        "for layer in model8.layers[:15]:\n",
        "   layer.trainable = True\n",
        "for layer in model8.layers[15:]:\n",
        "   layer.trainable = True"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7lgg-EjL0lkP",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "b1d072ff-9ff2-43b7-a7de-1a55a3c1704d"
      },
      "source": [
        "early_stopping_monitor = EarlyStopping(patience=5)\n",
        "model8.compile(optimizer=SGD(lr=0.001,momentum=0.9), loss='binary_crossentropy',metrics=['accuracy',f1_score])\n",
        "history8=model8.fit(train_sagittal ,trainACL_labels, validation_split=0.05,shuffle=True ,epochs=100, callbacks=[early_stopping_monitor], batch_size = 20)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/100\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/keras/optimizer_v2/gradient_descent.py:102: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.\n",
            "  super(SGD, self).__init__(name, **kwargs)\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "54/54 [==============================] - 49s 877ms/step - loss: 0.7382 - accuracy: 0.7512 - f1_score: nan - val_loss: 0.4712 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 2/100\n",
            "54/54 [==============================] - 47s 868ms/step - loss: 0.5241 - accuracy: 0.8108 - f1_score: nan - val_loss: 0.4668 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 3/100\n",
            "54/54 [==============================] - 47s 868ms/step - loss: 0.5167 - accuracy: 0.8145 - f1_score: nan - val_loss: 0.4254 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 4/100\n",
            "54/54 [==============================] - 47s 869ms/step - loss: 0.5186 - accuracy: 0.8145 - f1_score: nan - val_loss: 0.4406 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 5/100\n",
            "54/54 [==============================] - 47s 870ms/step - loss: 0.5028 - accuracy: 0.8145 - f1_score: nan - val_loss: 0.4417 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 6/100\n",
            "54/54 [==============================] - 47s 870ms/step - loss: 0.4797 - accuracy: 0.8145 - f1_score: nan - val_loss: 0.4310 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 7/100\n",
            "54/54 [==============================] - 47s 871ms/step - loss: 0.4969 - accuracy: 0.8145 - f1_score: nan - val_loss: 0.4470 - val_accuracy: 0.8421 - val_f1_score: nan\n",
            "Epoch 8/100\n",
            "54/54 [==============================] - 47s 872ms/step - loss: 0.4708 - accuracy: 0.8127 - f1_score: nan - val_loss: 0.4343 - val_accuracy: 0.8421 - val_f1_score: nan\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4GxoI6J90lkY",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "21a307bc-e8bc-40be-de1b-12857e254454"
      },
      "source": [
        "p = np.array(model8.predict(train_sagittal))\n",
        "train_ACL_pred.append(p)\n",
        "p = np.array(model8.predict(test_sagittal))\n",
        "test_ACL_pred.append(p)                       \n",
        "score8 = model8.evaluate(test_sagittal ,testACL_labels,verbose=1)       \n",
        "print('Test loss:',score8[0])\n",
        "print('Test accuracy:',score8[1])                                                                   \n",
        "plt.figure()\n",
        "plt.grid()                                                                                                         \n",
        "plt.title('Training performance')                          \n",
        "plt.plot(history8.epoch, history8.history['loss'], label='train loss+error')  \n",
        "plt.plot(history8.epoch, history8.history['val_loss'], label='val_error')\n",
        "plt.plot(history8.epoch, history8.history['f1_score'], label='f1_score')\n",
        "plt.legend()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "4/4 [==============================] - 1s 361ms/step - loss: 0.7610 - accuracy: 0.5500 - f1_score: nan\n",
            "Test loss: 0.7609687447547913\n",
            "Test accuracy: 0.550000011920929\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7fe3bcf0bb50>"
            ]
          },
          "metadata": {},
          "execution_count": 66
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "filename = '/content/drive/MyDrive/Medical/MRNET MODEL/VGGmodel8.h5'\n",
        "model8.save(filename)\n",
        "print('>Saved %s' % filename)\n",
        "clean_up(model=model8)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "czkeMRbjBStB",
        "outputId": "781ac821-e580-4427-aabe-887ca5abee77"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            ">Saved /content/drive/MyDrive/Medical/MRNET MODEL/VGGmodel8.h5\n",
