🧠 TeleNeuro

AI-Powered Telemedicine Platform for Early Alzheimer's Detection

---

📖 Overview

TeleNeuro is a full-stack Flutter telemedicine application that connects patients and neurologists, combined with an on-device AI diagnostic engine for early Alzheimer's detection from brain MRI scans.

The app runs a fine-tuned Vision Transformer (ViT-B/16) model — converted to TFLite and bundled directly into the app — so MRI classification happens entirely on-device, without sending medical images to any external inference server.

Patients can upload an MRI scan, receive an instant AI-generated diagnosis report (PDF), book appointments with doctors, chat in real time, and securely share their reports — all within one app.

---

✨ Features

👤 Patient Side

• Secure sign-up / login (Gmail-restricted email validation, 18+ age gate)
• Browse & search doctors by specialization, sorted by rating
• Book appointments with date/time picker + problem description
• Real-time chat with doctors after appointment acceptance
• AI MRI Diagnosis — upload a scan, get instant stage classification + confidence score
• Auto-generated PDF diagnostic report, saved to personal Lab Reports library
• Share MRI reports selectively with doctors (only those with an accepted appointment)
• In-app notifications (bookings, messages, session updates, shared reports)
• Editable profile with photo upload (Supabase-backed)
• Secure account deletion with password re-authentication

🩺 Doctor Side

• Separate doctor registration flow (specialization, qualification, hospital, experience)
• Dashboard with live stats: total patients, pending requests, rating, experience
• Accept / decline appointment requests in real time
• Dedicated chat inbox per patient, with "End Session" and delete conversation controls
• View shared MRI reports & clinical summaries from linked patients
• "My Patients" directory built from appointment history
• Patient reviews & ratings section on public profile
• In-app doctor guide / FAQ

🔐 Shared / Core

• Dual-role authentication routing (Patient vs Doctor) via Firestore role checks
• Firestore-backed real-time notification engine (deduplicated per event)
• Appointment-scoped chat with visibility rules (auto-hides chat if either account is deleted)
• Reusable UI kit: searchable dropdowns, animated buttons, custom date pickers, badge icons

---

📸 Screenshots

Patient Dashboard	MRI Reports	Appointments

Doctor Dashboard	Doctor Profile	User Guide

---

🧠 AI Model

Architecture	Vision Transformer — ViT-B/16 (torchvision)
Task	4-class classification: NonDemented, VeryMildDemented, MildDemented, ModerateDemented
Training framework	PyTorch (mixed-precision / AMP, early stopping)
Datasets	Alzheimer's MRI Dataset + OASIS Brain MRI Dataset
Export pipeline	PyTorch → ONNX (opset 14) → TFLite (via ai-edge-torch)
On-device runtime	tflite_flutter
Input	224×224×3 RGB tensor, ImageNet-normalized

Preprocessing pipeline (must match training exactly): Resize(256) → CenterCrop(192) → Grayscale→RGB → Resize(224) → Normalize(ImageNet mean/std) → CHW

The full training/export notebook is documented in final-fyp.ipynb.

---

🛠️ Tech Stack

Layer	Technology
Frontend	Flutter (Dart)
Authentication	Firebase Auth
Database	Cloud Firestore (real-time)
File Storage	Supabase Storage (MRI PDFs, images, profile photos)
AI Inference	TensorFlow Lite (on-device)
PDF Generation	pdf package
State/Local Cache	shared_preferences

---

📂 Project Structure

lib/
├── Splash/                       # Onboarding / splash screens
├── Profile Side/                 # Role selection (Patient / Doctor)
├── Patient Side/
│   ├── Auth/                     # Patient login, signup, portal
│   ├── Screens/                  # Dashboard, appointments, chat, MRI upload, reports...
│   └── services/                 # MRI report service
├── Doctor Side/
│   ├── Auth/                     # Doctor login, signup, portal
│   ├── Screens/                  # Dashboard, appointments, chat, patients, reports...
│   └── Widgets/                  # Doctor-specific cards
├── Widgets/                      # Shared reusable widgets (avatars, dropdowns, pickers...)
├── services/                     # Shared services (auth, notifications, validation...)
├── data/                         # Static option lists (specializations, qualifications)
├── firebase_options.dart
├── supabase_config.dart
└── main.dart

assets/
├── screen1.png ... screen4.jpg   # Onboarding illustrations
└── vit_fyp_direct.tflite         # AI model (NOT committed — see setup below)

---

🚀 Getting Started

Prerequisites

• Flutter SDK ^3.9.2
• A Firebase project (Auth + Firestore enabled)
• A Supabase project (Storage bucket)

1. Clone the repository

git clone https://github.com/<your-username>/teleneuro.git
cd teleneuro

2. Install dependencies

flutter pub get

3. Configure Firebase

• Create a Firebase project → enable Email/Password Auth and Firestore.
• Run:

  flutterfire configure

This regenerates lib/firebase_options.dart and android/app/google-services.json for your own project.

4. Configure Supabase

• Create a free project at supabase.com.
• Create a public Storage bucket (default name used in this project: mri-reports).
• Run the SQL policies in supabase_mri_reports_storage.sql inside the Supabase SQL Editor.
• Update lib/supabase_config.dart with your own Project URL and anon key:

  static const String supabaseUrl = 'YOUR_SUPABASE_URL';
  static const String supabaseAnonKey = 'YOUR_SUPABASE_ANON_KEY';
  static const String supabaseBucket = 'mri-reports';

5. ⚠️ AI Model Setup (required — not included in this repo)

The .tflite model file is too large for a normal Git push (GitHub blocks files over 100MB), so it is hosted separately instead of being committed to this repository.

Steps to set it up:

1. Download the model file from:(GitHub Release)
2. Place the downloaded file at: assets/vit_fyp_direct.tflite
3. Confirm it's registered in pubspec.yaml (already included in this repo):

   flutter:
     assets:
       - assets/vit_fyp_direct.tflite

4. Run flutter pub get again, then rebuild the app.

Model training and export details are documented in final-fyp.ipynb.

6. Run the app

flutter run

---

🔒 Security Notes

• Never commit real Firebase/Supabase keys to a public repository — use environment-specific config or --dart-define for production.
• Firestore & Storage rules should restrict read/write access appropriately before going to production (this project uses permissive anon policies for development — see supabase_mri_reports_storage.sql).

---

📜 License

This project is licensed under the MIT License — see the LICENSE file for details.

---

👨‍💻 Author

Malik Anas Ahmed 📧 anasahmed.appdev@gmail.com 🔗 LinkedIn

---

Made with ❤️ using Flutter, Firebase, Supabase & PyTorch