I designed and deployed a full-stack deep learning application for automated chest X-ray analysis. This project demonstrates a complete AI solution pipeline, from implementing a state-of-the-art convolutional neural network to building an interactive web application for real-time inference. The system utilizes the CheXNet architecture, a 121-layer DenseNet, to detect and provide probability scores for 14 different thoracic pathologies, emulating a foundational analytical process for medical image interpretation.
The core system was developed in PyTorch, implementing the CheXNet model with a modified classifier and sigmoid activation for multi-label classification. I engineered a pre-processing pipeline to prepare medical images for the network and built an intuitive, responsive web interface using Streamlit. This interface allows users to upload chest X-ray images and receive instant, color-coded results that highlight findings based on confidence levels. The entire application was deployed to the Streamlit Community Cloud for public access, and the project is showcased on a professionally developed WordPress site.
This system is designed not as a diagnostic tool but as a potential assistive technology, and its future evolution must be guided by close collaboration with healthcare professionals to ensure it meets clinical needs and integrates safely into existing workflows.
Key Tasks Completed:
Deep Learning & Computer Vision: PyTorch, CNN Architecture (DenseNet-121), Transfer Learning, Medical AI
Full-Stack Development: Streamlit, Web Application Deployment, API-less Interaction Design
MLOps & Engineering: End-to-End Pipeline Development, Model Preprocessing, Version Control (Git)
Deployment & Presentation: Cloud Deployment (Streamlit Community Cloud), WordPress CMS, Technical Documentation
⚠️ Attention
If you see the message “This app has gone to sleep due to inactivity”, it simply means the service we use is hosted on a free platform that automatically goes to sleep when there hasn’t been any activity.
Just click “Wake it up.”
After it loads, you will see the application shown in the image below.