⚕️ From scan upload → MobileNetV2 AI inference → 4-role clinical review → encrypted PDF report delivery — fully automated in one platform.

Medical Disclaimer: CSSS is an AI-assisted screening tool designed to support qualified medical professionals. All AI predictions require review by a licensed physician before any clinical decision is made.

🚀 Quick Start  •  🏗️ Architecture  •  🧠 AI Pipeline  •  📸 Screenshots  •  🔌 API  •  👥 Team  •  ☕ Support

Traditional Diagnostics   →   Manual review, paper reports, siloed roles, slow turnaround
CSSS                      →   AI pre-screening, automated PDF reports, unified 4-role platform

89.51% Accuracy MobileNetV2 on 217,875 medical images across 6 disease classes

4-Role Workflow Patient → Doctor → Pharmacist → Admin clinical pipeline

Auto PDF Reports WeasyPrint + Jinja2 professional diagnostic reports emailed automatically

Enterprise Security JWT + bcrypt + OTP 2FA with full role-based access control

Clinical Scan Support System (CSSS) is a production-grade, full-stack AI medical imaging platform built as a Final Year Project at Saveetha Engineering College. It automates the complete diagnostic pipeline — from patient scan upload through real-time MobileNetV2 inference, structured multi-role clinical review, to professional PDF diagnostic report generation and encrypted email delivery.

🎓 Academic Supervisor: Ms. V. Swedha & Dr. Selvakumar R, Saveetha Engineering College 📅 Academic Year: 2024–2025 📰 Research: Published in IEEE Conference & Journal 🏛️ Institution: Saveetha Engineering College, Chennai

Traditional diagnostic workflows are manual, siloed, and slow — patients wait days for reports that require multiple sign-offs across departments with no unified tracking. CSSS replaces this with a single platform where every stakeholder — patient, doctor, pharmacist, and admin — has a purpose-built dashboard, enforced RBAC, and a clearly defined role in an automated pipeline that ends with an encrypted PDF report in the patient's inbox.

                    ┌──────────────────────┐
                    │   Patient Uploads     │
                    │   Chest X-Ray / MRI  │
                    └──────────┬───────────┘
                               │
                    ┌──────────▼───────────┐
                    │  FastAPI REST API     │  ◄── JWT Auth + Role Guard
                    │  Port 8000           │
                    └──────────┬───────────┘
                               │
               ┌───────────────▼────────────────┐
               │        AI Inference Engine      │
               │  MobileNetV2 (224×224×3 input) │
               │  → 6-class softmax output       │
               │  → Confidence score + label     │
               └───────────────┬────────────────┘
                               │
          ┌────────────────────▼─────────────────────┐
          │           4-Role Clinical Pipeline        │
          │  Doctor Review → Pharmacist → Admin OK   │
          └────────────────────┬─────────────────────┘
                               │
                    ┌──────────▼───────────┐
                    │  WeasyPrint + Jinja2 │  ◄── PDF generation
                    │  PDF Report Engine   │
                    └──────────┬───────────┘
                               │
                    ┌──────────▼───────────┐
                    │  Gmail SMTP Service  │  ◄── Encrypted email delivery
                    │  → Patient Inbox     │
                    └──────────────────────┘

Input (224 × 224 × 3)
        ↓
MobileNetV2 — ImageNet pretrained (frozen feature extractor)
        ↓
GlobalAveragePooling2D
        ↓
Dense(256, relu)
        ↓
Dropout(0.4)
        ↓
Dense(6, softmax)  ← COVID · Lung_Opacity · NIH_MERGED · Normal · Sick · Viral_Pneumonia
        ↓
predicted_class + confidence_score + all_class_probabilities

optimizer  = Adam(learning_rate=1e-4)
loss       = categorical_crossentropy
img_size   = (224, 224)
batch_size = 16
epochs     = 15
callbacks  = [
    EarlyStopping(monitor="val_loss", patience=5, restore_best_weights=True),
    ReduceLROnPlateau(monitor="val_loss", factor=0.5, patience=3)
]
split      = 70% train / 15% val / 15% test

Training Curve	Confusion Matrix

🔑 Auth Flow

Login	Register	Report Email Received (Auto-Generated PDF)

