Decoding the Best Automated Segmentation Tools for Vascular White Matter Hyperintensities in the Aging Brain: A...

Area of Science:

• Neuroimaging
• Gerontology
• Vascular Neurology

Background:

• Small vessel disease (SVD) causes cerebrovascular damage, with white matter hyperintensities (WMH) being a key marker in aging.
• Current visual assessment of WMH is time-consuming and prone to rater variability, limiting clinical utility.
• Developing precise automated WMH segmentation is crucial for reliable clinical and research applications.

Approach:

• Evaluated four automated neuroimaging tools (LST's LPA/LGA, SAMSEG, BIANCA) for WMH segmentation using T1, FLAIR, and DWI MRI data from 300 healthy older adults.
• Established a manual segmentation gold standard from a subsample (n=45) for comparison.
• Assessed algorithm performance, correlation with the Fazekas scale, and impact of WMH volume on reliability.

Key Points:

• Supervised WMH segmentation algorithms demonstrated superior performance, especially in detecting small lesions.
• Combining supervised and unsupervised tools enhanced the consistency of WMH detection.
• A novel biomarker for moderate vascular damage was proposed, based on the 95th percentile of WMH volume in healthy individuals aged 50-60.

Conclusions:

• Automated tools, particularly supervised algorithms, offer a more precise and practical alternative to manual WMH assessment.
• The proposed biomarker effectively identifies subgroups with varying brain structure and behavioral correlates.
• This research advances the quantitative analysis of SVD and its impact on cognitive aging.