Automated detection of progression of white matter hyperintensities in cerebral small vessel disease using machine learning

Luijten, Matthijs

Automated detection of progression of white matter hyperintensities in cerebral small vessel disease using machine learning

Files

Luijten, M. s-4787757-MSc-MKI92-Thesis-2023.pdf (10.88 MB)

Authors

Luijten, Matthijs

Issue Date

2023-03-01

Language

en

Abstract

White matter hyperintensities (WMH) are often seen in the brain of healthy elderly people. However, WMH are also biomarkers for the prognosis or disease progression of cerebral small vessel disease (CSVD). WMH, seen as hyperintense on fluid-attenuated inversion recovery (FLAIR)- and T2-weighted MRI sequences, are amongst the foremost recurrent CSVD features observed on MRI scans and represent different degrees of axonal loss, demyelination, and gliosis. This research paper investigates the progression of WMH and normal-appearing white matter (NAWM) in human brains through in vivo and post-mortem MRI segmentation and clustering of different regions. The proposed method achieves high segmentation performance with Dice coefficients of 0.76 for in vivo (compared to 0.805 for an independent human observer) and up to 0.94 for post-mortem MRI data. Clustering experiments successfully distinguish different levels of severity for WMH and NAWM using FLAIR intensity. Additionally, the study provides new insights into the mechanisms of WMH and its positive relationship with myelin loss. These findings have important implications for the development of new therapies to slow or halt the progression of WMH and related disorders. Future research should explore the relationship between WMH and other markers of white matter damage, such as axonal injury and inflammation, and perform larger-scale studies to understand the pathogenesis of WMH further.