🏥 MedAI Suite

An AI-powered clinical diagnostic platform with an impressive accuracy of 92.7%

Features · Screenshots · Architecture · Quick Start · Algorithm

---

📋 Overview

MedAI Suite represents a cutting-edge medical diagnostic platform that utilizes TF-IDF weighted symptom matching to accurately identify diseases based on patient-reported symptoms across 41 conditions and 132 clinical indicators. The platform includes mental wellness assessments (PHQ-9), BMI tracking, AI-driven medical advice via Google Gemini, and the capability to generate clinical reports in PDF format.

Developed as a dual-release system: a standalone Python desktop application and a modern full-stack web application.

---

✨ Features

Module	Description	Technology
🔬 Diagnostic Engine	TF-IDF symptom matching with IDF rarity weighting, F-beta scoring, and penalties for missed symptoms	Python, pandas
🧠 Mental Wellness	PHQ-9 standardized depression screening with severity mapping (0–27)	React
📊 BMI & Nutrition	Age-specific dietary guidelines and a BMI calculator with reference charts	Recharts
🤖 AI Medical Advisor	Conversational health guidance powered by Google Gemini	Gemini API
📄 Clinical Reports	Export patient history as professionally formatted PDF documents	jsPDF
🚨 Emergency Directory	National emergency hotlines (112, 108, 104, 102, 1033)	React
📱 Responsive Design	Desktop sidebar that transitions to a mobile bottom navigation bar	CSS Grid

---

📸 Screenshots

Diagnostic Portal	Patient History

Nutrition & BMI	Mental Wellness (PHQ-9)

AI Medical Advisor	Emergency Contacts

---

🏗 Architecture

MedAI-Suite/
├── PYTHON/                  ← Release 1: Standalone Desktop Application
│   ├── askBot.py            # Decision tree diagnostic engine
│   ├── gptBot.py            # Gemini-powered chat interface
│   ├── mainWind.py          # Tkinter main window
│   ├── Data/                # Training datasets (10,000+ records)
│   └── MasterData/          # Symptom descriptions & precautions
│   
├── WEBDEV/                  ← Release 2: Full-Stack Web Platform
│   ├── backend/
│   │   └── app.py           # FastAPI + TF-IDF diagnostic engine
│   ├── frontend/
│   │   └── src/
│   │       ├── App.jsx      # Router & navigation
│   │       └── pages/       # 6 feature modules
│   ├── Data/                # 10,000+ synthesized medical records
│   ├── Screenshots/         # Application screenshots
│   └── start_dev.sh         # One-click development launcher
│   
├── .gitignore
└── README.md

---

🚀 Quick Start

Web Application (Release 2)

Prerequisites: Python 3.13+, Node.js 18+

# 1. Clone the repository
git clone https://github.com/sonararadhya/Healthcare_Chatbot.git
cd Healthcare_Chatbot/WEBDEV

# 2. Set up the backend
python3 -m venv venv
./venv/bin/pip install fastapi uvicorn google-generativeai python-dotenv pandas

# 3. Configure the API key
echo "GEMINI_API_KEY=your_key_here" > backend/.env

# 4. Set up the frontend
cd frontend && npm install && cd ..

# 5. Launch the application
./start_dev.sh

Backend: http://127.0.0.1:8000 · Frontend: http://127.0.0.1:5173

Python Desktop Application (Release 1)

cd PYTHON
pip install -r requirements.txt
python mainWind.py

---

🧬 Diagnostic Algorithm

The engine employs a TF-IDF Weighted Scoring methodology instead of traditional machine learning classifiers:

Why Not Machine Learning?

Issue with ML (Random Forest / Naive Bayes)	Our Solution
Inaccurate results with 1-2 symptoms (130+ features mostly zero)	Exact set-intersection matching
Non-deterministic outputs across multiple runs	100% deterministic results
Requires model retraining for new diseases	Simply add a CSV row
Black-box predictions	Fully explainable scoring

How It Works

For each disease in the knowledge base:

1. OVERLAP     = user_symptoms ∩ disease_symptoms
2. IDF WEIGHT  = Σ log(total_diseases / diseases_with_symptom)   ← rare symptoms count more
3. RECALL      = weighted_overlap / weighted_user_input           ← coverage of user symptoms
4. PRECISION   = weighted_overlap / weighted_disease_total        ← specificity to disease
5. F-BETA      = ((1 + β²) × recall × precision) / (β² × precision + recall)
6. MISS PENALTY = 1 - 0.7 × (missed_weight / input_weight)       ← penalize unexplained symptoms
7. FINAL SCORE = F-beta × miss_penalty

Benchmark

Metric	Score
Top-1 Accuracy	92.7% (38/41 diseases)
Top-3 Accuracy	100% (41/41 diseases)
Dataset	10,000 records, 41 diseases, 132 symptoms
Latency	< 5ms per query

---

🛠 Tech Stack

Backend: FastAPI · Python 3.13 · pandas · Google Gemini API
Frontend: React 19 · Vite · Recharts · Framer Motion · jsPDF · Lucide Icons
Data: 10,000+ synthesized clinical records · symptom severity mappings · precaution databases

---

_{Crafted with precision by sonararadhya}

---

📝 Last maintained: September 27, 2026 at 09:25 UTC