1 / 12
← → Arrow Keys to Navigate
AI FOR FUTURE WORKFORCE BOOTCAMP 2026

Multi-Disease
Prediction System

Advanced ML models for early detection of Diabetes · Heart Disease · Kidney Disease · Alzheimer's

Team HealthMatrix
Institution Chandigarh University
Program MCA – Data Science
4Disease Models
100msInference Speed
Full StackReact + Flask
LiveDeployed App
🎤 Speaker Notes: "Good [morning/afternoon]. I'm [Your Name], and today I'll be presenting HealthMatrix — a multi-disease prediction system that uses machine learning to predict four major diseases from a single platform. This project was built as part of the AI for Future Workforce Bootcamp 2026 at Chandigarh University."
02
THE CHALLENGE

Problem Statement

🏥

Fragmented Diagnosis

Existing systems are disease-specific — no single platform covers multiple conditions simultaneously

💸

High Cost & Delay

Multiple specialists, tests, and appointments cause delays and financial burden for patients

🌍

Limited Accessibility

Rural and underserved populations lack access to timely diagnostic tools and healthcare professionals

⚠️

Late Detection

Diseases like Diabetes, Heart Disease, Kidney Disease & Alzheimer's worsen significantly without early intervention

422M+Diabetics Worldwide
17.9MCardiovascular Deaths/yr
850MKidney Disease Patients
55MAlzheimer's Patients
🎤 Speaker Notes: "The core problem is that healthcare diagnosis is fragmented, expensive, and slow. People suffering from chronic diseases like diabetes or heart disease often don't get diagnosed until it's too late. Existing ML tools are siloed — they only predict one disease. We asked: why not build one unified platform that covers all major diseases?"
03
GOALS

Objectives

01

Build a Unified Multi-Disease Platform

Predict Diabetes, Heart Disease, Kidney Disease, and Alzheimer's from a single web application

🎯
02

Train High-Accuracy ML Models

Use Random Forest & XGBoost on validated clinical datasets with proper preprocessing pipelines

🤖
03

Deliver Real-Time Predictions

Provide instant results (<100ms inference) through a clean, accessible React-based UI

⚡
04

Enable Early Detection & Decision Support

Assist patients and healthcare professionals in making informed, data-driven decisions early

🩺
05

Deploy a Scalable, Production-Ready System

Full-stack deployment on Vercel (frontend) and Render (backend) for global accessibility

🚀
🎤 Speaker Notes: "Our objectives were clear: build one platform, train multiple models, and make it accessible to everyone. We didn't just want a prototype — we wanted a fully deployed, production-ready system that anyone with a browser could use. The key differentiator is the unified approach — four diseases, one platform, instant results."
04
OUR ANSWER

Proposed Solution

One Platform. Four Diseases. Instant Results.

A web-based intelligent clinical decision support system powered by ensemble machine learning — accessible to anyone, anywhere, in seconds.

🩸

Diabetes

Random Forest

Glucose, BMI, Insulin, Age parameters

❤️

Heart Disease

Random Forest

Cholesterol, BP, ECG, Chest Pain type

🫘

Kidney Disease

XGBoost + KNN

Creatinine, GFR, Hemoglobin, Albumin

🧠

Alzheimer's

Random Forest

MMSE Score, CDR, Age, Brain Volume

User Inputs Parameters
→
Flask API Processes
→
ML Model Predicts
→
Instant Result
🎤 Speaker Notes: "Our solution is HealthMatrix — a web application where a user selects a disease, enters their medical parameters, and gets an instant prediction. The backend is a Flask API that routes the input to the appropriate trained ML model. We used Random Forest for three diseases and XGBoost for kidney disease due to its superior handling of missing values."
05
DESIGN

