
MediScan AI is a complete health triage platform designed for Bangladesh. Users can describe symptoms via text or Bengali voice, upload medical scans (X-ray, MRI, CT, blood test images), and receive AI-powered preliminary diagnoses and risk assessments. Features include a dashboard, scan upload flow with drag-and-drop, real-time analysis results, diagnostic history, and a symptom checker with voice input. Built with Next.js 15 (frontend) and FastAPI (backend), using Gemini 2.5 Flash for multimodal analysis.
Integrating Bengali voice recognition with real-time transcription was complex due to limited libraries. Handling large medical image uploads (up to 10MB) and processing them through Gemini API while maintaining performance required careful optimisation. Managing authentication state across the dashboard and scan history also posed challenges.
Add support for more scan types (ultrasound, pathology slides), offline mode for rural areas, integration with local hospital APIs, and multi-language support (Chittagonian, Sylheti).