Assessing Spatial Probabilistic Distributional Differences in the Common Space Between Schizophrenics and Normal Controls Based on A Novel Automated Probabilistic Pattern Analysis Method

Abstract

Because of the complex nature of the human brain, a full understanding of its various group specific variation factors such as volume, shape, and location related to age, gender, ethnic, and disease might be provided in both structural and functional neuroimaging studies. To serve this purpose, a novel approach for characterizing the group variability information using group specific labeled probabilistic maps was introduced in this article. An automatic labeling technique was applied to encode group specific probabilistic information for each region of interests (ROIs) covering the overall cortical region and a probabilistic pattern analytic method was proposed to assess the difference in the spatial extent between 70 schizophrenics and 70 controls in the common space. From our proposed method, we found major differences in 17 ROIs that had shown large variation in schizophrenics. Most of these ROIs were in the frontal and the temporal lobe and only three ROIs were in the parietal and the occipital lobe. The ROIs highlighted through our propose method could be connected with previous morphological findings on schizophrenia and it also might be considered in functional analysis. As a result, our method could provide intuitive information on group difference relevant to the overall anatomical variability in the sub-structural level. Thus, it could be used as a prompting system to search and examine the regions of the brain that are worthy of further precise analysis by various sub-cortical region based group studies in assessing specific patterns related to diseases.