System Architecture

FRONTEND — Vercel
⚛️ React App
🎨 Disease UI Forms
🔗 Axios HTTP
↕ REST API (JSON)
BACKEND — Render (Flask)
🐍 Flask API
⚙️ StandardScaler + KNN Imputer
🔄 sklearn Pipeline
↕ Model Inference
ML MODELS — Trained & Serialized (.pkl)
🩸 Diabetes RF
❤️ Heart RF
🫘 Kidney XGB
🧠 Alzheimer's RF
📌 Placeholder: Insert your system architecture diagram here for maximum visual impact
🎤 Speaker Notes: "The architecture is a classic three-tier system. The React frontend communicates with the Flask backend via REST API calls. The backend preprocesses the input using StandardScaler and KNN Imputer, then passes it to the appropriate serialized model. The result is returned as JSON and displayed instantly on the frontend. Frontend is on Vercel, backend on Render."
06
TOOLS & FRAMEWORKS

Technology Stack

⚛️ Frontend
React.jsComponent-based UI
JavaScriptCore logic
AxiosHTTP requests
VercelDeployment
🐍 Backend
FlaskREST API server
Python 3.xCore language
PickleModel serialization
RenderCloud deployment
🤖 Machine Learning
Scikit-learnRandom Forest, Pipeline
XGBoostKidney disease model
Pandas / NumPyData processing
StandardScalerFeature normalization
📊 Data & Datasets
Pima IndiansDiabetes dataset
ClevelandHeart disease dataset
UCI CKDKidney disease dataset
OASISAlzheimer's dataset
🎤 Speaker Notes: "For the tech stack, we chose React for the frontend because of its component reusability and fast rendering. Flask was chosen for the backend due to its lightweight nature and easy integration with Python ML libraries. For ML, we used Scikit-learn's Random Forest and XGBoost. All models are serialized with Pickle and loaded at runtime for fast inference."
07
HOW IT WORKS

Methodology & ML Pipeline

1

Data Collection

4 public clinical datasets: Pima Indians, Cleveland, UCI CKD, OASIS

2

Preprocessing

KNN Imputer for missing values · StandardScaler for normalization · Feature selection

3

Model Training

Random Forest (Diabetes, Heart, Alzheimer's) · XGBoost (Kidney) · sklearn Pipelines

4

Evaluation

Accuracy, Precision, Recall, F1-Score on test split · Cross-validation

5

Serialization & Deployment

Models saved as .pkl · Flask API serves predictions · React UI displays results

Why Random Forest? ✓ Handles non-linearity ✓ Robust to overfitting ✓ Works well with tabular medical data ✓ Feature importance built-in
🎤 Speaker Notes: "Our ML pipeline follows five stages. First, we collected four publicly available clinical datasets. Then we preprocessed them — handling missing values with KNN Imputer, normalizing features with StandardScaler. We trained Random Forest models for three diseases and XGBoost for kidney disease. After evaluation, models were serialized and deployed via Flask. The entire pipeline is reproducible from the GitHub repo."
08
WHAT MAKES IT SPECIAL

Key Features

🔬

4-in-1 Disease Prediction

Single platform covering Diabetes, Heart Disease, Kidney Disease & Alzheimer's — no switching between tools

UNIQUE
⚡

Real-Time Inference

Sub-100ms prediction response via optimized Flask API with pre-loaded models

🧹

Robust Preprocessing

KNN Imputer handles real-world missing data; StandardScaler ensures consistent input

📱

Responsive UI

Clean React interface accessible on desktop and mobile — no installation required

☁️

Cloud Deployed

Live on Vercel + Render — globally accessible, zero downtime, 99% uptime

🔄

Modular Architecture

Each disease module is independent — easy to add new diseases or update models

🎤 Speaker Notes: "The standout feature is the 4-in-1 approach — no other student project in this space covers all four diseases in one deployed application. The system is live, not just a prototype. The modular architecture means we can easily add more diseases in the future. The preprocessing pipeline handles real-world messy data, which is critical for medical applications."
09
LIVE DEMO

