Reliable Segmentation of Subcortical Gray Matter Structures Using Magnetization Transfer (mt) Maps

Introduction: While much work has been devoted to MRI-based segmentation and voxel-based morphometry (VBM) of cortical anatomy, comparatively few studies have investigated deep grey matter structures. These are of particular interest for study of the structural correlates of neurological and neuropsychiatric disorders associated with basal ganglia dysfunction (1). The relatively small number of segmentation studies is partly due to difficulties delineating these structures using standard methods of structural T1-weighted 3D MRI (for review, see (2)). Automatic segmentation of the basal ganglia and thalamic structures is less reliable, since unlike the cortex, the basal ganglia and thalamus are made up of a large number of neuronal nuclei that are connected by complex and intertwined axonal tracts. Also the high iron content of some midbrain nuclei of the extrapyramidal pathway (e.g., substantia nigra) shortens T1 significantly (3). The partial volume effect and shortened T1 reduce the contrast in T1w images and lead to segmentation problems. We propose the use of parameter maps based on magnetization transfer (MT) contrast to overcome these problems, since MT is a more direct measure of the content of “structural material” such as myelin, which is unlike T1 contrast which chiefly reflects the physical properties of tissue water. We compared the segmentation results derived from MT maps (4) to an established T1w MDEFT method (5) in a cohort of 49 subjects. Considerable improvements were observed in putamen, pallidum, substantia nigra and thalamus.