NeuroScan AI - Brain Tumor Segmentation

Overview

NeuroScan AI is a deep learning-powered brain tumor segmentation tool that leverages a U-Net architecture to automatically detect and segment brain tumors in MRI scans. This application provides a user-friendly interface for medical professionals and researchers to analyze brain MRI images with high accuracy.

Features

• U-Net Architecture: State-of-the-art segmentation model for medical imaging
• Test-Time Augmentation (TTA): 4-fold ensemble for more robust predictions
• CLAHE Enhancement: Contrast Limited Adaptive Histogram Equalization for improved image quality
• Adaptive Thresholding: Intelligent threshold selection for accurate segmentation
• Post-processing: Morphological operations to refine segmentation masks
• Multi-format Support: Accepts JPG, PNG, TIF, and BMP image formats
• Clinical Visualization: Overlay, heatmap, and binary mask outputs
• Performance Metrics: Tumor area percentage, confidence scores, and severity classification

Architecture

The application uses a custom U-Net implementation with the following components:

1. Data Preprocessing:

   • CLAHE enhancement in LAB color space
   • Multi-channel image handling
   • Normalization to model requirements

2. Model Architecture:

   • Encoder-decoder structure with skip connections
   • Double convolution blocks with batch normalization
   • 4-level feature extraction (64, 128, 256, 512 features)

3. Inference Pipeline:

   • Test-time augmentation (4-fold)
   • Adaptive threshold selection using Otsu's method
   • Post-processing with morphological operations

4. Visualization:

   • Overlay visualization with customizable opacity
   • Heatmap generation for confidence visualization
   • Binary mask output for further analysis

Installation

Prerequisites

• Python 3.8 or higher
• PyTorch 2.11.0
• CUDA-compatible GPU (recommended for faster inference)

Setup

1. Clone the repository:

git clone <repository-url>
cd neuroscan-ai

2. Create a virtual environment:

python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

4. Download the pre-trained model:
   • Place unet_brain_tumor.pth in the project root directory

Usage

1. Run the Streamlit application:

streamlit run app.py

2. Upload a brain MRI scan (JPG, PNG, TIF, or BMP format)

3. Configure analysis settings in the sidebar:

   • Enable/disable Test-Time Augmentation
   • Adjust overlay opacity

4. Click "Run Segmentation" to process the image

5. View results in the tabs:

   • Overlay: Original image with tumor overlay
   • Side-by-Side: Original and segmented images
   • Heatmap: Confidence probability map
   • Mask: Binary segmentation mask

6. Download results using the export buttons

Model Details

• Architecture: U-Net with skip connections
• Input Size: 256×256 pixels
• Training Data: TCGA-LGG and similar brain MRI datasets
• Output: Binary segmentation mask of tumor regions
• Device: Automatically uses GPU if available, falls back to CPU

Technical Implementation

Key Components

1. Preprocessing:

   • CLAHE enhancement in LAB color space for improved contrast
   • Multi-channel handling for various input formats

2. Inference:

   • 4-fold Test-Time Augmentation (horizontal flip + brightness variations)
   • Adaptive thresholding using Otsu's method with fallback

3. Post-processing:

   • Morphological closing and opening operations
   • Minimum area filtering to remove noise

4. Visualization:

   • Color-coded overlays (blue for tumor regions)
   • Heatmap visualization with Inferno colormap
   • Export functionality for all results

Performance Optimizations

• Efficient U-Net implementation with skip connections
• GPU acceleration when available
• Optimized image processing pipeline
• Memory-efficient inference with batch processing

Contributing

We welcome contributions to improve NeuroScan AI! Please follow these steps:

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Submit a pull request

Please ensure your code follows the existing style and includes appropriate tests.

License

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

Disclaimer

⚠️ Research Use Only - This tool is not a medical device. Always seek a qualified radiologist for clinical diagnosis.

Contact

For questions or support, please open an issue on the GitHub repository.