Ishii K, Willoch F, Minoshima S, Drzezga A, Ficaro E P, Cross D J, Kuhl D E, Schwaiger M
Nuklearmedizinische Klinik und Poliklinik, Klinikum rechts der Isar, Technische Universität München, München, Germany.
J Nucl Med. 2001 Apr;42(4):548-57.
Despite the increased use of statistical mapping to detect brain functional changes in Alzheimer's disease (AD), potential artifacts introduced by stereotactic anatomic standardization of atrophied brains have not been examined carefully. We investigated the effects of anatomic standardization by Statistical Parametric Mapping (SPM) and NEUROSTAT.
First, 10 AD patients and 10 age-matched healthy volunteers underwent 18F-FDG brain PET imaging. Each image set was standardized to a stereotactic brain template using SPM or NEUROSTAT, followed by pixel normalization to the global or cerebellar activity. Within-group comparisons of standardized image sets by each method and a between-group comparison of healthy volunteers and AD patients were performed using the statistical analysis routines of SPM. Second, simulated PET image sets were generated from segmented MR image sets of 5 healthy volunteers and 5 AD patients. Using the anatomic standardization parameters estimated on the simulated image sets, original gray matter MR image sets were transformed to the stereotactic coordinate system. Between-group subtraction analyses of the transformed gray matter image sets between healthy volunteers and AD groups were performed to examine the accuracy of cortical gray matter matching.
Between-group comparison by SPM or NEUROSTAT showed generally similar areas of hypometabolism in bilateral temporoparietal, posterior cingulate, and left frontal cortices. Both methods showed possible deformation artifacts in the anterior part of the corpus callosum. The localization of the peak hypometabolism varied considerably between the two methods when global normalization was applied. The use of a common brain template for standardization resulted in asymmetric differences in cortical margins, indicating systematic differences in the deformation algorithms. The realistic simulation study revealed gray matter mismatches to be 20% greater with SPM than with NEUROSTAT.
Although different statistical mapping methods may yield grossly similar patterns of hypometabolism in AD, the extent, severity, and peak location of metabolic changes can be inconsistent. Deformation accuracy appears to be more prone to atrophy. These limitations need to be considered carefully in the application and interpretation of brain mapping analysis in atrophied brains.
尽管越来越多地使用统计映射来检测阿尔茨海默病（AD）中的脑功能变化，但萎缩脑的立体定向解剖标准化所引入的潜在伪影尚未得到仔细研究。我们研究了统计参数映射（SPM）和NEUROSTAT进行解剖标准化的效果。
首先，10名AD患者和10名年龄匹配的健康志愿者接受了18F-FDG脑PET成像。每个图像集使用SPM或NEUROSTAT标准化为立体定向脑模板，然后对像素进行全局或小脑活动归一化。使用SPM的统计分析程序对每种方法标准化后的图像集进行组内比较，并对健康志愿者和AD患者进行组间比较。其次，从5名健康志愿者和5名AD患者的分割MR图像集中生成模拟PET图像集。使用在模拟图像集上估计的解剖标准化参数，将原始灰质MR图像集转换到立体定向坐标系。对健康志愿者和AD组之间转换后的灰质图像集进行组间减法分析，以检查皮质灰质匹配的准确性。
通过SPM或NEUROSTAT进行的组间比较显示，双侧颞顶叶、后扣带回和左额叶皮质的代谢减低区域总体相似。两种方法均显示胼胝体前部可能存在变形伪影。当应用全局归一化时，两种方法之间代谢减低峰值的定位差异很大。使用共同的脑模板进行标准化导致皮质边缘出现不对称差异，表明变形算法存在系统差异。实际模拟研究显示，与NEUROSTAT相比，SPM导致的灰质不匹配高20%。
尽管不同的统计映射方法在AD中可能产生大致相似的代谢减低模式，但代谢变化的程度、严重程度和峰值位置可能不一致。变形准确性似乎更容易受到萎缩的影响。在萎缩脑的脑图谱分析的应用和解释中，需要仔细考虑这些局限性。
