Brain Tumor Detection using CNN & FastAPI
This repository contains a FastAPI-based API for detecting brain tumors using a Convolutional Neural Network (CNN) model. The model is trained on a dataset of MRI images labeled as either "tumor" or "no tumor." The API allows users to upload an MRI image and receive a prediction indicating whether the image contains a brain tumor.
Brain tumor detection is a critical task in medical imaging, as early and accurate detection can significantly impact patient outcomes. This project leverages deep learning techniques to build a CNN model capable of classifying MRI images into two categories: "tumor" or "no tumor." The trained model is then deployed as an API using FastAPI, providing an accessible interface for users to interact with the model and make predictions.
- Image Upload: Users can upload MRI images directly through the web interface provided by the
index.htmlpage. Thestatic/styles.cssfile ensures that the page is styled for a better user experience. - Prediction: The FastAPI API endpoint
/predict/processes the uploaded image and returns a prediction indicating whether the image contains a brain tumor. Predictions are displayed on the web interface below the uploaded image. - Model Summary: A Convolutional Neural Network (CNN) is used for prediction. The model includes layers for convolution, pooling, flattening, and dense connections, and has been trained to achieve a high accuracy rate.
- Data Augmentation: The model utilizes data augmentation techniques to enhance generalization and robustness. This helps in improving the model's performance by exposing it to a wider variety of training examples.
- Visualization: Scripts are available for visualizing training results, including loss curves and confusion matrices, which help in evaluating the model's performance and understanding its learning progress.
- User Interface: A simple and user-friendly web interface allows users to upload images and view predictions. The interface is styled using a separate
styles.cssfile for a clean and professional look.
- Anaconda or Miniconda
- Python 3.8 (included in the Conda environment)
git clone https://github.com/Islam-hady9/BrainTumorDetection-API.git
cd BrainTumorDetection-APIIt's recommended to create a Conda virtual environment to manage dependencies.
conda create -n brain-tumor-detection python=3.8
conda activate brain-tumor-detectionInstall the required Python packages using the requirements.txt file:
pip install -r requirements.txtIf you haven’t trained the model yet, use the Jupyter Notebook brain_tumor_detection_using_cnn.ipynb provided in the repository. Ensure you have the MRI images in the Dataset/ directory and in the correct format before running the notebook or use the pre-trained model brain_tumor_cnn_model.h5 directly that I trained.
Start the FastAPI server to expose the API:
uvicorn app:app --reloadThe server will start running at http://127.0.0.1:8000/.
Open your browser and navigate to http://127.0.0.1:8000 to access the HTML interface. You can upload images through this interface and receive predictions.
You can make predictions by sending a POST request to the /predict/ endpoint with an image file.
Example using curl:
curl -X POST "http://127.0.0.1:8000/predict/" -F "file=@path_to_your_image.jpg"Replace path_to_your_image.jpg with the actual path to your MRI image file.
You can test the API using the test_api.py script provided in the repository:
python test_api.pyThis script will send a sample image to the API and print the prediction result. Make sure to modify the script if necessary to use the correct path to your test image.
app.py: The FastAPI application script that defines the API endpoints and handles image uploads and predictions.test_api.py: A script for testing the API by sending a sample image and printing the prediction result.requirements.txt: A file listing all Python dependencies required for the project.brain_tumor_detection_using_cnn.ipynb: Jupyter Notebook used for training the CNN model (Accuracy: 92.16%).brain_tumor_model.h5: The trained Keras model file (to be generated after training).Dataset/: Directory containing MRI images used for training and testing the model.templates/: Directory containing HTML templates.index.html: The main HTML file providing the user interface for image uploads and predictions.
static/: Directory containing static files such as CSS.styles.css: The CSS file used for styling theindex.htmlpage.
Project Presentation.pptx: Project presentation file.README.md: This file, providing an overview and instructions for the project.
The dataset used for training the model should be placed in the Dataset/brain_tumor_dataset directory, with subdirectories yes and no for images containing tumors and images without tumors, respectively.
- Yes Tumor Directory:
Dataset/brain_tumor_dataset/yes - No Tumor Directory:
Dataset/brain_tumor_dataset/no
Ensure that all images are in a compatible format (e.g., .jpg, .png).
The CNN model used in this project has the following architecture:
- Conv2D Layers: For feature extraction from the input images.
- MaxPooling2D Layers: For downsampling the feature maps.
- Flatten Layer: To convert 2D feature maps into 1D feature vectors.
- Dense Layers: Fully connected layers for classification.
- Output Layer: A softmax layer with 2 units (tumor, no tumor).
The repository includes code for visualizing the training process, such as loss curves and confusion matrices. These visualizations can help in understanding the model's performance and diagnosing potential issues.
Contributions to the project are welcome. If you encounter any issues or have suggestions for improvements, feel free to open an issue or submit a pull request.
This project is licensed under the MIT License.
For questions or inquiries, please contact [Islam Abd_Elhady] at [eslamabdo71239@gmail.com].