Optimizing automated white matter hyperintensity segmentation in individuals with stroke

Abstract

White matter hyperintensities (WMHs) are a risk factor for stroke. Consequently, many individuals who suffer a stroke have comorbid WMHs. The impact of WMHs on stroke recovery is an active area of research. Automated WMH segmentation methods are often employed as they require minimal user input and reduce risk of rater bias; however, these automated methods have not been specifically validated for use in individuals with stroke. Here, we present methodological validation of automated WMH segmentation methods in individuals with stroke. We first optimized parameters for FSL's publicly available WMH segmentation software BIANCA in two independent (multi-site) datasets. Our optimized BIANCA protocol achieved good performance within each independent dataset, when the BIANCA model was trained and tested in the same dataset or trained on mixed-sample data. BIANCA segmentation failed when generalizing a trained model to a new testing dataset. We therefore contrasted BIANCA's performance with SAMSEG, an unsupervised WMH segmentation tool available through FreeSurfer. SAMSEG does not require prior WMH masks for model training and was more robust to handling multi-site data. However, SAMSEG performance was slightly lower than BIANCA when data from a single site were tested. This manuscript will serve as a guide for the development and utilization of WMH analysis pipelines for individuals with stroke.

White matter hyperintensities modify relationships between corticospinal tract damage and motor outcomes after stroke

“…(Lo et al ., 2023) Moreover, we have previously demonstrated that our automated WMH segmentation protocol is robust across multiple sites in individuals with stroke. (Ferris et al ., 2023) A limitation of the multisite and secondary nature of our study sample is that we had limited availability of additional covariates that may influence WMH severity and stroke outcomes, most notably cardiometabolic risk factors such as diabetes and hypertension. (Jeerakathil et al ., 2004) Another limitation is that our sample had mild motor impairment overall (median impairment of 5%), which is typical of neuroimaging samples of motor impairment after stroke.…”

“…(Liew et al ., 2018; Lo et al ., 2023) WMHs were segmented with Freesurfer’s SAMSEG, which we previously established has robust performance in multi-site data from individuals with stroke. (Ferris et al ., 2023) Linearly co-registered T1 and FLAIR/T2 scans were used as inputs into SAMSEG, and a 0.1 probability threshold was applied to tissue segmentation. We subtracted stroke lesion masks from segmented WMHs to prevent any misclassification of stroke lesions as WMHs.…”

“…(Boyd et al, 2017) WMH damage was operationalized by whole-brain WMH volume, measured with automated WMH segmentation. (Ferris et al, 2023) We hypothesized that the relationship between motor impairment and CST-LL would be attenuated in individuals with higher WMH volumes, indicating a greater influence of concurrent WMHs on motor outcomes after stroke. This hypothesis was tested over two analyses: 1) a beta regression analysis testing the effects of WMH volume, CST-LL, and their interaction on motor impairment across the whole sample, and 2) beta regression analyses in subsamples stratified by severity of WMHs (mild versus moderate-severe WMHs) to assess if WMH severity modifies relationships between motor impairment, CST-LL, and WMH volume.…”

“…(Boyd et al ., 2017) WMH damage was operationalized by whole-brain WMH volume, measured with automated WMH segmentation. (Ferris et al ., 2023)…”