Application Screenshots

🖥️
[ Insert Landing Page / Dashboard Screenshot ]
MedPredict.AI — Main Dashboard with Disease Cards
🩸
[ Diabetes Prediction Form ]
Input form with glucose, BMI, insulin fields
❤️
[ Heart Disease Prediction ]
Cardiovascular risk assessment form
✅
[ Prediction Result Screen ]
Disease / No Disease result with confidence
🧠
[ Alzheimer's / Kidney Module ]
MMSE score and clinical parameter inputs
💡 Tip: Replace placeholders with actual screenshots from your deployed app at your-app.vercel.app
🎤 Speaker Notes: "Here you can see the application interface. The landing page shows all four disease modules as cards. When a user selects a disease, they're taken to a form where they enter their medical parameters. After submission, the result is displayed instantly — either 'Disease Detected' or 'No Disease Detected'. The UI is clean, modern, and designed to be non-intimidating for non-technical users."
10
PERFORMANCE

Results & Model Output

🩸 Diabetes
Random Forest
DatasetPima Indians (768 rows)
AlgorithmRandom Forest Classifier
PreprocessingStandardScaler
OutputDiabetic / Non-Diabetic
❤️ Heart Disease
Random Forest
DatasetCleveland (303 rows)
AlgorithmRandom Forest Classifier
PreprocessingStandardScaler
OutputDisease / No Disease
🫘 Kidney Disease
XGBoost + KNN
DatasetUCI CKD (400 rows)
AlgorithmXGBoost + KNN Imputer
PreprocessingKNN Imputer + Scaler
OutputCKD / Not CKD
🧠 Alzheimer's
Random Forest
DatasetOASIS Dataset
AlgorithmRandom Forest Classifier
PreprocessingStandardScaler
OutputDemented / Non-Demented
💡 Key Finding: Feature scaling significantly improved accuracy across all models. Random Forest performed consistently well on all tabular medical datasets. XGBoost was essential for the kidney dataset due to heavy missing values.
🎤 Speaker Notes: "Our models were trained on well-known public clinical datasets. The key finding was that feature scaling dramatically improved accuracy. Random Forest was our go-to algorithm because it handles non-linearity and is robust to overfitting. For kidney disease, we used XGBoost combined with KNN Imputer because the UCI CKD dataset had significant missing values that needed intelligent imputation."
11
GROWTH

Challenges & Learnings

⚡ Challenges Faced

🔴
Missing Data in Kidney Dataset

UCI CKD had significant missing values — solved with KNN Imputer for intelligent filling

🔴
Class Imbalance

Unequal disease/no-disease samples — addressed through careful model selection and evaluation metrics

🔴
CORS & API Integration

Cross-origin issues between React (Vercel) and Flask (Render) — resolved with Flask-CORS configuration

🔴
Model Consistency

Ensuring the same preprocessing pipeline at training and inference time — solved with sklearn Pipelines

✨ Key Learnings

🟢
End-to-End ML Deployment

Learned the full pipeline from data cleaning to production deployment on cloud platforms

🟢
Full-Stack Integration

Gained hands-on experience connecting React frontend with Python Flask backend via REST APIs

🟢
Real-World Data Challenges

Medical datasets are messy — preprocessing is as important as model selection

🟢
Modular System Design

Building independent modules for each disease made the system scalable and maintainable

🎤 Speaker Notes: "The biggest technical challenge was the kidney disease dataset — it had significant missing values that couldn't just be dropped. We solved this with KNN Imputer. Another challenge was ensuring the preprocessing pipeline was identical at training and inference time — sklearn Pipelines solved this elegantly. The biggest learning was that real-world ML is 80% data engineering and 20% model selection."
12
WRAP UP

Conclusion & Future Scope

✅ What We Built

✓ Unified multi-disease prediction platform — 4 diseases, 1 app
✓ Production-deployed full-stack application (React + Flask)
✓ Robust ML pipeline with proper preprocessing and model serialization
✓ Real-time predictions accessible to anyone, anywhere
✓ Scalable, modular architecture ready for extension
"Early detection saves lives. Technology should make it accessible to all."

