🩺 HealthGuard AI — ICU Diabetes Prediction System

Table of Contents

• Overview
• Features
• Directory structure
• Technology stack
• Installation
• Usage
• Project architecture
• File details
• Contributing
• License
• Acknowledgments

---

Overview

HealthGuard AI is a clinical intelligence platform for predicting diabetes risk in ICU patients and for providing patient monitoring visualizations. It pairs a Next.js TypeScript frontend with MATLAB scripts for model development and explainability.

---

Features

• Diabetes risk prediction using ensemble models and SVM
• Real-time vital-sign monitoring (demo/mock data)
• Interactive dashboard and charts
• Explainability: feature importance and partial dependence plots
• Patient registry and risk gauges

---

Directory structure (high level)

MatHackathon/
├── health-guard-ai/       # Next.js frontend
│   ├── app/
│   ├── components/
│   ├── lib/
│   ├── model/
│   └── public/
├── model/                 # MATLAB model scripts and datasets
└── README.md              # This file

---

Technology stack

• Frontend: Next.js (App Router), React, TypeScript, Tailwind CSS
• Visuals: Recharts, react-gauge-chart
• ML / Models: MATLAB (Statistics & Machine Learning Toolbox)
• Dev tools: ESLint, PostCSS, Node.js

---

Installation

Prerequisites

• Node.js (recommended v20+)
• A package manager: npm, yarn, pnpm, or bun
• MATLAB (for model scripts; R2020b or newer recommended)

Frontend setup

1. Clone the repository using your Git client.
2. Change to the frontend folder: health-guard-ai
3. Install dependencies:

npm install
# or use your preferred package manager

4. Run the development server:

npm run dev

5. Open the application in your browser at the address your local dev server reports (usually http://localhost:3000).

Model setup (MATLAB)

1. Open MATLAB and set the working directory to the model folder.
2. Ensure that the dataset file diabetes_15_important_columns.csv is present in the model/datasets folder.
3. Run the preprocessing and training scripts:

run('data_prep.m')
run('train_model.m')
run('explainability.m')

---

Usage

Development

cd health-guard-ai
npm run dev

Production

npm run build
npm start

Frontend routes (examples)

• Login page
• Dashboard (patient registry and statistics)
• Patient analysis (individual vitals and charts)
• Risk checker (interactive form)

MATLAB scripts

• data_prep.m — data cleaning and preprocessing
• train_model.m — model training (bagged trees, SVM)
• explainability.m — feature importance and PDPs
• dashboard_concept.m — MATLAB GUI prototype (optional)

---

Project architecture (summary)

• Next.js frontend uses mock data for demos and displays results.
• MATLAB pipeline prepares data, trains models offline, and creates explainability plots.
• Frontend and MATLAB pieces are separate; model outputs can be exported and consumed by the frontend if desired.

---

File details (high level)

• Frontend:
  • components: StatsGrid, PatientTable, GlucoseChart, RiskGauge
  • lib: mockData (patient generator), utils
• Model:
  • MATLAB scripts and datasets (diabetes_15_important_columns.csv)

---

Contributing

Suggested workflow:

1. Fork the repository
2. Create a feature branch
3. Commit changes with clear messages
4. Open a pull request

Coding conventions:

• TypeScript with strict typing
• Functional React components and hooks
• Tailwind CSS utilities for styling
• MATLAB scripts follow standard MATLAB style guidelines

---

License

This project was created as a hackathon submission. Add a LICENSE file to specify terms when ready.

---

Acknowledgments

Thanks to contributors, anonymized clinical datasets used for prototyping, and open-source libraries used during development.

---