🧬 Clinical Scan Support System

AI-Powered Medical Image Diagnosis & Automated Clinical Reporting Platform

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

🚀 Quick Start • 🏗️ Architecture • 🧠 AI Pipeline • 🔌 API Docs • 🧪 Workflow Demo • 🎓 Model Training • 👥 Team

---

⚕️ 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.

---

🎯 Overview

Clinical Scan Support System (CSSS) is a production-grade, full-stack AI medical imaging platform that 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.

Built as a final-year engineering mini-project, CSSS tackles a real clinical bottleneck: traditional diagnostic workflows are manual, siloed, and slow. CSSS replaces this with a unified platform where every stakeholder — patient, doctor, pharmacist, and admin — has a purpose-built dashboard, enforced access control, and a clearly defined role in the pipeline.

✨ Key Features

Feature	Detail
🔍 Multi-Modal Scan Support	Chest X-ray (NIH), COVID-19 radiography, cardiac MRI
🧠 Deep Learning Inference	MobileNetV2 — 89.51% test accuracy, sub-second prediction
🔄 4-Role Clinical Workflow	Patient → Doctor → Pharmacist → Admin approval pipeline
📄 Auto PDF Report Generation	WeasyPrint + Jinja2 HTML → professional diagnostic report
📧 SMTP Email Delivery	Encrypted PDF automatically emailed to patient on approval
🔒 JWT + OTP 2FA Security	Role-based access control, two-factor login for admin
💬 Medical AI Chatbot	Context-aware assistant for patient FAQ and workflow help
🎛️ 4 Role Dashboards	Tailored UIs — Admin, Doctor, Pharmacist, Patient
⚡ Live Status Tracking	Real-time scan status badges across all role dashboards
📊 Confidence Scoring	Per-class probability output with automatic uncertainty flagging
🗄️ SQLite / PostgreSQL	Lightweight local dev DB with production migration path
🖥️ Dark Theme SPA	Responsive Next.js 14 frontend with cyan-accented dark UI

🎯 Use Cases

• Hospital Radiology Departments — Streamline chest X-ray and CT report generation with AI pre-screening
• COVID-19 Screening Clinics — Real-time detection of COVID, Viral Pneumonia, and Lung Opacity patterns
• Cardiac Screening Centers — Automated normal/abnormal cardiac MRI triage
• Telemedicine Platforms — Remote scan submission with automated encrypted report delivery
• Medical Education — Demonstrate AI-assisted diagnostics with an end-to-end working system

---

🏗️ System Architecture

🧩 Component Summary

Component	Technology	Responsibility
Frontend SPA	Next.js 14 + React 18	Role-based dashboards, scan upload, report download
REST API	FastAPI (Python)	Endpoints, business logic, auth enforcement
AI Inference	TensorFlow 2.x / Keras	MobileNetV2 prediction, confidence scoring
PDF Engine	WeasyPrint + Jinja2	Clinical diagnostic report rendering
Auth Stack	JWT + bcrypt + OTP	Stateless auth, password hashing, admin 2FA
Database	SQLAlchemy + SQLite	ORM, scan lifecycle, user accounts, OTP records
Email Service	smtplib + Gmail SMTP	OTP codes + PDF report delivery
Chatbot	Rule-based NLP	Keyword-matched medical FAQ assistant

