harshtd/lung-nodule-ct-detection

🫁 Lung Nodule CT Detection

A comprehensive medical imaging application for detecting lung nodules in CT scans using the MONAI framework and deep learning.

📋 Overview

This Hugging Face Space provides an easy-to-use interface for lung nodule detection in CT scans. The application leverages the power of MONAI (Medical Open Network for AI), a PyTorch-based framework designed specifically for healthcare imaging applications.

🚀 Features

• —Medical Image Support: Accepts multiple medical imaging formats including NIfTI (.nii, .nii.gz) and DICOM (.dcm)
• —Deep Learning Detection: Uses the MONAI/lungnodulect_detection model for accurate nodule identification
• —Visual Results: Provides both textual summaries and visual overlays of detected nodules
• —Confidence Scoring: Each detection includes confidence scores to help assess reliability
• —User-Friendly Interface: Clean, intuitive Gradio interface designed for both medical professionals and researchers

🛠 How to Use

1. 1.Upload CT Scan: Click on the file upload area and select your CT scan file
2. 2.Supported formats: .nii, .nii.gz, .dcm, .png, .jpg, .jpeg, .tiff
3. 3.Recommended: NIfTI or DICOM formats for best results

1. 1.Run Analysis: Click the "🔍 Analyze CT Scan" button to start the detection process

1. 1.Review Results:
2. 2.Detection Summary: Text-based results showing number of nodules found, confidence scores, and bounding box coordinates
3. 3.Visualization: Visual overlay showing detected nodules highlighted with red bounding boxes

📊 Model Information

• —Base Model: MONAI/lungnodulect_detection
• —Framework: MONAI (Medical Open Network for AI)
• —Task: Object detection for lung nodules in CT scans
• —Input: 3D CT volumes
• —Output: Bounding boxes with confidence scores

🔧 Technical Details

Dependencies

• —MONAI: Medical imaging framework with comprehensive transforms and networks
• —PyTorch: Deep learning backend
• —Gradio: Web interface framework
• —nibabel: Medical imaging file I/O
• —matplotlib: Visualization
• —PIL: Image processing

Preprocessing Pipeline

1. 1.Loading: Medical images loaded using MONAI's LoadImage transform
2. 2.Normalization: Intensity scaling to standardize pixel values
3. 3.Resampling: Images resized to model input requirements (512x512x64)
4. 4.Channel Management: Ensures proper channel dimensions for model input

Model Architecture

The underlying model uses advanced computer vision techniques optimized for medical imaging:

• —Specialized for 3D volumetric data
• —Trained on lung CT datasets
• —Optimized for nodule detection tasks

⚠️ Medical Disclaimer

IMPORTANT: This application is intended for research and educational purposes only. It should NOT be used as a substitute for professional medical diagnosis or treatment. Key considerations:

• —Results should always be reviewed by qualified healthcare professionals
• —This tool is for demonstration and research purposes
• —Clinical decisions should never be based solely on automated analysis
• —Always consult with radiologists and physicians for medical diagnosis
• —The model may have limitations and false positives/negatives

🔍 Example Use Cases

• —Research: Academic studies on lung nodule detection algorithms
• —Education: Teaching medical imaging and AI applications
• —Prototyping: Developing medical imaging workflows
• —Screening: Preliminary analysis (with professional oversight)

📈 Performance Notes

• —Processing Time: Varies based on image size and hardware availability
• —Hardware: CPU processing supported; GPU acceleration available with upgraded hardware
• —Memory: Large CT volumes may require significant memory
• —Accuracy: Results depend on image quality and nodule characteristics

🤝 Contributing

This Space demonstrates the integration of MONAI models with Hugging Face Spaces. Contributions for improvements are welcome:

• —Model optimizations
• —UI/UX enhancements
• —Additional preprocessing options
• —Performance improvements

📚 References

• —MONAI Framework
• —MONAI Documentation
• —Original Model Repository
• —Gradio Documentation

📄 License

This project is licensed under the Apache License 2.0. See the model repository for specific model licensing terms.

Note: Always ensure you have the necessary permissions and ethical approvals before using medical imaging data.