Related Experiment Video
Updated: Jun 4, 2025
11:50
A Standardized Pipeline for Examining Human Cerebellar Grey Matter Morphometry using Structural Magnetic Resonance Imaging
Published on: February 4, 2022
3.9K
segcsvdWMH: A Convolutional Neural Network-Based Tool for Quantifying White Matter Hyperintensities in Heterogeneous
Erin Gibson1,2, Joel Ramirez1, Lauren Abby Woods1
1SEB Centre for Brain Resilience & Recovery, Hurvitz Brain Sciences Program, Sunnybrook Research Institute, Toronto, Canada.
Human Brain Mapping
|December 26, 2024
Summary
A new AI tool, segcsvdWMH, accurately quantifies white matter hyperintensities (WMH) in brain MRI scans. This tool improves the measurement of WMH, a key marker for cerebral small vessel disease (CSVD), across diverse patient populations and imaging data.
Area of Science:
- Neuroimaging
- Artificial Intelligence in Medicine
- Cerebrovascular Diseases
Background:
- White matter hyperintensities (WMH) are MRI biomarkers for cerebral small vessel disease (CSVD).
- WMH are linked to cognitive decline, stroke, and dementia.
- Accurate WMH measurement is challenging in large, diverse clinical studies due to imaging variability.
Purpose of the Study:
- To develop and validate segcsvdWMH, a convolutional neural network tool for reliable WMH quantification.
- To improve WMH measurement accuracy across heterogeneous clinical datasets and imaging protocols.
Main Methods:
- Developed segcsvdWMH using a large dataset (>700 FLAIR MRI scans) from seven multisite studies.
- Employed a hierarchical segmentation approach and extensive data augmentation for model training.
- Benchmarked segcsvdWMH against HyperMapp3r, SAMSEG, and WMH-SynthSeg on diverse test datasets.
Main Results:
- segcsvdWMH demonstrated superior accuracy over existing tools, with significant Dice score improvements.
- Achieved a high mean Dice score (0.86 ± 0.08) across diverse test datasets.
- Showed robust performance against noise artifacts and varying WMH burdens, with strong correlation to ground truth volumes (mean r = 0.99 ± 0.01).
Conclusions:
- segcsvdWMH offers accurate and robust WMH segmentation for diverse clinical data.
- The tool is suitable for large-scale studies involving patients with varying CSVD severity.
- Improved WMH quantification can aid in understanding CSVD progression and clinical outcomes.
