Abstract
The paper is oriented to the computer classification of Alzheimer disease via analysis of 3D SPECT image of human brain perfusion, which were scanned via scintillation camera and standardized by SPM technique. After the thresholding (60 % of maximum intensity) and space segmentation, the digital morphological analysis was performed including the analysis of convex hulls.
The linear classifier was applied to the set of morphological characteristics with three aims: unit sensitivity, unit specificity and minimum norm of weight vector to obtain robust recognition system with generalization power. The best classifier consists of two morphological characteristics (internal sphere radii, one half of maximum voxel distance) for nine space regions of 3D map.