Fast and Efficient Diffusion-based Thalamic Segmentation Using Spectral Clustering

Motivation: Current diffusion-based thalamus segmentation methods are limited by time-consuming cortical parcellation and inefficiencies in handling complex intricate small-scale fiber tract geometries. Goal(s): Develop a fast and efficient segmentation approach to overcome the limitations of traditional methods, such as k-means clustering and lengthy parcellation processes. Approach: We utilize THOMAS for automated thalamus masking, MSMT-CSD for enhanced ODFs, and spectral clustering to effectively capture the intricate geometry of thalamic nuclei. Results: Tested on 15 subjects, our approach successfully segmented the thalamus into seven clusters and the pulvinar nucleus into three sub-clusters, demonstrating fast and efficient segmentation, and the ability to handle complex anatomical structures. Impact: Fast and accurate subthalamic segmentation can enable more accurate thalamic studies and interventions, thereby improving both our understanding of brain pathology and patient outcomes in various neurological conditions