Zuendorf Gerhard, Kerrouche Nacer, Herholz Karl, Baron Jean-Claude
Neurological Clinic and Max-Planck-Institute for Neurological Research, University Cologne, Gleueler Strasse 50, 50931 Cologne, Germany.
Hum Brain Mapp. 2003 Jan;18(1):13-21. doi: 10.1002/hbm.10069.
Principal component analysis (PCA) is a well-known technique for reduction of dimensionality of functional imaging data. PCA can be looked at as the projection of the original images onto a new orthogonal coordinate system with lower dimensions. The new axes explain the variance in the images in decreasing order of importance, showing correlations between brain regions. We used an efficient, stable and analytical method to work out the PCA of Positron Emission Tomography (PET) images of 74 normal subjects using [(18)F]fluoro-2-deoxy-D-glucose (FDG) as a tracer. Principal components (PCs) and their relation to age effects were investigated. Correlations between the projections of the images on the new axes and the age of the subjects were carried out. The first two PCs could be identified as being the only PCs significantly correlated to age. The first principal component, which explained 10% of the data set variance, was reduced only in subjects of age 55 or older and was related to loss of signal in and adjacent to ventricles and basal cisterns, reflecting expected age-related brain atrophy with enlarging CSF spaces. The second principal component, which accounted for 8% of the total variance, had high loadings from prefrontal, posterior parietal and posterior cingulate cortices and showed the strongest correlation with age (r = -0.56), entirely consistent with previously documented age-related declines in brain glucose utilization. Thus, our method showed that the effect of aging on brain metabolism has at least two independent dimensions. This method should have widespread applications in multivariate analysis of brain functional images.
主成分分析（PCA）是一种用于降低功能成像数据维度的知名技术。PCA可以看作是将原始图像投影到一个维度更低的新正交坐标系上。新轴按重要性递减顺序解释图像中的方差，显示脑区之间的相关性。我们使用一种高效、稳定且解析的方法，以[（18）F]氟-2-脱氧-D-葡萄糖（FDG）作为示踪剂，计算了74名正常受试者的正电子发射断层扫描（PET）图像的PCA。研究了主成分（PCs）及其与年龄效应的关系。对图像在新轴上的投影与受试者年龄之间进行了相关性分析。前两个PCs可被确定为仅与年龄显著相关的PCs。解释数据集方差10%的第一主成分仅在55岁及以上的受试者中降低，且与脑室和基底池内及相邻区域的信号丢失有关，反映了预期的与年龄相关的脑萎缩以及脑脊液空间扩大。占总方差8%的第二主成分在前额叶、顶叶后部和扣带回后部皮质有高负荷，并显示出与年龄的最强相关性（r = -0.56），这与先前记录的与年龄相关的脑葡萄糖利用下降完全一致。因此，我们的方法表明衰老对脑代谢的影响至少有两个独立维度。该方法应在脑功能图像的多变量分析中有广泛应用。