            "...Finished clean up...\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "QntZqwr22Qqu"
      },
      "source": [
        "#### Finetuning VGG16 Model for Sagittal Plane Scanes checking for Abnormal Condition"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "W6WQFKWN2Qqx"
      },
      "source": [
        "input_tensor = Input(shape=(256, 256,3))  \n",
        "base_model_Sa_Ab = VGG16(input_tensor=input_tensor, weights='imagenet', include_top=False)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0-M3wkWd2Qq5"
      },
      "source": [
        "x = base_model_Sa_Ab.output\n",
        "x = GlobalAveragePooling2D()(x)\n",
        "x=Dropout(0.2)(x)\n",
        "x = Dense(1024, activation='relu')(x)\n",
        "x=Dropout(0.2)(x)\n",
        "x=Dropout(0.2)(x)\n",
        "predictions = Dense(1, activation='sigmoid')(x)\n",
        "model9 = Model(inputs=base_model_Sa_Ab.input,outputs=predictions)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "l-F3Exud2Qq_"
      },
      "source": [
        "for layer in model9.layers[:15]:\n",
        "   layer.trainable = True\n",
        "for layer in model9.layers[15:]:\n",
        "   layer.trainable = True"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WAMH3it52QrK",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "63ad6158-4ba3-4aaa-d81f-f5d90ff7b540"
      },
      "source": [
        "early_stopping_monitor = EarlyStopping(patience=10)    \n",
        "model9.compile(optimizer=SGD(lr=0.001,momentum=0.9), loss='binary_crossentropy',metrics=['accuracy',f1_score])\n",
        "history9=model9.fit(train_sagittal ,trainAb_labels, validation_split=0.05,shuffle=True ,epochs=100, callbacks=[early_stopping_monitor], batch_size = 20)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/keras/optimizer_v2/gradient_descent.py:102: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.\n",
            "  super(SGD, self).__init__(name, **kwargs)\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/100\n",
            "54/54 [==============================] - 85s 1s/step - loss: 1.0111 - accuracy: 0.5974 - f1_score: nan - val_loss: 0.6822 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 2/100\n",
            "54/54 [==============================] - 55s 1s/step - loss: 0.6714 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6741 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 3/100\n",
            "54/54 [==============================] - 55s 1s/step - loss: 0.6612 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6692 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 4/100\n",
            "54/54 [==============================] - 55s 1s/step - loss: 0.6554 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6675 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 5/100\n",
            "54/54 [==============================] - 55s 1s/step - loss: 0.6523 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6670 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 6/100\n",
            "54/54 [==============================] - 55s 1s/step - loss: 0.6497 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6670 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 7/100\n",
            "54/54 [==============================] - 55s 1s/step - loss: 0.6484 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6673 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 8/100\n",
            "54/54 [==============================] - 55s 1s/step - loss: 0.6480 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6676 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 9/100\n",
            "54/54 [==============================] - 55s 1s/step - loss: 0.6478 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6679 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 10/100\n",
            "54/54 [==============================] - 55s 1s/step - loss: 0.6478 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6684 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 11/100\n",
            "54/54 [==============================] - 55s 1s/step - loss: 0.6471 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6686 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 12/100\n",
            "54/54 [==============================] - 55s 1s/step - loss: 0.6475 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6689 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 13/100\n",
