Empowering ASHA Workers with AI-Powered Health Screening
Offline-First • Multi-Disease • Rural India
Rural India faces critical healthcare challenges:
- Doctor Shortage: 1 doctor per 10,000+ people in villages
- Poor Connectivity: Limited internet access for telemedicine
- Late Diagnosis: Preventable diseases detected too late
- High Costs: Expensive diagnostic tests unavailable locally
Aasha.AI brings AI-powered health screening directly to villages through ASHA workers' smartphones:
- ✅ Works 100% Offline - No internet required
- ✅ Multi-Disease Screening - TB, Skin, Anemia, Maternal Health
- ✅ Instant Results - AI analysis in seconds
- ✅ Smart Referrals - Nearest PHC recommendations
- ✅ Patient Records - Offline-first with cloud sync
| Module | Input | AI Model | Risk Detection |
|---|---|---|---|
| TB Screening | 🎤 Cough Audio | Audio Classification | TB risk indicators |
| Skin Disease | 📷 Photo | Image Classification | Fungal, Eczema, Ringworm |
| Anemia Check | 📷 Palm/Eye Photo | Pallor Analysis | Hemoglobin estimation |
| Maternal Health | 📋 Questionnaire | Risk Scoring | Pregnancy danger signs |
| Symptom Triage | 📋 Symptoms List | Urgency Scoring | Care urgency level |
- Flutter 3.0+ installed
- Android Studio / VS Code
- Android device or emulator
# Clone the repository
git clone https://github.com/yourusername/aasha-ai.git
cd aasha-ai
# Install dependencies
flutter pub get
# Run the app
flutter runflutter build apk --releaseSplash Screen
│
▼
ASHA Dashboard
│
├── New Screening ──► Patient Registration ──► Select Module
│ │
│ ┌──────────────────────────────┼──────────────────────────────┐
│ │ │ │
│ ▼ ▼ ▼
│ TB Module Skin Module Anemia Module
│ (Record Cough) (Capture Photo) (Capture Palm/Eye)
│ │ │ │
│ └──────────────┬───────────────┘ │
│ ▼ │
│ AI Processing ◄─────────────────────────────────────────┘
│ │
│ ▼
│ Result Screen
│ (Risk + Recommendation)
│ │
│ ▼
│ Nearest PHC Map
│
└── Patient History ──► Filter by Risk ──► View Past Results
aasha_ai/
├── lib/
│ ├── main.dart # App entry point
│ ├── models/
│ │ ├── patient.dart # Patient data model
│ │ └── screening_result.dart # Screening result model
│ ├── screens/
│ │ ├── splash_screen.dart # Animated splash
│ │ ├── login_screen.dart # ASHA worker login
│ │ ├── home_screen.dart # Dashboard with modules
│ │ ├── patient_registration_screen.dart
│ │ ├── cough_recording_screen.dart
│ │ ├── skin_scan_screen.dart
│ │ ├── anemia_scan_screen.dart
│ │ ├── maternal_health_screen.dart
│ │ ├── symptom_triage_screen.dart
│ │ ├── ai_processing_screen.dart
│ │ ├── result_screen.dart
│ │ └── patient_history_screen.dart
│ ├── services/
│ │ ├── storage_service.dart # Hive local database
│ │ ├── tflite_service.dart # AI inference
│ │ ├── audio_service.dart # Audio recording
│ │ ├── image_service.dart # Camera capture
│ │ └── location_service.dart # PHC finder
│ ├── widgets/
│ │ ├── risk_indicator.dart # Risk level badges
│ │ ├── module_card.dart # Dashboard cards
│ │ ├── patient_card.dart # Patient list items
│ │ ├── action_button.dart # CTA buttons
│ │ └── symptom_checkbox.dart # Touch-friendly checkboxes
│ └── utils/
│ ├── constants.dart # Colors, styles, config
│ └── routes.dart # Navigation routes
├── assets/
│ ├── models/ # TFLite model files
│ ├── icons/ # App icons
│ └── images/ # Logos and illustrations
└── README.md
The app uses TensorFlow Lite for offline inference:
| Model | Size | Input | Output |
|---|---|---|---|
| TB Cough | ~2MB | Audio spectrogram | Risk probability |
| Skin Disease | ~3MB | 224x224 image | Disease class + confidence |
| Anemia | ~2MB | 224x224 palm/eye | Pallor level |
| Maternal Risk | ~500KB | Feature vector | Risk score |
Place your trained .tflite models in:
assets/models/
├── tb_cough.tflite
├── skin_disease.tflite
├── anemia_screen.tflite
└── maternal_risk.tflite
Note: The app includes mock predictions for demo purposes. Replace with real models for production.
- Flutter - Cross-platform mobile framework
- TensorFlow Lite - On-device ML inference
- Hive - Fast, lightweight local database
- Firebase - Optional cloud sync (when online)
- Geolocator - GPS for nearest PHC
┌─────────────────────────────────────────────────────────────┐
│ ASHA Worker Device │
├─────────────────────────────────────────────────────────────┤
│ ┌─────────────┐ ┌─────────────┐ ┌─────────────────────┐ │
│ │ Flutter │ │ TFLite │ │ Hive DB │ │
│ │ UI │◄─┤ Models │ │ (Patients, Results)│ │
│ └─────────────┘ └─────────────┘ └─────────────────────┘ │
│ │ │ │
│ └────────────────┬───────────────────┘ │
│ │ │
│ ┌─────▼─────┐ │
│ │ Sync │ │
│ │ Service │ │
│ └─────┬─────┘ │
└──────────────────────────┼───────────────────────────────────┘
│ When Online
┌──────▼──────┐
│ Firebase │
│ Firestore │
└─────────────┘
Add to android/app/src/main/AndroidManifest.xml:
<uses-permission android:name="android.permission.RECORD_AUDIO"/>
<uses-permission android:name="android.permission.CAMERA"/>
<uses-permission android:name="android.permission.ACCESS_FINE_LOCATION"/>
<uses-permission android:name="android.permission.ACCESS_COARSE_LOCATION"/>
<uses-permission android:name="android.permission.INTERNET"/>- Rural-Friendly UI - Large buttons, high contrast, simple navigation
- Minimal Text - Icon-based interface, easy for low-literacy users
- Multi-Language Ready - String resources extracted for i18n
- Low Resource - Optimized for low-end Android devices
- ✅ Complete multi-disease screening platform
- ✅ Offline AI inference
- ✅ Animated recording with waveform
- ✅ Step-by-step AI processing visualization
- ✅ Color-coded risk levels
- ✅ Nearest PHC with maps integration
- ✅ Patient history with filters
- ✅ Premium UI with gradients and animations
- Hindi/Odia voice guidance
- Emergency SMS alerts
- High-risk zone heatmap
- Government PHC API integration
- Diabetes screening module
- Eye disease detection
- Dental health screening
Built with ❤️ for rural India
MIT License - See LICENSE for details
🌟 Star this repo if Aasha.AI helps rural healthcare! 🌟