Abstract

Tissue clustering and classification are among the most challenging tasks in DT image analysis. While classification identifies the tissue type within a voxel, clustering identifies regions of interest (ROI) in which tissue properties are similar. The aim of this work is to propose and investigate the effectiveness of unsupervised tissue clustering and classification algorithms for DTI data. The former employs four possible models to describe diffusion in each seed voxel; a model selection framework, adapted from Snedecors F-test is used to choose the most parsimonious model. The latter assesses the spatial homogeneity of the distribution of the entire diffusion tensor using the statistical framework of Hext and Snedecor, in which the null hypothesis of diffusion tensors having a similar parameter distribution is determined by an F-test. Both numerical phantoms and DWI data obtained from excised rat spinal cord are used to test and validate these tissue clustering and classification approaches.



Close Window