Rorden Chris, Fridriksson Julius, Karnath Hans-Otto
Department of Communication Sciences and Disorders, University of South Carolina, SC 29208, USA.
Neuroimage. 2009 Feb 15;44(4):1355-62. doi: 10.1016/j.neuroimage.2008.09.031. Epub 2008 Oct 7.
Kinkingnéhun et al. (Kinkingnéhun, S., Volle, E., Pélégrini-Issac, M., Golmard, J.L., Lehéricy, S., du Boisguéheneuc, F., Zhang-Nunes, S., Sosson, D., Duffau, H., Samson, Y., Levy, R., Dubois, B., 2007. A novel approach to clinical-radiological correlations: Anatomo-Clinical Overlapping Maps (AnaCOM): method and validation. NeuroImage 37: 1237-1249.) have recently described a novel approach for lesion-behavior mapping (LBM), referred to as Anatomo-Clinical Overlapping Maps (AnaCOM). Conventional voxelwise LBM tools apply statistics to contrast behavioral performance of patients with lesions that encompass given voxels to control patients where these voxels are spared. In contrast, AnaCOM contrasts performance of patients with injury involving given voxels to the performance of neurologically healthy participants. The authors correctly note that their procedure can offer substantially more statistical power than conventional LBM methods. We compared AnaCOM to conventional LBM techniques by examining hemiparesis (a common consequence of stroke) as the behavior of interest. We found that AnaCOM detected many regions of the middle cerebral artery territory not associated with the motor system. We suggest that conventional LBM techniques detect regions that are damaged in patients with a deficit while spared in those without a deficit, while AnaCOM detects regions that are associated with a deficit. Therefore, this new measure may offer poor specificity. Furthermore, on theoretical grounds we suggest that permutation-based thresholding will be a more sensitive method for controlling familywise error than the method of counting lesion-overlap clusters used by AnaCOM. Finally, we note that the within group variability tends to be smaller for neurologically healthy controls than in neurological patients, due to ceiling effects. Therefore, we suggest that nonparametric measures or the Welch's t-test are more appropriate than the conventional pooled variance t-test used by AnaCOM.
金克内洪等人（金克内洪，S.，沃勒，E.，佩莱格里尼 - 伊萨克，M.，戈尔马尔，J.L.，勒埃里西，S.，迪布瓦盖内克，F.，张 - 努内斯，S.，索松，D.，迪福，H.，萨姆森，Y.，利维，R.，迪布瓦，B.，2007年。临床 - 放射学相关性的一种新方法：解剖 - 临床重叠图谱（AnaCOM）：方法与验证。《神经影像学》37卷：1237 - 1249页。）最近描述了一种用于病变 - 行为映射（LBM）的新方法，称为解剖 - 临床重叠图谱（AnaCOM）。传统的体素级LBM工具运用统计学方法，对比包含特定体素的病变患者与这些体素未受影响的对照患者的行为表现。相比之下，AnaCOM对比涉及特定体素损伤的患者与神经功能正常参与者的表现。作者正确地指出，他们的方法比传统的LBM方法能提供更大的统计功效。我们通过将偏瘫（中风的常见后果）作为感兴趣的行为，把AnaCOM与传统的LBM技术进行了比较。我们发现AnaCOM检测到许多大脑中动脉区域中与运动系统无关的区域。我们认为传统的LBM技术检测的是有缺陷患者中受损而无缺陷患者中未受损的区域，但AnaCOM检测的是与缺陷相关的区域。因此，这种新方法可能特异性较差。此外，从理论角度我们认为，基于置换的阈值设定将是一种比AnaCOM所使用的计算病变重叠簇的方法更敏感的控制家族性误差的方法。最后，我们注意到由于天花板效应，神经功能正常对照者组内变异性往往比神经疾病患者更小。因此，我们认为非参数测量或韦尔奇t检验比AnaCOM所使用的传统合并方差t检验更合适。
