Multi-Disease Chest X-ray AI System
A modular deep learning system that runs multiple independent chest X-ray models
(TB, COVID-19, and NIH multi-disease) in parallel on a single input image.
The system is designed with strict model isolation, unified orchestration, and
explainability where clinically appropriate. It is deployable locally and not
tied to notebook-based workflows.
Features
- Independent disease-specific deep learning models
- Parallel inference on a single X-ray input
- Grad-CAM explainability where applicable
- Multi-label prediction support (NIH model)
- Unified Streamlit-based interface
- Local deployment (VS Code ready)
System Overview
The application orchestrates three independent pipelines:
- Tuberculosis Detection (ResNet-based, single-label)
- COVID-19 Detection (DenseNet-based, single-label with risk mapping)
- NIH Multi-Disease Detection (DenseNet-based, multi-label sigmoid output)
Each model is trained, loaded, and executed independently to avoid label
interference and to preserve interpretability.
How to Run
- pip install -r requirements.txt
- streamlit run app.py
Contributors
| Name | Role | Responsibilities |
|---|
| Shreya R Chittaragi | Lead ML Engineer | TBX11K (Tuberculosis) pipeline, COVID-19 model integration, multi-model orchestration, Streamlit deployment |
| Devika N D | Lead ML Engineer | NIH multi-disease model training, multi-label inference, dataset preparation |
| Jagadeesh R S | Project Operations & Coordination | Team coordination, task management, submission support |
| Kiran H | Technical Documentation & Presentation | Documentation structuring, project presentation, demo support |
