This repository contains the implementation of the Fuzzy Information Seeded Region Growing (FISRG) algorithm for segmenting stroke lesions in brain MRI images. Developed as part of a research project, this algorithm combines fuzzy logic with Seeded Region Growing techniques to tackle the complex task of delineating irregular and diverse textures of stroke lesions.
- Robust Segmentation: Capable of handling complex textures and irregular boundaries of stroke lesions in brain MRI scans.
- Optimized Performance: Fine-tuned parameters for balancing segmentation accuracy and computational efficiency.
- Tested on ATLAS Dataset: Validated and optimized using the comprehensive ATLAS dataset for stroke lesion MRI images.
This project requires the following libraries:
- OpenCV (cv2): For image processing tasks.
- Pandas: For data manipulation and analysis.
- NumPy: For numerical operations on arrays and matrices.
- Matplotlib: For creating static, interactive, and animated visualizations.
- NiBabel: For reading and writing neuroimaging data formats.
- Scikit-learn (sklearn): For KMeans clustering and other machine learning tasks.
- os: For interacting with the operating system.
To install these dependencies, run the following commands:
pip install opencv-python
pip install pandas
pip install numpy
pip install matplotlib
pip install nibabel
pip install scikit-learn
Contributions to this project are welcome.
This project is licensed under the MIT License.
Special thanks to Dr. Enrique Nava Baro and Dr. Ezequiel Lopez Rubio for their guidance and support throughout this project.
If you use this algorithm or find this project helpful, please cite:
@article{FISRGStrokeMario,
title={Fuzzy Information Seeded Region Growing for Automated Lesions After Stroke Segmentation in MR Brain Images},
author={Mario Pascual González},
journal={ArXiv},
year={2023},
eprint={2311.11742},
archivePrefix={arXiv},
primaryClass={eess.IV},
doi={10.48550/arXiv.2311.11742},
pages={10},
note={14 figures. Associated code and data available online}
}
Dice coefficient for each Experiment between the predicted mask and the ground truth mask. The algorithm effectively segments the stroke lesion with a maximum dice coefficient of 94.2% and archives the highest average accuracy in Experiment 3, with an average dice coefficient of 88.1%
Statistic | Experiment 1 | Experiment 2 | Experiment 3 |
---|---|---|---|
mean | 0.849 | 0.865 | 0.881 |
std | 0.071 | 0.039 | 0.034 |
min | 0.481 | 0.742 | 0.771 |
max | 0.934 | 0.942 | 0.942 |
For any queries, please reach out to my LinkedIn.