🩺 Chest Cancer Detection using Deep Learning

An AI-powered web application for automated lung cancer detection from chest X-ray images

🚀 Live Demo | 📊 Report Bug | ✨ Request Feature

---

📋 Table of Contents

• Overview
• Features
• Technology Stack
• Model Architecture
• How It Works
• Installation
• Usage
• Project Structure
• Model Performance
• Screenshots
• Future Enhancements
• Contributing
• License
• Contact

---

🎯 Overview

This project implements a deep learning-based medical imaging system designed to assist radiologists and healthcare professionals in the early detection of lung cancer from chest X-ray images. Using a fine-tuned InceptionV3 convolutional neural network, the application can classify chest X-rays into four distinct categories:

1. Adenocarcinoma - Lung Cancer
2. Large Cell Carcinoma - Lung Cancer
3. Squamous Cell Carcinoma - Lung Cancer
4. Normal - No Cancer Detected

The model is deployed as an interactive Streamlit web application, making it accessible to medical professionals without requiring deep technical expertise.

🎯 Problem Statement

Lung cancer is one of the leading causes of cancer-related deaths worldwide. Early detection significantly improves survival rates, but manual analysis of X-rays is time-consuming and requires expert radiologists. This project aims to provide an AI-assisted diagnostic tool to:

• ✅ Speed up the diagnostic process
• ✅ Reduce human error and improve accuracy
• ✅ Make preliminary screening more accessible
• ✅ Assist in areas with limited access to expert radiologists

---

✨ Features

Core Features

• 🔍 Real-Time Prediction - Upload chest X-rays and receive instant AI-powered predictions
• 🎯 Multi-Class Classification - Detects 3 types of lung cancer + normal cases
• 📊 Confidence Scores - Displays prediction confidence percentage for transparency
• 🖼️ Image Preprocessing - Automatic image resizing and normalization
• 🌐 Web-Based Interface - No installation required for end-users (via live deployment)
• ⚡ Fast Inference - Optimized model for quick predictions

Technical Features

• 🧠 Pre-trained InceptionV3 architecture with transfer learning
• 📁 Support for multiple image formats (PNG, JPG, JPEG)
• 🔄 Automatic image conversion to RGB
• 📈 Confidence-based result visualization
• 🎨 User-friendly Streamlit interface

---

🛠️ Technology Stack

Machine Learning & Deep Learning

• TensorFlow 2.x - Deep learning framework
• Keras - High-level neural networks API
• InceptionV3 - Pre-trained CNN architecture for image classification
• NumPy - Numerical computing

Web Framework

• Streamlit - Interactive web application framework

Image Processing

• Pillow (PIL) - Image manipulation and processing
• Matplotlib - Visualization (optional)

Deployment

• Streamlit Cloud - Production deployment
• Git LFS - Large file storage for model weights

---

🧠 Model Architecture

Transfer Learning Approach

This project uses InceptionV3, a state-of-the-art convolutional neural network architecture developed by Google, pre-trained on ImageNet. The model leverages:

• Inception Modules - Multi-scale feature extraction using parallel convolutions
• Factorized Convolutions - Reduced computational cost
• Auxiliary Classifiers - Better gradient flow during training

Model Specifications

Parameter	Value
Architecture	InceptionV3
Input Size	224 × 224 × 3 (RGB)
Output Classes	4 (3 Cancer Types + Normal)
Activation	Softmax
Loss Function	Categorical Crossentropy
Optimizer	Adam (presumed)

Classification Categories

classes = [
    "Adenocarcinoma Chest Lung Cancer",
    "Large cell carcinoma Lung Cancer",
    "No Lung Cancer / NORMAL",
    "Squamous cell carcinoma Lung Cancer"
]

---

🔄 How It Works

Workflow

┌─────────────────┐
│  User Uploads   │
│   Chest X-Ray   │
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│ Image Loading & │
│  Preprocessing  │  ← Convert to RGB, Resize to 224×224
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│  Normalization  │  ← Pixel values scaled to [0, 1]
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│ InceptionV3 CNN │  ← Feature extraction & classification
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│  Softmax Layer  │  ← Probability distribution
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│   Prediction    │  ← Highest probability class
│  + Confidence   │
└─────────────────┘

Prediction Pipeline

1. Input - User uploads a chest X-ray image (PNG/JPG/JPEG)
2. Preprocessing - Image is converted to RGB and resized to 224×224 pixels
3. Normalization - Pixel values are scaled from [0, 255] to [0, 1]
4. Inference - The preprocessed image is fed through the InceptionV3 model
5. Prediction - Softmax layer outputs probability distribution across 4 classes
6. Result - The class with the highest probability is displayed along with confidence score

---

🚀 Installation

Prerequisites

Make sure you have the following installed:

