coderush-hackathon-project
Disease Predictor v2
Explainable disease risk workflow for fast, judge-friendly triage demos
An explainable disease risk workflow that turns symptoms, report text, and screening signals into triage guidance, follow-up planning, and a judge-ready demo.
This project is designed to be judged quickly and understood instantly:
- Paste report text or upload a file
- Get disease suggestions and ranked hospital presets
- Review smart triage guidance
- Generate a short follow-up plan
- Explore the supporting ML and Bayesian outputs
One-line pitch
We help users and judges understand disease risk faster by converting scattered clinical clues into one explainable decision-support flow.
Why this project stands out
- Solves a real workflow problem: fragmented health signals across symptoms, reports, and images
- Explains results instead of hiding them behind a black box
- Uses multiple signals: ML, Bayesian inference, AI guidance, and report parsing
- Feels like a product, not a notebook
- Has a clean judge-ready UI and a deployment-friendly structure
Core features
1. Symptom-based prediction
- Select a disease and symptoms
- Generate ML probability output
- Review missing symptoms and risk level
- Download the result summary
2. AI report intake
- Paste free text or upload report files
- Extract age, glucose, temperature, and BP patterns when available
- Rank likely disease presets from
hospital_data.csv - Auto-fill the form from the report analysis
3. Smart triage guardrail
- Detect red-flag patterns
- Classify cases as routine, urgent, or emergency
- Give a short, actionable safety message
4. Follow-up automation
- Generate a concise 3-day care plan
- Turn the result into next-step guidance
- Keep the demo useful even without Gemini enabled
5. Presentation-ready UI
- Hero section with a clear problem statement
- Judge-friendly explanation blocks
- Loading states, toast feedback, and polished cards
- Mobile-friendly layout
Tech stack
- Backend: Flask
- Frontend: Bootstrap 5, HTML, CSS, JavaScript
- AI support: Google Gemini integration
- ML and logic: custom prediction flow and Bayesian calculator
- Data: CSV-backed disease presets
- File handling: report text, PDF, and image intake support where available
Demo flow
- Open the homepage
- Paste a report or upload a file
- Click Analyze report
- Review the suggested disease, top presets, and triage output
- Apply the suggestions to auto-fill the form
- Run prediction
- Generate the follow-up plan
60-second pitch
“Healthcare data is often fragmented. A patient’s symptoms, report text, and screening signals may all exist, but they rarely come together in one place. Disease Predictor v2 solves that by combining symptom-based prediction, Bayesian reasoning, AI report intake, triage guardrails, and follow-up automation into one explainable workflow. Instead of showing a black-box result, the app shows why the result appeared, ranks likely disease presets from report text, flags urgent cases, and generates a short next-step plan. The goal is not to replace a doctor. The goal is to reduce friction, improve clarity, and make early risk screening easier to understand and demo.”
3-slide demo structure
- Problem and solution
- Fragmented symptom and report interpretation
- One explainable workflow that unifies intake, prediction, and triage
- Live product demo
- Paste report text
- Show ranked presets and triage output
- Apply suggestions and run prediction
- Impact and differentiation
- Explainability instead of black-box output
- Follow-up automation instead of a dead-end score
- Deployment-friendly architecture for real demos
Submission assets to include
If you are submitting this to a hackathon, include these visuals in your final package:
- Homepage hero and problem statement
- Report intake with ranked preset suggestions
- Triage result showing routine, urgent, or emergency classification
- Follow-up plan output
- A 30 to 60 second demo video
Quick start
python -m pip install -r requirements-minimal.txt
python run.py
Open:
http://127.0.0.1:5000/
Local setup
Recommended
python -m venv venv
venv\Scripts\activate
python -m pip install -r requirements-minimal.txt
python run.py
Optional full install
If you want the broader dependency set for local development, use the main requirements file instead.
python -m pip install -r requirements.txt
Environment variables
For Gemini-powered responses, set:
GEMINI_API_KEY=your_api_key_here
The app still works without Gemini, but AI-enhanced recommendations will be limited.
Project structure
Disease-predictor-v2/
├── run.py
├── requirements.txt
├── requirements-minimal.txt
├── hospital_data.csv
├── frontend/
│ ├── templates/
│ └── static/
├── backend/
│ ├── routes/
│ ├── models/
│ ├── utils/
│ ├── templates/
│ └── static/
├── README.md
└── docs/
Judge talking points
- “This is an explainable disease risk workflow, not just a prediction model.”
- “The app combines report parsing, Bayesian reasoning, and AI follow-up in one product demo.”
- “It is built to reduce manual triage friction and make the result easy to understand.”
- “The UI is presentation-ready and the architecture supports deployment.”
Why judges remember this
- It solves a specific workflow pain point instead of showing raw model output only
- It explains the result in plain language
- It shows multiple AI-adjacent layers working together in one flow
- It is easy to demo live in under two minutes
- It looks polished enough to trust at first glance
Important note
This project is for educational and demonstration purposes only.
It is not a medical device and must not be used for real-world diagnosis or treatment decisions.
Always consult a qualified healthcare professional for medical advice.
License
Add your preferred hackathon or open-source license here if needed.
Next improvements for a stronger submission
- Add one short recorded demo video
- Add a metrics section with latency or time-saved claims
- Add screenshots for the homepage, triage, and follow-up flow
- Add a one-paragraph “why we built this” section for judges