🚀 Future Scope

01
Deep Learning Models

Upgrade to CNN/LSTM for image-based diagnosis (X-rays, MRI scans)

02
More Disease Coverage

Add Cancer, Liver Disease, Thyroid, and Stroke prediction modules

03
EHR Integration

Connect with Electronic Health Records for real-time patient data

04
Doctor Dashboard

Professional interface for healthcare providers with patient history

05
Mobile App

React Native app for offline-capable predictions in rural areas

⭐ GitHub Repository
|
Team HealthMatrix · Chandigarh University · MCA Data Science · 2026
🎤 Speaker Notes: "To conclude — HealthMatrix is a fully deployed, production-ready multi-disease prediction system. We've demonstrated that ML can be applied to real healthcare problems with proper engineering. In the future, we plan to add deep learning for image-based diagnosis, more disease modules, and EHR integration. The goal is to make early disease detection accessible to everyone, regardless of location or income. Thank you — I'm happy to take any questions."
📝 Scripts

Presentation Scripts

⏱️ 1-Minute Script (Elevator Pitch)

"Hi, I'm [Your Name] from Chandigarh University, MCA Data Science. I built HealthMatrix — a multi-disease prediction system that uses machine learning to predict four major diseases: Diabetes, Heart Disease, Kidney Disease, and Alzheimer's — all from a single web platform.

The problem is that existing diagnostic tools are fragmented, expensive, and inaccessible. Our solution is a React + Flask full-stack application where users enter their medical parameters and get instant predictions in under 100 milliseconds.

We trained Random Forest and XGBoost models on validated clinical datasets, deployed on Vercel and Render. The system is live, scalable, and designed to support early detection — which is the most effective way to fight chronic disease. Thank you."

⏱️ 3-Minute Script (Full Presentation)

[Slide 1 — Title] "Good [morning/afternoon]. I'm [Your Name], and today I'll present HealthMatrix — a multi-disease prediction system built for the AI for Future Workforce Bootcamp 2026."

[Slide 2 — Problem] "The problem we're solving is fragmented healthcare diagnosis. Today, if you want to check for diabetes, heart disease, kidney disease, and Alzheimer's, you need four different specialists, multiple tests, and weeks of waiting. This is expensive, slow, and inaccessible — especially in rural areas. Over 422 million people have diabetes, 17.9 million die from cardiovascular disease annually. Early detection is critical."

[Slide 3-4 — Solution] "Our solution is HealthMatrix — one platform, four diseases, instant results. Users enter their medical parameters, and our ML models return a prediction in under 100 milliseconds. We used Random Forest for three diseases and XGBoost for kidney disease."

[Slide 5-6 — Architecture & Stack] "The system is a three-tier architecture: React frontend on Vercel, Flask backend on Render, and four serialized ML models. We used Scikit-learn, XGBoost, Pandas, and StandardScaler for the ML pipeline."

[Slide 7-8 — Methodology & Features] "Our ML pipeline covers data collection, preprocessing with KNN Imputer and StandardScaler, model training, evaluation, and deployment. The standout feature is the 4-in-1 approach — no other tool in this space covers all four diseases in one deployed application."

[Slide 10-11 — Results & Challenges] "Key finding: feature scaling significantly improved accuracy. The biggest challenge was the kidney dataset's missing values — solved with KNN Imputer. We also learned that real-world ML is 80% data engineering."

[Slide 12 — Conclusion] "To conclude, HealthMatrix is a fully deployed, production-ready system. Future plans include deep learning for image-based diagnosis, more disease modules, and EHR integration. Our goal: make early disease detection accessible to everyone. Thank you — happy to take questions."