Automatic Segmentation of the Subthalamic Nucleus: A Viable Option to Support Planning and Visualization of Patient-Specific Targeting in Deep Brain Stimulation

Abstract

Abstract BACKGROUND Automatic segmentation is gaining relevancy in image-based targeting of neural structures. OBJECTIVE To evaluate its feasibility, we retrospectively analyzed the concordance of magnetic resonance imaging (MRI)-based automatic segmentation of the subthalamic nucleus (STN) and intraoperative microelectrode recordings (MERs). METHODS Electrodes (n = 60) for deep brain stimulation were implanted in the STN of patients (n = 30; median age 57 yr) with Parkinson disease (n = 29) or rapid-onset dystonia parkinsonism (n = 1). Elements (Brainlab, Munich, Germany) were used to segment the STN, using 2 volumetric T1 (±contrast) and volumetric T2 images as input. The stereotactic computed tomography was coregistered with the imaging, and the original stereotactic coordinates were imported. MERs (0.5-1 mm steps) along the anterior, central, and lateral trajectories were used to determine differences between the image-segmented STN boundary and MER-based STN entry and exit. RESULTS Of 175 trajectories, 105 penetrated or touched (≤0.7 mm) the STN. The overall median deviation between the segmented STN boundary and electrophysiological recordings was 1.1 mm for MER-based STN entry and 2.0 mm for STN exit. Regarding the entry point of the STN, there was no statistically significant difference between MRI-based automatic segmentation and the electrophysiological trajectories analyzed with intraoperative MER. The exit point was significantly different between both methods in the central and lateral trajectories. CONCLUSION MRI-based automatic segmentation of the STN is a viable, patient-specific targeting approach that can be used alongside traditional targeting methods in deep brain stimulation to support preoperative planning and visualization of target structures and aid postoperative optimization of programming.