Validation Of A Thalamus Segmentation Based On Local Difusion Information

Introduction: Fast and accurate segmentation of thalamic nuclei is important for clinical applications such as treatment of Parkinson's disease or chronic pain syndrome for example. The intrinsic contrast in standard anatomical MR images in the thalamus is very low, therefore direct segmentation of thalamic nuclei is not possible. Since thalamic nuclei have a distinct and well-ordered fiber structure, DTI promises better contrast within the thalamus. Several methods for the segmentation of thalamic nuclei on DTI data have been proposed (see for example [1-2]). Even though these methods show impressive results they require manual segmentation of the whole thalamus or other brain regions, finetuned segmentation parameters and their computation time is in some cases [2] very high. We therefore refine a segmentation method based on local Dominant Diffusion Orientation (DDO) classification that is fast and independent of manual segmentation [3]. Using a group of 63 healthy subjects, we show the anatomical relevance of our DDO-based segmentation method. To prove that the segmented nuclei in individual subject data correspond well to each other we evaluate the spread of the center-of-mass (COM) coordinates for the individual subjects and clusters. Methods: Data: DTI data and T1 weighted images were acquired for 63 healthy right-handed volunteers (32 male and 31 female) between 20 and 40 years old (mean 26.29 years +/4.48; females: mean 25.52 years +/4.35) on a 1.5T MR-