Extracting and summarizing white matter hyperintensities using supervised segmentation methods in Alzheimer's disease...

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Gerontology

Background:

• White matter hyperintensities (WMH) in MRI scans are linked to aging and neurological disorders like Alzheimer's disease.
• WMH can indicate co-morbid neural injury or cerebrovascular disease burden.
• The small, diffuse, and heterogeneous nature of WMH presents segmentation challenges.

Purpose of the Study:

• To develop and evaluate supervised machine learning models for precise detection and quantification of WMH.
• To assess the association of WMH accumulation with cognitive status and family history of dementia.
• To provide an open-source library for WMH detection and accumulation analysis.

Main Methods:

• Adapted Support Vector Machines and Random Forests for WMH detection.
• Utilized texture features from texton filter banks for image segmentation.
• Evaluated methods on healthy middle-aged and older adults with varying Alzheimer's disease risk.

Main Results:

• The developed methods demonstrated reliable and robust segmentation of hyperintense regions.
• Normalized effective WMH volume was associated with dementia in older adults.
• Parental family history of dementia correlated with WMH in cognitively normal subjects.

Conclusions:

• Machine learning approaches, using texture features, offer effective WMH segmentation.
• WMH accumulation is a relevant biomarker for cognitive decline and dementia risk.
• The open-source library facilitates broader application in neuroimaging research.