Detection of cortical transition regions utilizing statistical analyses of excess masses

Abstract

A new statistical approach for observer-assisted detection of transition regions of adjacent cytoarchitectonic areas within the human cerebral cortex was developed. This method analyzes the structural information of cytoarchitectural profiles (e.g., the modality of a gray level intensity distribution) based on observed excess mass differences verified by a suitable statistical test. Profiles were generated by scanning the cerebral cortex over respective regions of interest that were oriented to trajectories running parallel to the orientation of cell columns. For each single profile, determination of excess masses provided evidence for a certain number of peaks in the cell density, thereby avoiding fluctuation due solely to sampling anomalies. Comparing such excess mass measurements by means of multiple local rank tests over a wide range of profiles allowed for the detection of cytoarchitectural inhomogeneities at respective given confidence levels. Special parameters (e.g., level of significance, width of targeted region, number of peaks) then could be adapted to specific pattern recognition problems in lamination analyses. Such analyses of excess masses provided a general tool for observer-assisted evaluation of profile arrays. This observer-assisted statistical method was applied to five different cortical examples. It detected the same transition regions that had been determined earlier through direct examination of samples, despite cortical convexities, concavities, and some minor staining inhomogeneities.