MediScan
i strapped three neural nets together and taught them to read chest x-rays — 15 diseases, heatmaps, confidence scores — then built an entire hospital around it. engineered like lives depend on it. one day they might.
◉ FRAME 01
◉ FRAME 02An enterprise-grade, full-stack clinical decision support system. A multi-model AI ensemble — YOLOv12m detection, EfficientNet-B0 classification, and a Lungmask U-Net segmenter — reads chest X-rays across 15 lung-disease classes with confidence scores and Grad-CAM heatmaps. Around the model sits a real clinical workflow: patient management and a full case lifecycle (upload → AI inference → radiologist markup → doctor diagnosis → automated PDF report), behind RBAC, 2FA, AES-256 field-level encryption and full audit logging. FastAPI + PostgreSQL backend, Next.js 16 / React 19 dashboard, Docker + Kubernetes, Prometheus + Celery ops.