Multi-Disease Medical Diagnosis AI
Deep learning system detecting 14 chest diseases from X-ray images with Grad-CAM explainability.
Project Overview
A clinical-grade AI system that analyzes chest X-ray images to detect 14 diseases simultaneously including Atelectasis, Cardiomegaly, Pneumonia, and more. Uses DenseNet121 with Grad-CAM to show radiologists exactly where disease indicators appear.
⚠ The Problem
Radiologists face enormous workloads screening chest X-rays. An AI assistant that can flag potential diseases and highlight regions of concern would significantly improve efficiency and accuracy.
✓ The Solution
Trained DenseNet121 on 4,999 real NIH chest X-ray images with weighted binary cross-entropy loss to handle class imbalance. Two-phase training: frozen base layers, then fine-tuned last 60 layers.
Results & Outcomes
Mean AUC 0.72 across 14 diseases. Approaches results from research papers using 50,000+ images. Grad-CAM heatmaps clearly show disease regions.
Visual Outputs