👤 Patient

Patient Dashboard	Scan Upload Interface

🩺 Clinical Staff

Doctor Dashboard	Pharmacist Dashboard	Admin Dashboard

💬 AI Chatbot

Chatbot Interface

Projectwork2/
│
├── backend/                          # FastAPI REST API
│   ├── models/
│   │   ├── user.py                   # User (id, name, email, role, password)
│   │   ├── schema.py                 # Scan (id, patient_id, prediction, status...)
│   │   └── otp.py                    # OTPRecord (email, otp, expires_at, used)
│   │
│   ├── routers/
│   │   ├── auth_router.py            # POST /auth/register, /auth/login
│   │   ├── patient.py                # POST /patient/upload, GET /patient/status/{id}
│   │   ├── doctor.py                 # GET /doctor/pending, POST /doctor/analyze/{id}
│   │   ├── pharmacist.py             # GET /pharmacist/queue, POST /pharmacist/complete/{id}
│   │   ├── admin.py                  # GET /admin/pending, POST /admin/approve/{id}
│   │   ├── otp.py                    # POST /otp/send, /otp/verify
│   │   ├── chatbot.py                # POST /chatbot/
│   │   └── reports.py                # GET /reports/pdf/{scan_id}
│   │
│   ├── security/
│   │   ├── jwt_handler.py            # Token creation + verification
│   │   ├── password.py               # bcrypt hash / verify
│   │   └── role_guard.py             # require_role() dependency factory
│   │
│   ├── services/
│   │   ├── ai_service.py             # TensorFlow MobileNetV2 inference
│   │   └── email_service.py          # SMTP: OTP + PDF attachment delivery
│   │
│   ├── templates/
│   │   └── report_template.html      # Jinja2 HTML → WeasyPrint PDF
│   │
│   ├── config.py / database.py / main.py
│   ├── init_db.py                    # Create all DB tables
│   └── seed_db.py                    # Seed 4 demo users
│
├── frontend/                         # Next.js 14 SPA
│   ├── components/
│   │   ├── ChatBot.js                # Floating chatbot widget
│   │   ├── Navbar.js                 # Role-aware navigation
│   │   ├── ScanTable.js              # Reusable scan history table
│   │   ├── UploadScan.js             # Drag-and-drop uploader
│   │   ├── OTPVerification.js        # 6-digit OTP input + resend timer
│   │   ├── ProtectedRoute.js         # JWT route guard
│   │   └── ReportDownloadButton.js   # Secure PDF fetch
│   │
│   └── pages/
│       ├── index.js / login.js / register.js
│       ├── patient/index.js
│       ├── doctor/index.js
│       ├── pharmacist/index.js
│       └── admin/index.js
│
├── models/
│   ├── lung_model.h5                 # Trained MobileNetV2 (~14 MB)
│   ├── metadata/class_labels.json    # Class index mapping
│   ├── metrics/training_metrics.json # Per-class precision/recall/F1
│   └── plots/                        # Confusion matrix, training curves
│
├── img/                              # All diagrams & screenshots (34 images)
├── documents/                        # IEEE Papers + Project Report
│   ├── IEEE_Conference_CSSS.pdf
│   ├── IEEE_Journal_Paper_CSSS.pdf
│   ├── FINAL PPT.pdf
│   └── SUPPLEMENTARY_MATERIAL_CSSS.pdf
│
├── split_lung_dataset.py             # 70/15/15 dataset splitter
├── train_lung_model.py               # Full training pipeline
├── requirements.txt
└── README.md

✓ Python 3.10+          ✓ Node.js 18.17+ & npm 9+
✓ Git                   ✓ Gmail account with App Password
✓ 8GB+ RAM              ✓ NVIDIA GPU (optional, for training)

git clone https://github.com/Darkwebnew/Projectwork2.git
cd Projectwork2

pip install -r requirements.txt

cp .env.example .env          # Edit with your credentials

python backend/init_db.py     # Create database tables
python backend/seed_db.py     # Seed 4 demo accounts

uvicorn backend.main:app --reload --host 0.0.0.0 --port 8000

cd frontend
npm install
npm run dev

Frontend runs at http://localhost:3001 · API docs at http://localhost:8000/docs

# Database
DATABASE_URL=sqlite:///database/csss.db

# JWT Security
JWT_SECRET_KEY=your_secret_key_here
JWT_ALGORITHM=HS256
JWT_ACCESS_TOKEN_EXPIRE_MINUTES=60

# Gmail SMTP (use App Password, not account password)
SMTP_HOST=smtp.gmail.com
SMTP_PORT=587
SMTP_USER=[email protected]
SMTP_PASSWORD=your_16char_app_password
EMAIL_SENDER=[email protected]

# AI Model
AI_MODEL_PATH=models/lung_model.h5
CLASS_LABELS_PATH=models/metadata/class_labels.json
CONFIDENCE_THRESHOLD=0.75
IMG_SIZE=224

# Storage
UPLOAD_DIR=uploads/patient_scans
REPORTS_DIR=reports/temp

# Frontend
NEXT_PUBLIC_API_URL=http://localhost:8000
FRONTEND_URL=http://localhost:3001

Gmail SMTP: Enable 2FA → Google Account → Security → App Passwords → generate 16-character password.

python split_lung_dataset.py    # Creates 70/15/15 dataset split
python train_lung_model.py      # Trains MobileNetV2, saves to models/lung_model.h5

💡 The login page includes clickable test account buttons that auto-fill credentials.