            "54/54 [==============================] - 55s 1s/step - loss: 0.6476 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6691 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 14/100\n",
            "54/54 [==============================] - 55s 1s/step - loss: 0.6474 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6691 - val_accuracy: 0.6140 - val_f1_score: nan\n",
            "Epoch 15/100\n",
            "54/54 [==============================] - 55s 1s/step - loss: 0.6469 - accuracy: 0.6505 - f1_score: nan - val_loss: 0.6693 - val_accuracy: 0.6140 - val_f1_score: nan\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bb_cHRLb2QrW",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 354
        },
        "outputId": "4e3d2bb4-e661-4117-d38f-55b2daf6883b"
      },
      "source": [
        "p = np.array(model9.predict(train_sagittal))\n",
        "train_Ab_pred.append(p)\n",
        "p = np.array(model9.predict(test_sagittal))\n",
        "test_Ab_pred.append(p)                       \n",
        "score9 = model9.evaluate(test_sagittal ,testAb_labels,verbose=1)       \n",
        "print('Test loss:',score9[0])\n",
        "print('Test accuracy:',score9[1])                                                                   \n",
        "plt.figure()\n",
        "plt.grid()                                                                                                       \n",
        "plt.title('Training performance')                          \n",
        "plt.plot(history9.epoch, history9.history['loss'], label='train loss+error')  \n",
        "plt.plot(history9.epoch, history9.history['val_loss'], label='val_error')\n",
        "plt.plot(history9.epoch, history9.history['f1_score'], label='f1_score')\n",
        "plt.legend()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "4/4 [==============================] - 2s 422ms/step - loss: 0.9162 - accuracy: 0.2083 - f1_score: nan\n",
            "Test loss: 0.9161697626113892\n",
            "Test accuracy: 0.2083333283662796\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f3cc9ed81d0>"
            ]
          },
          "metadata": {},
          "execution_count": 18
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "filename = '/content/drive/MyDrive/Medical/MRNET MODEL/VGGmodel9.h5'\n",
        "model9.save(filename)\n",
        "print('>Saved %s' % filename)\n",
        "clean_up(model=model9)"
      ],
      "metadata": {
        "id": "YCBVv3zEJv6z"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TlbunAuE2Qre"
      },
      "source": [
        "del train_sagittal\n",
        "del test_sagittal\n",
        "del sagittal_dir\n",
        "del sagittal_reloaded\n",
        "del sagittalTest_reloaded\n",
        "del sagittal_list\n",
        "del sagittalTest_list\n",
        "del model7\n",
        "del model8\n",
        "del model9\n",
        "del base_model_Sa_Men\n",
        "del base_model_Sa_ACL\n",
        "del base_model_Sa_Ab\n",
        "del p"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "p8rTxtG6vzbk"
      },
      "source": [
        "##Ensembling\n",
        "\n",
        "---\n",
        "We ensemble the models for each state detection for three planes Axial,Coronal and Sagittal respectively and train a Sequential model with input being probabitity of condition detection by indiviual model and output again being probability person has the condition. We give original label for evaluation of the method.\n",
        "\n",
        "`We couldnt run this for VGG16 as we ran out of Memory/RAM in colab but this worked with other baselines models like ResNet50`"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "D2thlG5_12K_"
      },
      "source": [
        "#### Meniscal Condition Detection"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "woSckwh419pD"
      },
      "source": [
        "train_Men_pred = np.array(train_Men_pred)\n",
        "print(train_Men_pred.shape)\n",
        "test_Men_pred = np.array(test_Men_pred)\n",
        "print(test_Men_pred.shape)\n",
        "train_Men_pred = train_Men_pred.reshape((1130,3))\n",
        "test_Men_pred = test_Men_pred.reshape((120,3))\n",
        "print(train_Men_pred)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4LsT7P9j31am"
      },
      "source": [
        "menModel = Sequential()\n",
        "menModel.add(Dense(1, activation='sigmoid', input_dim=3))\n",
        "menModel.compile(loss='binary_crossentropy', optimizer='adam', metrics=['acc',f1_score])\n",
        "h = menModel.fit(train_Men_pred, trainMen_labels, batch_size=32, epochs=50, verbose=1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NqS1xDpo4LjW"
      },
      "source": [