---

📂 Project Structure

📁 Click to expand full project tree

CLINICAL-SCAN-SUPPORT-SYSTEM/
│
├── .env                              # Environment configuration
├── requirements.txt                  # Python dependencies
├── README.md                         # This file
├── split_lung_dataset.py             # Dataset train/val/test splitter (70/15/15)
├── train_lung_model.py               # MobileNetV2 training with Grad-CAM
│
├── backend/                          # FastAPI REST API
│   ├── models/
│   │   ├── user.py                   # User (id, name, email, role, password)
│   │   ├── schema.py                 # Scan (id, patient_id, file_path, 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}, /verify/{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, get_current_user
│   │   ├── password.py               # bcrypt hash / verify
│   │   └── role_guard.py             # require_role() dependency factory
│   │
│   ├── services/
│   │   ├── ai_service.py             # TensorFlow MobileNetV2 inference (predict_scan)
│   │   └── email_service.py          # SMTP: OTP email + PDF attachment delivery
│   │
│   ├── templates/
│   │   └── report_template.html      # Jinja2 HTML → WeasyPrint PDF
│   │
│   ├── config.py                     # BASE_DIR + UPLOAD_DIR path config
│   ├── database.py                   # SQLAlchemy engine + session + Base
│   ├── init_db.py                    # Create all DB tables
│   ├── main.py                       # FastAPI app + CORS + router registration
│   └── seed_db.py                    # Seed 4 demo users
│
├── frontend/                         # Next.js 14 SPA
│   ├── components/
│   │   ├── ChatBot.js                # Floating chatbot with category badges
│   │   ├── Navbar.js                 # Role-aware navigation bar
│   │   ├── ScanTable.js              # Reusable scan history table component
│   │   ├── UploadScan.js             # Drag-and-drop file uploader
│   │   ├── OTPVerification.js        # 6-digit OTP input with resend timer
│   │   ├── ProtectedRoute.js         # JWT-validated route guard
│   │   └── ReportDownloadButton.js   # Secure PDF fetch and open
│   │
│   ├── pages/
│   │   ├── index.js                  # Landing page
│   │   ├── login.js                  # Login + OTP flow
│   │   ├── register.js               # Registration page
│   │   ├── patient/index.js          # Patient dashboard
│   │   ├── doctor/index.js           # Doctor dashboard
│   │   ├── pharmacist/index.js       # Pharmacist dashboard
│   │   └── admin/index.js            # Admin dashboard
│   │
│   ├── services/api.js               # Axios client + auth interceptors
│   ├── styles/globals.css            # Dark theme global CSS
│   └── package.json
│
├── dataset/                          # Split dataset (created by split_lung_dataset.py)
│   └── lung/
│       ├── train/                    # 70% — 6 class folders
│       ├── val/                      # 15% — 6 class folders
│       └── test/                     # 15% — 6 class folders
│
├── Dataset/                          # Raw Kaggle dataset (input for splitter)
│   ├── COVID/
│   ├── Lung_Opacity/
│   ├── NIH_MERGED/
│   ├── Normal/
│   ├── Sick/
│   └── Viral_Pneumonia/
│
├── models/
│   ├── lung_model.h5                 # Trained MobileNetV2 weights (~14 MB)
│   ├── metadata/class_labels.json    # ["COVID","Lung_Opacity","NIH_MERGED","Normal","Sick","Viral_Pneumonia"]
│   ├── metrics/training_metrics.json # Train / val / test accuracy
│   └── plots/                        # Confusion matrix, training curves, Grad-CAM misclassifications
│
└── uploads/
    └── patient_scans/                # Uploaded scans stored as {UUID}.{ext}

---

🧠 AI Model & Pipeline

📊 Dataset

Source	Classes	Images
NIH Chest X-ray Dataset	NIH_MERGED (14 pathologies combined)	112,120
COVID-19 Radiography Database	COVID · Normal · Lung_Opacity · Viral_Pneumonia	42,330
CAD Cardiac MRI Dataset	Normal · Sick	63,425
Total	6 disease classes	217,875

🏛️ Model Architecture

Input Image (224 × 224 × 3)
         ↓
MobileNetV2 (ImageNet pretrained — feature extractor, frozen)
         ↓
GlobalAveragePooling2D
         ↓
Dense(256, activation='relu')
         ↓
Dropout(0.4)
         ↓
Dense(6, activation='softmax')   ← 6 disease classes
         ↓
Output: predicted_class + confidence_score + all_class_probabilities

⚙️ Training Configuration

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)
]

📈 Performance Metrics

  

---

Metric	Value
Test Accuracy	89.51%
Validation Accuracy	89.31%
Training Accuracy	92.97%
Total Training Images	217,875
Inference Speed	< 1 second
Model Size	~14 MB (.h5)
Confidence Threshold	75% (below → "Uncertain")

Predicted Disease Classes: COVID · Lung_Opacity · NIH_MERGED · Normal · Sick · Viral_Pneumonia

🔬 Inference Code (backend/services/ai_service.py)

