BIASS: Benchmarking the Impact of Anatomical Segmentation in Spectroscopy

Motivation: Water-scaled metabolite estimates in brain magnetic resonance spectroscopy (MRS) usually require corrections accounting for differences in water content and water and metabolite relaxation behavior gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF). Goal(s): This study aims to determine how strongly commonly used segmentation tools differ in determining partial volumes from a typical MRS single-voxel volume. Approach: By analyzing GM, WM, and CSF fractions obtained from segmentation tools (e.g., FSL, ANTs, and SPM) and comparing them to manual segmentation as "ground truth". Results: Significant variability in tissue fraction estimates across segmentation methods for different brain regions, with CSF showing the highest inconsistencies. Impact: This study demonstrates that segmentation tools obtain different tissue volume fraction estimates in a typical MRS voxel. These findings help researchers and clinicians understand the variability that the choice of segmentation algorithm contributes to the uncertainty of water-scaled concentration estimates.