Meeting Banner
Abstract #3229

Comparing Two Atlas-Based Automatic Segmentation Methods for Subthalamic Nucleus Deep Brain Stimulation

Y. Xiao1, L. Bailey1, M. Mallar Chakravarty2, S. Beriault1, A. F. Sadikot3, G. Bruce Pike1, D. Louis Collins1

1McConnell Brain Imaging Centre, Montral Neurological Institute, McGill University, Montral, Qubec, Canada; 2Kimel Family Translational Imaging-Genetic Laboratory, Research Imaging Centre, Centre for Addiction, Toronto, Canada; 3Division of Neurosurgery, McGill University, Montral, Canada

Accurate automated segmentation of the subthalamic nucleus (STN) and its surrounding structures can help improve planning for deep brain stimulation procedures when treating Parkinsons disease. Two atlas-based automatic segmentation methods, one with histologically defined atlas, and the other with in vivo T2w-image-defined atlas, were compared for the segmentation of the STN, red nucleus, and substantia nigra, by using the manual segmentation on the consensus of T2*w image and R2* map acquired on a 3T scanner as the ground truth. Although they both provided acceptable results, the in vivo T2w-image-defined atlas performed better on average.

Keywords

abstract account achieve achieved addiction almost anterior assist associative assumption atlas atlases automatic averaging basal belong best better biases bilaterally black blue brain coarse comparable consensus constraints contrasts coordinate correspondence currently deep defined deformed derived described differ disease distance distances division echoes effective encoding enhance euclidean evaluation explained exponential fact family female fitting five fully ganglia generated genetic hand healthy histological important in vivo incorporating inferior inherent institute instrumental inter invisible kappa known labels laboratory larger lateral limbic linearly loss magnitude male manual manually maps matrix medial metric metrics motor neurological neuron neurosurgery neutralized novel nuclei nucleus outlines overlap painting partial particularly patient patients performance performances please population posterior previously proper providing referred registered relatively resolution rigid scanner segmentation segmentations segmenting significantly silver since slice snap space stimulation strategy structural structure structures studied subject subjects suggest superior susceptibility table target therapeutic transforming transforms treatment trio undesirable various visualize volume volumes warped warping worse