View inference pipeline code

def predict_scan(image_path: str) -> dict:
    img = cv2.imread(image_path, cv2.IMREAD_COLOR)
    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
    img = cv2.resize(img, (IMG_WIDTH, IMG_HEIGHT))   # auto-detected from model
    img = img.astype("float32") / 255.0
    img = np.expand_dims(img, axis=0)

    predictions   = model.predict(img, verbose=0)
    confidence    = float(np.max(predictions))
    class_index   = int(np.argmax(predictions))
    label         = CLASS_NAMES[class_index]

    if confidence < CONFIDENCE_THRESHOLD:   # default 0.75
        label = "Uncertain"

    return {
        "label":           label,
        "confidence":      confidence,
        "class_index":     class_index,
        "all_predictions": {CLASS_NAMES[i]: float(predictions[0][i])
                            for i in range(len(CLASS_NAMES))},
        "threshold_used":  CONFIDENCE_THRESHOLD,
    }

---

🎓 Model Training & Dataset Preparation

Dataset Splitting (split_lung_dataset.py)

Prepares the raw Kaggle dataset for training by creating train/validation/test splits:

# 70/15/15 split with sklearn train_test_split
SPLIT_RATIO = (0.7, 0.15, 0.15)   # train / val / test

# Preserves class folder structure for ImageDataGenerator
# Uses fixed random_state=42 for reproducibility
# Creates dataset/lung/train, dataset/lung/val, dataset/lung/test

Usage:

python split_lung_dataset.py

Output Structure:

dataset/lung/
├── train/
│   ├── COVID/           (70% of COVID images)
│   ├── Lung_Opacity/
│   ├── NIH_MERGED/
│   ├── Normal/
│   ├── Sick/
│   └── Viral_Pneumonia/
├── val/                 (15% of each class)
└── test/                (15% of each class)

Model Training (train_lung_model.py)

Trains MobileNetV2 on the split dataset with transfer learning and Grad-CAM visualization:

View training pipeline features

Key Features:

• Transfer Learning: MobileNetV2 base (ImageNet weights, frozen)
• Data Augmentation: Rotation (10°), zoom (0.1), horizontal flip
• Callbacks: EarlyStopping (patience=3), ModelCheckpoint (saves best model)
• Metrics Export: Training curves, confusion matrix, classification report
• Grad-CAM Visualization: Heatmap overlay on top 9 misclassified images

Training Code Snippet:

base_model = MobileNetV2(weights="imagenet", include_top=False, input_shape=(224,224,3))
base_model.trainable = False

x = base_model.output
x = GlobalAveragePooling2D()(x)
x = Dense(128, activation="relu")(x)
outputs = Dense(len(CLASS_NAMES), activation="softmax")(x)

model = Model(inputs=base_model.input, outputs=outputs)
model.compile(optimizer="adam", loss="categorical_crossentropy", metrics=["accuracy"])

callbacks = [
    EarlyStopping(patience=3, restore_best_weights=True),
    ModelCheckpoint("models/lung_model.h5", save_best_only=True)
]

history = model.fit(
    train_data, validation_data=val_data, epochs=15, callbacks=callbacks
)

Generated Artifacts:

• models/lung_model.h5 — Trained model weights
• models/metadata/class_labels.json — Class index mapping
• models/metrics/training_metrics.json — Precision, recall, F1 per class
• models/plots/training_graph.png — Accuracy/loss curves
• models/plots/confusion_matrix.png — Per-class performance heatmap
• models/plots/top_misclassified_gradcam.png — Grad-CAM visualizations

Usage:

# After splitting dataset
python train_lung_model.py

Training Environment:

• GPU: NVIDIA GPU (CUDA-enabled) recommended for faster training
• RAM: 16GB+ recommended
• Training Time: ~15 minutes on RTX 3060 (depends on GPU)

---

🔄 Clinical Workflow

The scan lifecycle moves through 5 statuses, each requiring a different authenticated role:

