HealthShield

Authentication Page
Dashboard page
Medical Assessment Page

Inspiration

Many people do not know when they are about to fall sick. Most people ignore early symptoms or misunderstand what their body is trying to tell them. This often leads to late treatment, higher medication costs or probably death. We were inspired by only one question: What if people could know they were likely to fall sick before it happens?

What it does

**HealthShield is a web app that allows users to describe how they feel and the symptoms they see using text or a voice note for those who do not know how to express themselves with words. The system analyzes their input using llama-3 AI. It, being trained on past data of over 18 diseases and their equivalent symptoms, identifies patterns and possible health risks for the user by comparing and analyzing them thoroughly, which then predicts possible illness the user may be at risk of. Finally, it suggests preventive steps and base care recommendation based on the input data.

How we built it

This solution works as a full website. Frontend: React.js and Framer-motion for a fast and interactive web UI. Backend: Python Flask for fast integration with AI, including data processing and predictions. AI Engine: Llama-3 for context-aware conversation on health basis and Whisper-v3-turbo for lightning speed speech recognition in real-time. AI Classifier: Used Random Forest Classifier to train the data model to predict health risks based on symptoms provided.

Challenges we ran into

Our biggest challenge was getting accurate and valid medical data. Medical records are often hard to access and sometimes incomplete. Accurate high-quality data depends on high-quality data.

Accomplishments that we're proud of

Data-Driven Personalized Health AI: We were able to find a free medical data available on Kaggle. We used it to train our model and it performed well in most diseases. It could predict risks for both young and old. We also focused on preventive guidance instead of making final medical diagnosis.