Morphological analysis of 3D SPECT images via nilpotent t-norms in diagnosis of Alzheimer's disease

Abstract

We propose a fuzzy image processing approach serving as a potential diagnostic tool for Alzheimer's disease. We have set up a sequence of several image-processing methods based on morphological fuzzy edge detection followed by a region-growing segmentation. We applied these operations to two sets of three-dimensional SPECT images of human brains: one set of images was composed of patients with Alzheimer's disease and the second was represented by control brains from non-demented patients. Then we undertook an analysis of nilpotent t-norms forming the edge detectors and the parameter of Gaussian filter, and in the end we carried out an evaluation of segments that was computed after the final watershed segmentation. This article provides a detailed description of these methods as well as the analysis of evaluated data using Student's two-sample t-test. Traditional gradient magnitude edge detectors are included and compared with fuzzy detectors. Our goal was to demonstrate the usefulness of the Łukasiewicz BL-algebra in feature extraction of 3D biomedical images by using enhanced methods of image morphology.