┌──────────┐     ┌────────────────┐     ┌─────────────────┐     ┌───────────────────────┐     ┌──────────────┐
│ PATIENT  │     │    DOCTOR      │     │     DOCTOR      │     │      PHARMACIST       │     │    ADMIN     │
│ uploads  │────▶│  runs AI scan  │────▶│ adds clinical   │────▶│  adds prescription    │────▶│  approves +  │
│  image   │     │                │     │     notes       │     │        notes          │     │  emails PDF  │
└──────────┘     └────────────────┘     └─────────────────┘     └───────────────────────┘     └──────────────┘
PENDING_AI    →    AI_ANALYZED      →    DOCTOR_VERIFIED     →    PHARMACIST_COMPLETED     →    REPORT_READY

Status	Meaning
PENDING_AI	Scan uploaded, awaiting AI analysis
AI_ANALYZED	MobileNetV2 prediction complete with confidence score
DOCTOR_VERIFIED	Doctor reviewed AI results and added clinical notes
PHARMACIST_COMPLETED	Pharmacist reviewed and added prescription notes
REPORT_READY	Admin approved — PDF auto-generated and emailed to patient

---

🎛️ User Interface Walkthrough

🏠 System Landing Page

• Modern dark-themed medical UI built with Next.js 14
• Clean navigation with role-based system access
• Secure login and registration entry points
• Fully responsive professional hospital interface
• Fast-loading optimized frontend architecture

📝 Registration Page

• Secure registration with role selection (Patient, Doctor, Pharmacist, Admin)
• Password hashing using bcrypt encryption
• Email validation and duplicate prevention
• Role-based account creation
• Clean medical-grade UI form design

🔑 Login Page

• Secure JWT authentication system
• Role-based automatic dashboard redirect
• Token-based session management
• Protected route authorization
• Fast and secure login workflow

📧 OTP Email Verification (Admin Only)

• Secure 6-digit OTP verification system
• Email-based admin two-factor authentication
• 10-minute OTP expiration protection
• One-time-use OTP validation
• Prevents unauthorized admin access

👤 Patient Dashboard

• Upload scan images via drag-and-drop interface
• Real-time upload progress tracking
• View scan history with status badges
• Display AI prediction and confidence score
• Access doctor and pharmacist notes
• Download approved diagnostic PDF reports

📤 Patient Scan Upload Interface

• Drag-and-drop scan upload system
• Supports JPG and PNG file formats
• Secure UUID-based file storage
• Automatic scan validation
• Instant status tracking after upload

🩺 Doctor Dashboard

• View all patient scans in doctor work queue
• Trigger AI analysis using MobileNetV2 model
• View prediction results and confidence score
• Add mandatory clinical verification notes
• Advance scan to DOCTOR_VERIFIED status

💊 Pharmacist Dashboard

• View DOCTOR_VERIFIED scans
• Review AI prediction and doctor notes
• Add medication and prescription details
• Use quick-fill prescription templates
• Mark scan PHARMACIST_COMPLETED

🛡️ Admin Dashboard

• Secure OTP-protected admin login
• View PHARMACIST_COMPLETED scans
• Approve diagnostic reports
• Automatic PDF report generation
• Secure report delivery to patient email
• Full diagnostic workflow control

💬 CSSS Medical Chatbot

• Floating chatbot widget across dashboards
• Provides instant system and medical assistance
• Supports keyword-based intelligent responses
• Chat history persistence
• Typing animation and clean conversational UI

---

🔌 API Endpoints

Full interactive docs at http://localhost:8000/docs (Swagger UI) after starting the server.

Authentication /auth

Method	Endpoint	Description
POST	/auth/register	Register new user (name, email, password, role)
POST	/auth/login	Login → JWT token (admin requires OTP step)

OTP /otp

Method	Endpoint	Description
POST	/otp/send	Send 6-digit OTP to admin email (10-min expiry)
POST	/otp/verify	Verify OTP code → returns full access JWT

Patient /patient

Method	Endpoint	Description
POST	/patient/upload	Upload scan image (JPG/PNG, multipart, max 10MB)
GET	/patient/status/{patient_id}	Get all scans for this patient

Doctor /doctor

Method	Endpoint	Description
GET	/doctor/pending	All scans in work queue
POST	/doctor/analyze/{scan_id}	Run MobileNetV2 inference on scan
POST	/doctor/verify/{scan_id}	Add clinical notes → status: DOCTOR_VERIFIED

