Beware of white matter hyperintensities causing systematic errors in FreeSurfer gray matter segmentations!

Area of Science:

• Neuroimaging
• Neurology
• Medical image analysis

Background:

• Volumetric estimates of subcortical and cortical structures from T1-weighted MRIs are crucial for clinical and research applications.
• White matter hyperintensities (WMHs) are common in aging and neurological conditions and may affect brain structure volume estimations.
• The FreeSurfer software is widely used for automated brain structure segmentation, but its accuracy in the presence of WMHs is not fully understood.

Purpose of the Study:

• To investigate the impact of WMHs on FreeSurfer's gray matter (GM) volume estimates.
• To assess whether WMHs introduce bias in the relationships between GM volumes and clinical variables like age, diagnosis, and cognitive performance.
• To determine if correcting for WMHs improves the accuracy of these associations.

Main Methods:

• T1-weighted MRI data from 1,077 participants (4,321 timepoints) from the Alzheimer's Disease Neuroimaging Initiative were processed using FreeSurfer v6.0.0.
• WMHs were segmented using a validated algorithm on T2-weighted or FLAIR images.
• Mixed-effects models were employed to analyze the relationships between WMH overlap, GM volumes, and clinical factors.

Main Results:

• Higher WMH volumes showed significant overlap with GM volumes in several brain regions, including the caudate, cerebral cortex, putamen, thalamus, pallidum, and accumbens.
• Uncorrected caudate volumes appeared to increase with age and did not differ between healthy controls and Alzheimer's disease patients.
• After WMH correction, caudate volumes decreased with age, were lower in Alzheimer's disease patients, and showed a significant association with cognitive performance.

Conclusions:

• WMHs cause systematic inaccuracies in FreeSurfer's GM segmentation, particularly affecting the caudate nucleus.
• These inaccuracies can alter clinically relevant associations with age, Alzheimer's disease diagnosis, and cognitive function.
• Accounting for WMHs is essential for accurate brain volumetry and reliable clinical interpretations, likely impacting various segmentation algorithms.