Brain Tumor Detection System
AI-assisted full-stack diagnostic platform where users upload MRI scans and receive YOLO-powered tumor detection with confidence scores and annotated output images
Project Overview
Brain Tumor Detection System is an AI-assisted medical imaging platform combining a Next.js frontend, a NestJS API, and a Python FastAPI + YOLO inference microservice. Authenticated users upload brain scan images which are stored in Cloudinary, forwarded to the ML service for YOLO-based tumor detection, and returned with confidence scores and bounding-box-annotated output images. The backend orchestrates the full pipeline — Cloudinary upload, ML inference request, result persistence in PostgreSQL via TypeORM — and exposes modular REST API domains for scans, predictions, reports, histories, treatments, experts, and users. Auth includes JWT, Google OAuth hooks, email verification, and password reset flows.
Specifications
Core Challenges
- [1]Orchestrating a multi-step inference pipeline: Cloudinary upload → ML inference → output image re-upload → database persistence in a single request lifecycle
- [2]Implementing production-style auth (JWT, Google OAuth, email verification, password reset, refresh tokens) with centralized response handling and global exception filters
- [3]Designing modular NestJS domain architecture across 8+ entities (scans, predictions, reports, histories, treatments, experts, users) with paginated and filterable endpoints
Key Outcomes & Outcomes
- ✓Delivered end-to-end scan-to-prediction pipeline from upload to annotated result in a single user interaction
- ✓Implemented complete auth suite: JWT, Google OAuth, email verification, password reset, and refresh token rotation
- ✓Structured REST API surface with 30+ endpoints across modular clinical domains
- ✓Integrated YOLO inference service returning tumor class, confidence score, and bounding-box output image per scan