STEP 1 ─ PATIENT
  → Login → Upload chest X-ray via drag-and-drop
  → Scan appears as "Pending AI" in history

STEP 2 ─ DOCTOR
  → Login → View pending queue
  → Click "Analyze" → MobileNetV2 runs inference
  → Prediction + confidence score displayed
  → Add clinical notes → "Verify" → status: DOCTOR_VERIFIED

STEP 3 ─ PHARMACIST
  → Login → See DOCTOR_VERIFIED scans
  → Review AI + doctor notes → add prescription
  → Click "Complete" → status: PHARMACIST_COMPLETED

STEP 4 ─ ADMIN (OTP 2FA required)
  → Login → Verify 6-digit email OTP
  → Review full scan → Click "Approve"
  → PDF auto-generated (WeasyPrint + Jinja2)
  → PDF emailed to patient automatically

STEP 5 ─ PATIENT
  → Receives encrypted PDF report in email
  → Can also download directly from dashboard
  → Report includes: scan image, AI prediction, confidence,
    risk level, doctor findings, prescription, 3 signatures

Interactive Swagger UI → http://localhost:8000/docs

📐 View All Architecture & Flow Diagrams

Workflow Diagram	Use Case Diagram

Sequence — Scan Upload	Sequence — Doctor Analysis

Sequence — Admin Approval	Activity Diagram

DFD Level 0	DFD Level 1

DFD Level 2 — Auth	JWT Auth Flow

OTP Auth Flow	RBAC Matrix

Class Diagram	Entity Relationship

State Diagram — Scan Lifecycle	MobileNetV2 Architecture

Classification Head	Grad-CAM Visualization

• PostgreSQL migration for production scale
• Docker Compose full-stack deployment
• DICOM file format support
• Grad-CAM heatmap overlays embedded in PDF reports
• WhatsApp report delivery via Twilio

• Mobile-responsive PWA frontend
• WebSocket real-time scan status push notifications
• Multi-language reports (English, Tamil, Hindi)
• EMR / EHR integration (HL7 FHIR)
• HIPAA audit logging
• Federated learning across hospital nodes

🏆 Core Development Team

Sriram V 🚀 Project Lead & AI/ML Engineer MobileNetV2 · FastAPI · DevOps · Full-Stack

Surothaaman R ⚙️ Backend Developer & Security JWT Auth · Role Guard · OTP · Database

Tamizharasi S 🎨 Frontend Developer Next.js 14 · React · Dashboard UI · Chatbot

Praveen CK 🗄️ Database & Testing SQLAlchemy · ORM · Integration Tests · QA

🎓 Academic Advisors

Role	Name	Institution
Medical Domain Supervisor	Ms. V. Swedha	Saveetha Engineering College, Chennai
AI/ML Advisor	Dr. Selvakumar R	Saveetha Engineering College, Chennai

⚠️ Important: This project is under a restrictive proprietary license. Contributions are welcome strictly for educational improvement purposes only. By submitting a pull request, you agree that your contribution becomes part of this project under the same license terms. No contributor may independently use, redistribute, or commercialize any part of this code.

1. Open an Issue first — discuss your idea before writing any code
2. Fork the repository
3. Create a branch — git checkout -b feature/YourFeature
4. Write tests for all changes
5. Run checks — pytest && black backend/ && flake8 backend/
6. Commit — git commit -m 'feat: Add YourFeature'
7. Push & open a Pull Request with a detailed description

Please read CODE_OF_CONDUCT.md before contributing.

If CSSS helped your research, institution, or medical project — consider supporting continued development!

Your support helps maintain this project, publish more IEEE research, and build better AI healthcare tools for the community.

╔══════════════════════════════════════════════════════════════════╗
║              PROPRIETARY SOFTWARE LICENSE                        ║
║     Copyright (c) 2024–2025  Sriram V & CSSS Development Team   ║
║                   All Rights Reserved                            ║
╚══════════════════════════════════════════════════════════════════╝

This software and all associated source code, documentation, trained ML models, research papers, datasets, configurations, UI designs, screenshots, and assets are the exclusive intellectual property of the authors and are fully protected under applicable copyright law and the Indian Copyright Act, 1957.

• Copy, reproduce, or redistribute this code in whole or in part
• Use this project or any portion of it in commercial medical products or services
• Modify, adapt, translate, or create derivative works based on this project
• Sublicense, sell, rent, lease, or transfer rights to any third party
• Use this project's name, branding, ML models, or research in your own publications without explicit written permission
• Deploy this system in any clinical, production, or commercial environment without written authorization from the authors
• Reverse engineer any trained model weights, binaries, or compiled components
• Present this work as your own in academic, medical, or professional contexts

• View and study the source code for personal educational purposes only
• Fork the repository on GitHub solely to submit pull requests
• Reference this project in academic citations with proper attribution
• Use general concepts and ideas (not code or models) as inspiration for entirely original work

Any unauthorized use, reproduction, distribution, or clinical deployment of this software — in whole or in part — is strictly prohibited and may result in civil and criminal penalties under applicable intellectual property and medical device regulation law. The authors reserve all rights and will pursue all available legal remedies for any violations.

For licensing inquiries, clinical deployment requests, or research collaboration: 📧 Contact: [email protected] · @Darkwebnew via GitHub Issues

See the full LICENSE.txt for complete terms.

⭐ Star this repository if CSSS helped your project!

Made with ❤️ for better healthcare diagnostics · Saveetha Engineering College · Tamil Nadu, India 🇮🇳

🐛 Report Bug · 💡 Request Feature · 📖 API Docs