• Python 3.7+ - Download Python
• Git - Download Git
• Git LFS - Install Git LFS (Required for model file)

Step-by-Step Setup

1. Clone the Repository

git clone https://github.com/amitkumar2308/Cancer-detection-streamlit.git
cd Cancer-detection-streamlit

2. Set Up Virtual Environment (Recommended)

On Windows:

python -m venv .venv
.venv\Scripts\activate

On macOS/Linux:

python3 -m venv .venv
source .venv/bin/activate

3. Install Dependencies

pip install -r requirements.txt

4. Download Model File (If Using Git LFS)

If the model file (inception_chest.h5) wasn't downloaded automatically:

git lfs install
git lfs pull

5. Run the Application

streamlit run model.py

The application will open in your default browser at http://localhost:8501

---

💻 Usage

Running Locally

1. Start the application:

   streamlit run model.py

2. Upload an X-ray image:

   • Click on "Choose an image..." button
   • Select a chest X-ray image (PNG, JPG, or JPEG format)

3. Get Prediction:

   • Click the "Predict" button
   • View the predicted cancer type and confidence score

Using the Live Demo

Visit the deployed application: 🚀 Live Demo

---

📁 Project Structure

Cancer-detection-streamlit/
│
├── assests/
│   └── project.png              # Screenshot of the application
│
├── inception_chest.h5           # Pre-trained InceptionV3 model weights
├── model.py                     # Main Streamlit application
├── requirements.txt             # Python dependencies
│
├── .gitattributes              # Git LFS configuration
├── README.md                   # Project documentation
└── LICENSE                     # MIT License

Key Files

• model.py - Main application file containing:

  • Model loading function
  • Image preprocessing pipeline
  • Prediction logic
  • Streamlit UI components

• inception_chest.h5 - Fine-tuned InceptionV3 model (stored via Git LFS)

• requirements.txt - Dependencies:

  tensorflow
  numpy
  pillow
  streamlit
  matplotlib
  

---

📊 Model Performance

Dataset Information

The model was trained on a chest X-ray dataset containing labeled images of:

• Adenocarcinoma cases
• Large cell carcinoma cases
• Squamous cell carcinoma cases
• Normal/healthy chest X-rays

Key Metrics

Note: Add your actual metrics here if available

• Accuracy: ~XX%
• Precision: ~XX%
• Recall: ~XX%
• F1-Score: ~XX%

---

📸 Screenshots

Screenshot showing the Streamlit interface with prediction results

---

🚀 Future Enhancements

Planned Features

• Grad-CAM Visualization - Highlight regions of the X-ray that influenced the prediction
• Batch Processing - Upload and analyze multiple X-rays simultaneously
• Patient History Tracking - Store and compare previous scans
• PDF Report Generation - Export diagnostic reports
• Model Ensemble - Combine multiple models for improved accuracy
• API Endpoint - RESTful API for integration with hospital systems
• Mobile App - Native mobile application for iOS/Android
• Multi-language Support - Internationalization for global accessibility

Technical Improvements

• Model quantization for faster inference
• A/B testing with other architectures (ResNet, EfficientNet, Vision Transformer)
• Active learning pipeline for continuous model improvement
• DICOM format support for medical imaging standards
• Integration with PACS (Picture Archiving and Communication System)

---

🤝 Contributing

Contributions are what make the open-source community such an amazing place to learn, inspire, and create! Any contributions you make are greatly appreciated.

How to Contribute

1. Fork the Project

   git clone https://github.com/your-username/Cancer-detection-streamlit.git

2. Create a Feature Branch

   git checkout -b feature/AmazingFeature

3. Commit Your Changes

   git commit -m 'Add some AmazingFeature'

4. Push to the Branch

   git push origin feature/AmazingFeature

5. Open a Pull Request

Development Guidelines

• Follow PEP 8 style guide for Python code
• Use black for code formatting
• Write meaningful commit messages
• Add comments for complex logic
• Test thoroughly before submitting PRs

---

📜 License

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

---

📞 Contact

Amit Kumar

• 💼 GitHub: @amitkumar2308
• 📧 Email: amitkumar@example.com
• 🌐 Live Application: Cancer Detection App

---

🙏 Acknowledgments

• TensorFlow & Keras - For providing robust deep learning frameworks
• Streamlit - For the amazing web app framework
• Google Research - For the InceptionV3 architecture
• Chest X-ray Dataset Contributors - For making medical imaging data available for research

---

⚠️ Disclaimer

This application is for educational and research purposes only. It should not be used as a substitute for professional medical advice, diagnosis, or treatment. Always seek the advice of qualified health providers with any questions regarding a medical condition.

---

⭐ If you found this project helpful, please consider giving it a star!

Made with ❤️ by Amit Kumar