        "loss, acc, f1 = menModel.evaluate(test_Men_pred, testMen_labels)\n",
        "p1 = menModel.predict(test_Men_pred)\n",
        "print(\"Loss: \", loss)\n",
        "print('Accuracy: ', acc)                                                  \n",
        "plt.figure()\n",
        "plt.grid()                                                           \n",
        "plt.title('Testing performance')                          \n",
        "plt.plot(h.epoch, h.history['loss'], label='train loss+error')  \n",
        "plt.plot(h.epoch, h.history['f1_score'], label='f1_score')\n",
        "plt.legend()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oiKi1DebC4X8"
      },
      "source": [
        "del train_Men_pred\n",
        "del menModel"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pJs7J20U4lGu"
      },
      "source": [
        "#### ACL Condition Detection"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Za99XogRCz-n"
      },
      "source": [
        "train_ACL_pred = np.array(train_ACL_pred)\n",
        "print(train_ACL_pred.shape)\n",
        "test_ACL_pred = np.array(test_ACL_pred)\n",
        "print(test_Men_pred.shape)\n",
        "train_ACL_pred = train_ACL_pred.reshape((1130,3))\n",
        "test_ACL_pred = test_ACL_pred.reshape((120,3))\n",
        "print(test_ACL_pred)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sHllxCwuCz-2"
      },
      "source": [
        "menModel = Sequential()\n",
        "menModel.add(Dense(1, activation='sigmoid', input_dim=3))\n",
        "menModel.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy',f1_score])\n",
        "h = menModel.fit(train_ACL_pred, trainACL_labels, batch_size=32, epochs=50, verbose=1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LPmLLhj1Cz_A"
      },
      "source": [
        "loss, acc, f1 = menModel.evaluate(test_ACL_pred, testACL_labels)\n",
        "p2 = menModel.predict(test_ACL_pred)\n",
        "print(\"Loss: \", loss)\n",
        "print('Accuracy: ', acc)                                \n",
        "plt.figure()\n",
        "plt.grid()                                                                                                       \n",
        "plt.title('Testing performance')                          \n",
        "plt.plot(h.epoch, h.history['loss'], label='train loss+error')  \n",
        "plt.plot(h.epoch, h.history['f1_score'], label='f1_score')\n",
        "plt.legend()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dm6Y7ofiDmgr"
      },
      "source": [
        "del train_ACL_pred\n",
        "del menModel"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VXE1feJ7D_qQ"
      },
      "source": [
        "#### Abnormal Condition Detection"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "np.isnan(np.sum(trainAb_labels))"
      ],
      "metadata": {
        "id": "_6_8RiNlrRWT"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UvKiF1LmEOlO"
      },
      "source": [
        "train_Ab_pred = np.array(train_Ab_pred)\n",
        "print(train_Ab_pred.shape)\n",
        "test_Ab_pred = np.array(test_Ab_pred)\n",
        "print(test_Ab_pred.shape)\n",
        "train_Ab_pred = train_Ab_pred.reshape((1130,3))\n",
        "test_Ab_pred = test_Ab_pred.reshape((120,3))\n",
        "print(test_ACL_pred)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WM7sx5NGEOlj"
      },
      "source": [
        "abModel = Sequential()\n",
        "abModel.add(Dense(1, activation='sigmoid', input_dim=3))\n",
        "abModel.compile(loss='binary_crossentropy', optimizer='adam', metrics=['acc',f1_score] )\n",
        "h = abModel.fit(train_Ab_pred, trainAb_labels, batch_size=32, epochs=50, verbose=1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RiP0cFwPEOly"
      },
      "source": [
        "loss, acc, f1 = abModel.evaluate(test_Ab_pred, testAb_labels)\n",
        "p3 = abModel.predict(test_Ab_pred)\n",
        "print(\"Loss: \", loss)\n",
        "print('Accuracy: ', acc)                                  \n",
        "plt.figure()\n",
        "plt.grid()                                                                                                       \n",
        "plt.title('Testing performance')                          \n",
        "plt.plot(h.epoch, h.history['loss'], label='train loss+error')  \n",
        "plt.plot(h.epoch, h.history['f1_score'], label='f1_score')\n",
        "plt.legend()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EBX_Lnb8EOl8"
      },
      "source": [
        "del train_Ab_pred\n",
        "del abModel"
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}