SynthStroke SPM Toolbox

Automated Stroke Lesion Segmentation for SPM12

SPM12 toolbox implementation of "Synthetic Data for Robust Stroke Segmentation" published in Machine Learning for Biomedical Imaging (MELBA) 2025.

---

About

This SPM12 toolbox provides automated stroke lesion segmentation for brain MRI and CT images using deep learning. The toolbox integrates directly into SPM's batch system and includes a pre-trained U-Net model trained on synthetic stroke data for robust performance across different imaging protocols.

Features:

• Pre-trained U-Net model for automated stroke lesion segmentation
• Native SPM12 batch interface integration
• Support for both T1-weighted MRI and CT imaging
• Test Time Augmentation (TTA) for improved accuracy
• Automatic 1mm isotropic resampling with restoration to original resolution
• Morphological hole filling and configurable probability thresholds

Paper: Chalcroft, L., Pappas, I., Price, C. J., & Ashburner, J. (2025). Synthetic Data for Robust Stroke Segmentation. Machine Learning for Biomedical Imaging, 3, 317–346.

Note: This tool is currently in early development. Please report any issues on GitHub to help improve the tool's reliability.

---

Installation

Prerequisites

• MATLAB (R2024a or later recommended)
• SPM12 (SPM25 should work but is untested)
• Deep Learning Toolbox
• Image Processing Toolbox
• Deep Learning Toolbox Converter for ONNX Model Format (Support Package)

Installation Steps

1. Download the toolbox:

   git clone https://github.com/liamchalcroft/SynthStrokeSPM.git

2. Install ONNX support package:

   • Open MATLAB → HOME tab → Add-Ons → Get Add-Ons
   • Search for "Deep Learning Toolbox Converter for ONNX Model Format"
   • Install the package

3. Place in SPM toolbox directory:

   mv SynthStrokeSPM /path/to/spm12/toolbox/

4. Restart SPM12 or refresh the toolbox menu

The toolbox requires no compilation and works entirely within MATLAB.

---

Usage

Quick Start

1. Open SPM12 and navigate to Batch → SPM → Tools → SynthStroke
2. Test with sample data: Select test_input/ct_img.nii or test_input/mprage_img.nii
3. Run batch to generate segmentation results

Standard Workflow

1. Load images: Select your T1-weighted MRI or CT scans (.nii/.img files)
2. Configure options:
   • Use TTA: Enable for improved accuracy (slower processing)
   • Fill Holes: Remove small gaps in lesion masks
   • Lesion Threshold: Adjust sensitivity (default: 0.5)
   • Output Directory: Specify save location (optional)
3. Run batch processing

Expected Processing Time

• Single image: 1-2 minutes (standard), 5-10 minutes (with TTA)
• Batch processing: Scales linearly with number of images

---

Output Files

The toolbox generates the following files:

• c1* - Background tissue posterior probability map
• c2* - Stroke lesion posterior probability map
• lesion_* - Binary lesion mask (thresholded with optional hole filling)

---

Support

For issues, questions, or contributions:

• Bug reports: Open an issue
• PyTorch implementation: SynthStroke repository
• Questions: Contact via GitHub issues

---

Citation

If you use this work in your research, please cite:

@article{Chalcroft2025,
  title = {Synthetic Data for Robust Stroke Segmentation},
  volume = {3},
  ISSN = {2766-905X},
  url = {http://dx.doi.org/10.59275/j.melba.2025-f3g6},
  DOI = {10.59275/j.melba.2025-f3g6},
  number = {August 2025},
  journal = {Machine Learning for Biomedical Imaging},
  publisher = {Machine Learning for Biomedical Imaging},
  author = {Chalcroft, Liam and Pappas, Ioannis and Price, Cathy J. and Ashburner, John},
  year = {2025},
  month = aug,
  pages = {317–346}
}

---

License

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