Skip to content

kanasva/ai-chest-xray-diagnosis-full-stack

Repository files navigation

AI Chest X-ray Diagnosis Full-stack

Visit aichestxray.kanasva.me

TL;DR

This project utilises Next.js 14 as a full-stack framework to deliver diagnoses for 18 chest X-ray pathologies using TorchXRayVision, with the model deployed separately on AWS Lambda, see ai-chest-xray-diagnosis-api

Features

  • Diagnosis for 18 chest X-ray pathologies, including Atelectasis, Cardiomegaly, Consolidation, Edema, Effusion, Emphysema, Enlarged Cardiomediastinum, Fibrosis, Fracture, Hernia, Infiltration, Lung Lesion, Lung Opacity, Mass, Nodule, Pleural Thickening, Pneumonia, and Pneumothorax.
  • Detection of out-of-distribution images using an auto-encoder to prevent predictions on images that are different from the training data.
  • Grad-CAM visualisation to highlight regions of interest in X-ray images
  • GitHub OAuth authentication
  • Global and user-specific quotas

Technologies

Design

  • This project follows the Material Design 3 guidelines, especially the colour system.

Project Architecture

This project uses a layered architecture to ensure modularity, maintainability, and scalability:

  • User Interface: The front-end layer where users interact with the application.
  • Server Actions: The backend layer to handle API interactions.
  • Use Cases: The layer for business logic and application-specific functionality, such as quota checking.
  • Data Access: The layer that manages data retrieval and storage operations.

Further Implementation

  • Improve error handling
  • Google Login that needs to submit app verification

About

A full-stack framework to deliver diagnoses for 18 chest X-ray pathologies.

Resources

Stars

Watchers

Forks

Languages