Pharmacist /pharmacist

Method	Endpoint	Description
GET	/pharmacist/queue	Scans in DOCTOR_VERIFIED status
POST	/pharmacist/complete/{scan_id}	Add prescription → status: PHARMACIST_COMPLETED

Admin /admin

Method	Endpoint	Description
GET	/admin/pending	Scans awaiting final approval
POST	/admin/approve/{scan_id}	Generate PDF + email patient + REPORT_READY

Reports /reports

Method	Endpoint	Description
GET	/reports/pdf/{scan_id}	Download diagnostic PDF (JWT protected, role scoped)

Chatbot /chatbot

Method	Endpoint	Description
POST	/chatbot/	Send message → keyword-matched medical response + category

---

🔒 Security Architecture

🔐 Role Permission Matrix

Permission	Admin	Doctor	Pharmacist	Patient
Upload scan	❌	❌	❌	✅
Run AI analysis	❌	✅	❌	❌
Add clinical notes	❌	✅	❌	❌
Add prescription notes	❌	❌	✅	❌
Approve & email report	✅	❌	❌	❌
Download own PDF report	❌	❌	❌	✅
Download any PDF report	✅	✅	✅	❌
Use chatbot	✅	✅	✅	✅
OTP 2FA login	✅	❌	❌	❌

---

🛠️ Installation & Setup

Prerequisites

• Python 3.10+
• Node.js 18.17+ and npm 9+
• Git
• (Optional) NVIDIA GPU with CUDA for model training

1. Clone the Repository

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

2. Backend Setup (FastAPI)

# Install Python dependencies
pip install -r requirements.txt

# Configure environment variables
cp .env.example .env
# Edit .env with your SMTP credentials and JWT secret

# Initialize the database
python backend/init_db.py

# Seed 4 demo users
python backend/seed_db.py

# Start the FastAPI server
uvicorn backend.main:app --reload --host 0.0.0.0 --port 8000

API: http://localhost:8000 · Swagger UI: http://localhost:8000/docs

3. Frontend Setup (Next.js)

cd frontend
npm install
npm run dev

Frontend: http://localhost:3001

4. Environment Variables

# ── 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

# ── SMTP (Gmail App Password) ─────────────────────
SMTP_HOST=smtp.gmail.com
SMTP_PORT=587
SMTP_USER=your_email@gmail.com
SMTP_PASSWORD=your_16char_app_password
EMAIL_SENDER=your_email@gmail.com

# ── 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 Note: Enable 2FA on your Google account → Google Account → Security → App Passwords → generate 16-character password. Use that as SMTP_PASSWORD.

5. (Optional) Retrain the Model

If you want to retrain the model with your own dataset:

# Step 1: Split dataset
python split_lung_dataset.py

# Step 2: Train model (requires GPU for faster training)
python train_lung_model.py

# Trained model will be saved to models/lung_model.h5

---

🚀 Quick Start

Demo Accounts (after seed_db.py)

Role	Email	Password
Admin	admin@csss.com	Admin123
Doctor	doctor@csss.com	Doctor123
Pharmacist	pharma@csss.com	Pharma123
Patient	patient@csss.com	Patient123

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

---

🧪 End-to-End Workflow Demo

Step 1 ── PATIENT logs in
          → Clicks "Upload Scan"
          → Drags a chest X-ray JPG into the upload zone
          → Real-time progress bar shows upload completion
          → Scan appears in history as "Pending AI"

Step 2 ── DOCTOR logs in
          → Sees the scan in the work queue
          → Clicks "Analyze" → MobileNetV2 runs inference
          → Prediction (e.g. "COVID") + confidence score displayed
          → Doctor adds clinical notes → clicks "Verify"
          → Status updates to "Doctor Verified"

Step 3 ── PHARMACIST logs in
          → Sees scan in prescription queue
          → Reviews AI prediction, confidence, and doctor notes
          → Selects quick template or writes custom prescription
          → Clicks "Complete" → status: "Pharmacist Completed"

Step 4 ── ADMIN logs in (OTP email verification required)
          → Sees scan in pending approvals list
          → Reviews full scan details
          → Clicks "Approve"
          → PDF auto-generated (WeasyPrint + Jinja2 template)
          → PDF emailed to patient automatically in background

