Spatial consistency in 3D tract-based clustering statistics

• Published in: Medical image computing and computer-assisted intervention : MICCAI ... International Conference on Medical Image Computing and Computer-Assisted Intervention Vol. 11; no. Pt 1; p. 535
• Main Authors: Caan, Matthan W A, van Vliet, Lucas J, Majoie, Charles B L M, Aukema, Eline J, Grimbergen, C A, Vos, Frans M
• Format: Journal Article
• Language: English
• Published: Germany 2008
• Subjects: Algorithms; Artificial Intelligence; Cerebellar Neoplasms - diagnosis; Child; Cluster Analysis; Computer Simulation; Diffusion Magnetic Resonance Imaging - methods; Humans; Image Enhancement - methods; Image Interpretation, Computer-Assisted - methods; Imaging, Three-Dimensional - methods; Medulloblastoma - diagnosis; Models, Biological; Models, Statistical; Nerve Fibers, Myelinated - pathology; Pattern Recognition, Automated - methods; Reproducibility of Results; Sensitivity and Specificity
• DOI: 10.1007/978-3-540-85988-8_64

We propose a novel technique for tract-based comparison of DTI-indices between groups, based on a representation that is estimated while matching fiber tracts. The method involves a non-rigid registration based on a joint clustering and matching approach, after which a 3D-atlas of cluster center points is used as a frame of reference for statistics. Patient and control FA-distributions are compared per cluster. Spatial consistency is taken to reflect a significant difference between groups. Accordingly, a non-parametric classification is performed to assess the continuity of pathology over larger tract regions. In a study to infant survivors treated for medulloblastoma with intravenous methotrexate and cranial radiotherapy, significant decreases in FA in major parts of the corpus callosum were found.