Department of Neurosurgery, Klinikum Bogenhausen, Städtisches Klinikum München, Englschalkinger Straße 77, 81925, Munich, Germany.
Acta Neurochir (Wien). 2013 Apr;155(4):675-83. doi: 10.1007/s00701-013-1624-1. Epub 2013 Feb 6.
Functional magnetic resonance imaging (fMRI) is a widely available method and is therefore progressively utilized in neurosurgical practice. This study was carried out to determine fMRI sensitivity and specificity and to emphasize the threshold dependence of fMRI data.
A total of 17 consecutive patients, scheduled for surgery on intracerebral lesions near eloquent brain areas, underwent preoperative motor (N = 12) and language (N = 5) fMRI. Functional data were analyzed with SPM software and displayed on a neuronavigation system for intraoperative guidance. High-risk maps for motor and language deficits obtained from direct electric cortical stimulation (DECS) were used for validation of functional activated areas. In a first analysis step, sensitivity and specificity were calculated in terms of a side-by-side correlation. The next step, the threshold dependence of fMRI data sensitivity and specificity, was estimated according to four statistical thresholds (p1 < 0.05, p2 < 0.0005, p3 < 0.00001, p4 < 0.0000001).
Both functional imaging and DECS revealed definite results for the investigated areas in all patients. Calculation of sensitivity and specificity resulted in 100 % and 68 % for the motor group and a sensitivity of 75 % and specificity of 68 % for the language group at the fixed threshold analysis. Threshold-dependent analysis of the obtained data revealed a sensitivity/specificity relationship from 100 %/0 % at threshold (p1), 100 %/5 % at (p2), 74 %/9 % at (p3), and 37 %/36 % at (p4) for the motor group. Evaluation of threshold-dependent sensitivity and specificity for the language group resulted in 78 %/51 % at threshold (p1), 67 %/75 % at (p2), 50 %/78 % at (p3), and 33 %/89 % at (p4).
The present findings on the threshold dependence of fMRI data demonstrate why individualized thresholds should be obtained in case of fMRI evaluation. Although the results are satisfying in most cases, fMRI is apparently not sufficient for critical intraoperative decision-making.
功能磁共振成像（fMRI）是一种广泛应用的方法，因此在神经外科实践中逐渐得到应用。本研究旨在确定 fMRI 的敏感性和特异性，并强调 fMRI 数据的阈值依赖性。
17 例连续患者，计划对位于语言功能区附近的颅内病变进行手术，术前进行运动（N=12）和语言（N=5）功能磁共振成像。功能数据采用 SPM 软件进行分析，并显示在神经导航系统上，用于术中指导。使用直接皮质电刺激（DECS）获得的运动和语言缺陷高风险图验证功能激活区。在第一步分析中，根据侧对侧相关性计算敏感性和特异性。下一步，根据四个统计阈值（p1<0.05、p2<0.0005、p3<0.00001、p4<0.0000001），估计 fMRI 数据敏感性和特异性的阈值依赖性。
所有患者的研究区域均通过功能成像和 DECS 得出明确结果。在固定阈值分析中，运动组的敏感性和特异性分别为 100%和 68%，语言组的敏感性和特异性分别为 75%和 68%。获得数据的阈值依赖性分析显示，运动组的敏感性/特异性关系从阈值（p1）的 100%/0%，到（p2）的 100%/5%，（p3）的 74%/9%，到（p4）的 37%/36%。语言组的阈值依赖性敏感性和特异性评估结果为阈值（p1）的 78%/51%，（p2）的 67%/75%，（p3）的 50%/78%，以及（p4）的 33%/89%。
本研究关于 fMRI 数据阈值依赖性的结果表明，在 fMRI 评估中应获得个体化阈值。尽管大多数情况下结果令人满意，但 fMRI 显然不足以进行关键的术中决策。
