X-Ray Vision
Deep learning chest X-ray pathology classifier built with PyTorch, OpenCV, and custom CNN architectures for automated pulmonary anomaly detection.
Overview
Automated computer vision screening tool for chest radiography, distinguishing healthy scans from pneumonia and pulmonary opacity anomalies with high sensitivity.
Architecture & Design
PyTorch deep CNN with transfer learning (DenseNet-121 / ResNet-50), Grad-CAM heatmap visualization highlighting anomaly regions in the lung fields.
Key Capabilities
- Grad-CAM visual saliency maps indicating diagnostic focus regions
- Contrast-limited adaptive histogram equalization (CLAHE) image preprocessing
- High sensitivity rate (>95%) tuned for clinical triage screening