Skeleton-based gyri sulci separation for improved assessment of cortical thickness

In order to improve classification of neurological diseases involving cortical thinning, this work proposes an approach for separating gyral and sulcal regions of the human cortex. Using data from magnetic resonance imaging, the skeleton of the brain's white matter was reconstructed and a geodesic distance measure was applied to separate gyri and sulci. Cortical thickness per subregion was measured for the entire cortex and for gyri and sulci individually in 21 patients with Alzheimer's disease, 10 patients with frontotemporal lobar degeneration composed of two subgroups and 13 control subjects. For discrimination using logistic regressions, which was assessed using leave-one-out cross-validation, improved results were obtained in five out of six group comparisons when cortical thickness measurements were constrained to gyral or sulcal regions