Project Overview
This medical image viewer and analyzer processes dicom and image files to suggest regions of interest. By leveraging PyTorch Convolutional Neural Networks (CNNs), it detects abnormalities such as fractures or lung nodules, ranking them by confidence scores.
Key Features & Scope
Web DICOM viewer supporting canvas zooms, pan settings, and contrast filters
Heatmap visualizer showing ML focus coordinates (Grad-CAM visualization)
Patient data masking to comply with HIPAA healthcare security regulations
PDF audit reports detailing detected issues and probabilities
System Architecture
React client utilizing custom HTML5 canvas rendering. Flask serves as the backend gateway, delegating heavy neural network calculations to PyTorch nodes.
Client InterfaceNext.js / React
──────▶
Backend CoreNode.js / Express
──────▶
Database NodePostgreSQL & MinIO storage
Database: PostgreSQL & MinIO storage
Deployment: AWS ECS with GPU clusters
Performance & Vitals
94
97
99
100
Verified Production Metrics
Technologies Used
ReactPythonFlaskPyTorchOpenCVDockerAWS S3