README.md

metadata

title: PulmoScanAI
emoji: 🫁
colorFrom: blue
colorTo: green
sdk: docker
app_file: app.py
pinned: false

PulmoScanAI - AI Lung Cancer Detection System

An advanced web-based application for detecting lung cancer from histopathology images using a deep learning CNN model with feature-based analysis.

Features

• Real-time AI Analysis: Uses TensorFlow/Keras deep learning model
• Feature-based Detection: Analyzes darkness, purple staining, and edge density
• Beautiful UI: Modern, responsive design with animated backgrounds
• Drag & Drop Upload: Easy image upload with preview
• Confidence Score: Displays detection confidence percentage
• CORS Enabled: Seamless frontend-backend communication

How It Works

1. Upload Image: Drag & drop a histopathology image
2. CNN Processing: Model analyzes tissue patterns
3. Feature Analysis: Evaluates darkness, staining, and texture
4. Result: Shows diagnosis with confidence score
   • Green: Normal tissue detected
   • Red: Cancer detected

API Endpoints

Health Check

GET /api/health

Prediction

POST /api/predict
Content-Type: multipart/form-data

Request: Image file in multipart form data Response:

{
  "is_cancer": false,
  "confidence": 0.92,
  "diagnosis": "No Cancer Found",
  "confidence_percentage": 92.0
}

Model Information

• Architecture: Convolutional Neural Network (CNN)
• Input: 150×150 RGB images
• Output: 3-class classification (Adenocarcinoma, Normal, Squamous Cell Carcinoma)
• Framework: TensorFlow 2.13.0 / Keras

Technical Stack

• Frontend: HTML5, CSS3, JavaScript (Vanilla)
• Backend: Python Flask with Flask-CORS
• ML Framework: TensorFlow 2.x / Keras
• Image Processing: OpenCV, Pillow, NumPy

Project Structure

├── app.py                    # Flask backend server
├── best_lung_model.h5       # Trained CNN model
├── PulmoScanAI.html         # Web frontend
├── requirements.txt         # Python dependencies
├── Dockerfile               # Container configuration
└── README.md               # This file

License

© 2025 PulmoScanAI • Next-Gen AI Pathology Platform