MediScanAI

Inspiration

Access to affordable and early healthcare screening is still a major challenge in India. People often ignore early symptoms due to cost, distance, or lack of awareness, which leads to serious complications later. We wanted to build a tool that makes preliminary health checks as simple as taking a selfie or recording a cough. This inspired us to create MediScanAI — an AI-driven health triage assistant available to anyone with a smartphone.

What it does

MediScanAI uses your phone’s camera and microphone to analyze:

Face

Eyes

Tongue

Skin

Cough audio

Using computer vision and machine learning models, it provides early indicators of conditions like:

Anemia

Jaundice

Dehydration

Skin abnormalities

Respiratory issues

The app generates a risk score, health summary, and personalized recommendations, making early detection easy and accessible.

How we built it

We developed MediScanAI using:

Frontend

React Native / Expo

Lottie animations

Glassmorphism UI

Backend

Node.js + Express

JWT-based authentication

AWS S3 / MinIO for file storage

ML Models

FastAPI-based Python microservice

Vision models for eye, tongue, skin, and face analysis

Audio classifier for cough patterns

Custom risk scoring engine

Database

MongoDB Atlas for user data and scan history

The system works on a modular pipeline: user scan → backend → ML microservice → results → app dashboard.

Challenges we ran into

Tuning ML models to work with varying lighting conditions and low-quality mobile cameras

Ensuring fast image/audio processing for smooth user experience

Designing a clean UI that feels futuristic but still simple

Integrating multiple scan types (face, eyes, tongue, skin, audio) into one workflow

Maintaining data privacy and secure upload handling

Time constraints during integration and debugging

Accomplishments that we're proud of

Successfully built a multi-scan AI health triage system

Achieved smooth, animated UI with a futuristic MedTech look

Integrated real-time scanning overlays

Implemented a working cough classifier and skin anomaly detector

Created a full end-to-end pipeline (app → backend → ML → results)

Made health screening more accessible for users

What we learned

How to integrate CV + audio ML models into a mobile app

Best practices in API design, token auth, and secure file handling

Importance of user experience in healthcare applications

How to structure modular AI services and microservice architectures

The potential of AI to transform preventive healthcare

Real-world challenges in building healthcare-focused solutions

What's next for MediScanAI

Adding heart rate and stress analysis using face video (rPPG)

Expanding dataset and improving model accuracy

Launching multi-language support for Indian users

Integrating with local clinics and telemedicine partners

Adding offline-mode scanning for rural areas

Developing a chatbot for symptom-based Q&A

Publishing the app on Play Store/App Store

Built With

• base44