Title :
Rank-2 model-order selection in diffusion tensor MRI: Infromation complexity based on the total Kullback-Leibler divergence
Author :
Yang, Jianfei ; Poot, Dirk H. J. ; Caan, Matthan W. A. ; Vos, Frans M. ; van Vliet, Lucas J.
Author_Institution :
Quantitative Imaging Group, Delft Univ. of Technol., Delft, Netherlands
Abstract :
Diffusion-weighted MRI (DW-MRI) can assess the integrity of white matter (WM) structures in the human brain. Multi-compartment analysis of DW-MRI requires an estimate of the number of compartments to permit unbiased estimation of the diffusion shape in a single fibers as well as crossing fascicles. We propose a new, rotation-invariant measure to assess the suitability of a model by a measure for information complexity (ICOMP) based on the total Kullback-Leibler divergence (TKLD). ICOMP-TKLD is evaluated on simulated data and on data from the Human Connectome Project. Compared to the state-of-the-art, ICOMP-TKLD is the only method that yields reliable model-order selection in both homogeneous and heterogeneous WM regions. Therefore, ICOM-TKLD may open the way for structure-adaptive estimation of diffusion properties of the entire brain.
Keywords :
biodiffusion; biological tissues; biomedical MRI; brain; information theory; medical computing; neurophysiology; physiological models; DW-MRI; Human Connectome Project data; ICOMP-TKLD evaluation; WM structure integrity assessment; compartment number estimation; crossing fascicle; diffusion shape estimation; diffusion tensor MRI; diffusion-weighted MRI; heterogeneous WM region; homogeneous WM region; human brain diffusion property estimation; information complexity measure; model suitability assessment; multicompartment analysis; rank-2 model-order selection; rotation-invariant measure; simulated data; single fiber; structure-adaptive diffusion property estimation; total Kullback-Leibler divergence; unbiased estimation; white matter; Brain modeling; Complexity theory; Computational modeling; Data models; Diffusion tensor imaging; Estimation; Tensile stress; DTI; model selection;
Conference_Titel :
Biomedical Imaging (ISBI), 2015 IEEE 12th International Symposium on
Conference_Location :
New York, NY
DOI :
10.1109/ISBI.2015.7164022
