This repository contains the original implementation of "Tuberculosis chest X-ray detection using CNN-based hybrid segmentation and classification approach" in Keras (Tensorflow as backend). This paper has been published in "Biomedical Signal Processing and Control - Elsevier, IF:5.076"

Proposed Architecture

TB-UNet:

TB-DenseNet:

Grad-CAM:

Web-based Prototype

A real-time demo of the Tuberculosis detection app has been uploaded to YouTube: https://youtu.be/jv3N5F91LbI

Requirements

• Python 3.7.11
• TensorFlow: 2.1.0
• Keras: 2.2.4
• OpenCV: 4.5.3
• Numpy: 1.19.1
• Matplotlib: 3.4.3

Cite:

If you use TB-UNet or TB-DenseNet in your project, please cite the following paper:

Iqbal, A., Usman, M., & Ahmed, Z. (2023). Tuberculosis chest X-ray detection using CNN-based hybrid segmentation and classification approach. Biomedical Signal Processing and Control, Volume 84, July 2023, 104667. DOI: https://doi.org/10.1016/j.bspc.2023.104667