Alvira - Alzheimer's Early Detection Platform

Multi-modal AI system for early Alzheimer's disease detection using speech patterns, medical images, and symptom analysis.

Problem & Solution

Problem: Early Alzheimer's detection requires multiple diagnostic approaches and is often delayed due to limited access to specialized testing.

Solution: AI-powered platform that analyzes:

• Speech patterns via WaveNet neural network
• Medical images via computer vision
• Symptom descriptions via NLP

Technology Stack

Backend (Flask + AI)

• WaveNet Model: Speech pattern analysis for cognitive decline detection
• SVM Model: Symptom-based risk assessment
• OpenSMILE: Acoustic feature extraction (jitter, shimmer, MFCC)
• OpenCV: Medical image processing
• PyTorch: Deep learning inference

Frontend (React + Vite)

• Multi-modal Interface: Image, audio, and text input
• Real-time Analysis: Instant AI predictions
• Responsive Design: Mobile-friendly interface

Quick Start

# Install dependencies
npm install
cd backend && pip install -r requirements.txt --break-system-packages

# Start application
./start.sh

API Endpoints

• POST /api/predict - Multi-modal prediction
  • type: "image" | "audio" | "text"
  • data: Base64 encoded file or text

Models

• wn_model.pth - WaveNet for speech analysis
• svm_alzheimer_model.pkl - SVM for symptom analysis
• scaler.pkl, pca.pkl, label_encoder.pkl - Data preprocessing

Architecture

Input → Feature Extraction → AI Models → Risk Assessment → Recommendations
  ↓           ↓              ↓            ↓               ↓
Image     OpenCV         SVM Model    Risk Level     Personalized
Audio     OpenSMILE      WaveNet      Confidence     Guidance
Text      NLP            NLP          Predictions    Actions

Performance

• Audio Analysis: 3-class classification (Healthy/Alzheimer's/Parkinson's)
• Image Analysis: Medical image feature extraction
• Text Analysis: Symptom parsing and risk scoring
• Response Time: < 5 seconds for all modalities

Deployment

# Backend
cd backend && python3 app.py

# Frontend
npm run build && npm run preview

License

MIT License