Scanalyse AI

Developer Names: Hamza Issa, Jared Paul, Ahmad Hamadi, Gurnoor Bal

Date of project start: 21 September 2024

This project is develop a convolutional neural network to identify lung and cardiac conditions in chest X-ray images.

Project Overview

The Scanalyse AI project is designed to assist radiologists and healthcare professionals by automating the detection of lung and cardiac conditions in chest X-rays. The system leverages deep learning techniques to provide accurate and interpretable predictions for multiple diseases. The project includes the following components:

1. Machine Learning Model

• A custom MobileNetV2 CNN model trained on the NIH chest X-ray dataset which has over 100,000+ images.
• Supports multi-disease classification for 13 conditions, including:
  • Atelectasis, Cardiomegaly, Consolidation, Edema, Effusion, Emphysema, Fibrosis, Infiltration, Mass, Nodule, Pleural Thickening, Pneumonia, and Pneumothorax.
• Custom training
  • Data augmentation
  • Class balancing with weighted loss functions.
  • Sparsity and margin loss regularization for improved generalization.

2. Backend API

• Built using Flask to serve predictions and handle requests.
• Key routes:
  • /test: Verifies API connectivity.
  • /predict: Accepts an X-ray image and returns disease predictions along with probabilities.
• Integrates with the trained PyTorch model for real-time inference.

3. Frontend User Interface

• Developed using React for a modern and responsive design.
• Features:
  • Drag-and-drop upload area for chest X-ray images.
  • Displays predictions and probabilities in a user-friendly format.
  • Styled with Tailwind CSS.

Prerequisites

• Python 3.8 or higher
• Node.js and npm
• PyTorch, Pillow, Albumentation Python libraries

The folders and files for this project are as follows:

docs - Documentation for the project

refs - did not use

src - Source code

test - Test cases

etc.

Style

Branch names must only be lower case letters, numbers, and hyphens.

Code style to be determined later.