Supplementary Material for: Quantifying the Variability Associated with Postoperative Localization of Deep Brain Stimulation Electrodes

Introduction: Computational models of deep brain stimulation (DBS) have become common tools in clinical research studies that attempt to establish correlations between stimulation locations in the brain and behavioral outcome measures. However, the accuracy of any patient-specific DBS model depends heavily upon accurate localization of the DBS electrodes within the anatomy, which is typically defined via co-registration of clinical CT and MRI datasets. Several different approaches exist for this challenging registration problem, and each approach will result in a slightly different electrode localization. The goal of this study was to better understand how different processing steps (e.g., cost-function masking, brain extraction, intensity remapping) affect the estimate of the DBS electrode location in the brain. Methods: No “gold standard” exists for this kind of analysis, as the exact location of the electrode in the living human brain cannot be determined with existing clinical imaging approaches. However, we can estimate the uncertainty associated with the electrode position, which can be used to guide statistical analyses in DBS mapping studies. Therefore, we used high-quality clinical datasets from 10 subthalamic DBS subjects and co-registered their long-term postoperative CT with their preoperative surgical targeting MRI using 9 different approaches. The distances separating all of the electrode location estimates were calculated for each subject. Results: On average, electrodes were located within a median distance of 0.57 mm (0.49–0.74) of one another across the different registration approaches. However, when considering electrode location estimates from short-term postoperative CTs, the median distance increased to 2.01 mm (1.55–2.78). Conclusions: The results of this study suggest that electrode location uncertainty needs to be factored into statistical analyses that attempt to define correlations between stimulation locations and clinical outcomes.

引言：脑深部电刺激（deep brain stimulation, DBS）的计算模型已成为临床研究中的常用工具，这类研究旨在探究脑内刺激位点与行为学结局指标之间的相关性。然而，任何患者特异性DBS模型的准确性高度依赖于脑解剖结构内DBS电极的精准定位，而该定位通常通过临床计算机断层扫描（computed tomography, CT）与磁共振成像（magnetic resonance imaging, MRI）数据集的共配准来确定。针对这一颇具挑战性的配准问题，目前存在多种不同的解决方案，且每种方案得到的电极定位结果均存在细微差异。本研究的目标在于深入探究不同处理步骤（例如代价函数掩膜、脑区提取、强度重映射）如何影响脑内DBS电极位置的估计结果。
方法：此类分析尚无"gold standard"，因为现有临床成像手段无法确定活体人脑内电极的精确位置。不过，我们可以估算与电极位置相关的不确定性，该结果可用于指导DBS定位研究中的统计分析。因此，我们纳入了10名接受丘脑底核DBS手术的受试者的高质量临床数据集，采用9种不同的配准方案，将其术后长期计算机断层扫描图像与术前手术靶向磁共振成像图像进行共配准。针对每名受试者，我们计算了所有电极位置估计结果之间的间距。
结果：总体而言，不同配准方案得到的电极位置估计结果的中位间距为0.57 mm（四分位距：0.49–0.74 mm）。但当采用术后短期计算机断层扫描图像进行电极位置估计时，中位间距升至2.01 mm（四分位距：1.55–2.78 mm）。
结论：本研究结果表明，在旨在明确刺激位点与临床结局之间相关性的统计分析中，需将电极位置不确定性纳入考量范围。