Step 5 ── PATIENT
          → Receives email with professional PDF report attached
          → Can also download directly from Patient Dashboard
          → Report contains: scan image, AI prediction, confidence score,
            risk level, doctor findings, prescription, and 3-party signatures

---

📄 Generated PDF Report

The auto-generated diagnostic report includes:

• 🏥 Hospital header — clinic name, report ID, generation timestamp
• 👤 Patient information — name, ID, age, gender, contact details
• 🩻 Scan information — scan type, date, machine, contrast, quality
• 🤖 AI Analysis Results — model, version, prediction badge, confidence %, risk level
• 📝 Clinical Findings — doctor verification notes
• 💊 Prescription — pharmacist medication notes
• 🖼️ Scan Image — original scan embedded as base64
• ✍️ Signatures — Radiologist, Pharmacist, and AI System signature blocks
• ⚠️ AI Notice — professional disclaimer watermark

---

💬 CSSS Medical Chatbot

The built-in CSSS Medical Assistant is a keyword-matching chatbot (not an LLM). It handles common questions across 10 categories:

Category	Topics Covered
🩺 Medical	CT, MRI, X-ray scan type information
🔄 Workflow	Patient → Doctor → Pharmacist → Admin pipeline
🤖 AI	What predictions mean, confidence scores, uncertainty
🔓 Auth	Login, OTP, registration, password requirements
🖼️ Scan	Upload formats, file size limits, accepted types
📊 Result	How to read and understand your diagnostic report
📞 Support	Contact, feedback, troubleshooting
🔑 Admin	Admin OTP process explanation
👋 Greeting	Welcome and general help
❓ Fallback	Topic list when query is not recognized

---

🗺️ Roadmap

v2.0 — Q3 2026

• PostgreSQL migration for production scale
• Docker Compose full-stack deployment
• WhatsApp report delivery via Twilio (already configured in .env)
• DICOM file format support
• Grad-CAM heatmap overlays in PDF reports for explainable AI

v2.5 — Q4 2026

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

---

👥 Team

	Name	Role	GitHub
	Sriram V	Lead Developer & AI/ML Engineer	@Darkwebnew
	Surothaaman R	Backend Development & Security	@surothaaman
	Andrew Varghese V S	Frontend Development	@Andrewvarghese653
	Praveen CK	Database & Testing	@praveenck23009864
	Dr. Swedha V	Medical Domain Expertise	@swedha333
	Dr. Selvakumar R	AI/ML Advisor	@selvasachein

---

🙏 Acknowledgements

Datasets (Kaggle)

• NIH Clinical Center — Chest X-Ray8 Dataset (112,120 images)
• Tawsifur Rahman et al. — COVID-19 Radiography Database (42,330 images)
• CAD Cardiac MRI Dataset (63,425 images)

Core Libraries

• FastAPI — High-performance Python REST API
• Next.js — React framework with file-based routing
• TensorFlow / Keras — Deep learning inference
• WeasyPrint — HTML → PDF report generation
• SQLAlchemy — Python ORM
• python-jose — JWT implementation
• Axios — HTTP client for frontend API calls

Academic References

• Howard et al., "MobileNets: Efficient CNNs for Mobile Vision Applications" (2017)
• Sandler et al., "MobileNetV2: Inverted Residuals and Linear Bottlenecks" (2018)
• Selvaraju et al., "Grad-CAM: Visual Explanations from Deep Networks" (2017)
• Wang et al., "ChestX-ray8: Hospital-scale Chest X-ray Database and Benchmarks" (2017)
• Chowdhury et al., "Can AI Help in Screening Viral and COVID-19 Pneumonia?" IEEE Access (2020)

---

📄 License

This project is proprietary software. All rights reserved © 2026.

⚠️ Unauthorized copying, distribution, or use of this software is strictly prohibited under the Indian Copyright Act, 1957. See LICENSE.txt for full terms.

---

📞 Contact & Support

Channel	Link
📧 Email	sriramnvks@gmail.com
🐛 Bug Reports	GitHub Issues
💬 Discussions	GitHub Discussions
📖 API Docs	http://localhost:8000/docs (Swagger UI)

---

Built with ❤️ for better healthcare diagnostics — Tamil Nadu, India 🇮🇳

⭐ Star this repository if CSSS helped your project!