Poline J B, Mazoyer B M
Service Hospitalier Frédéric Joliot, DRIPP-CEA, Orsay, France.
J Cereb Blood Flow Metab. 1993 May;13(3):425-37. doi: 10.1038/jcbfm.1993.57.
We present a new method for the analysis of individual brain positron emission tomography (PET) activation maps that looks for activated areas of a certain size rather than pixels with maximum values. High signal-to-noise-ratio pixel clusters (HSC) are identified and their sizes are statistically tested with respect to a Monte-Carlo-derived distribution of cluster sizes in pure noise images. From multiple HSC size tests, a strategy is proposed for control of the overall type I error. The sensitivity and specificity of this method have been assessed using realistic Monte Carlo simulations of brain activation maps. When compared with the gamma 2 statistic of the local maxima distribution, the proposed method showed enhanced sensitivity, particularly for signals of low magnitude and/or large size. Its potential for the individual analysis of PET activation studies is presented in two sets of subjects who underwent two cognitive protocols. Although it can be viewed as an alternative to the classical stereotactic averaging approach, this new method is intended to be a first step toward the analysis of single-subject PET activation studies.
我们提出了一种分析个体脑正电子发射断层扫描（PET）激活图的新方法，该方法寻找特定大小的激活区域，而非具有最大值的像素。识别出高信噪比像素簇（HSC），并根据纯噪声图像中通过蒙特卡罗方法得出的簇大小分布对其大小进行统计检验。通过多次HSC大小测试，提出了一种控制总体I型错误的策略。已使用逼真的脑激活图蒙特卡罗模拟评估了该方法的敏感性和特异性。与局部最大值分布的伽马2统计量相比，该方法显示出更高的敏感性，尤其是对于低强度和/或大尺寸的信号。在两组接受两种认知方案的受试者中展示了其在PET激活研究个体分析中的潜力。尽管该方法可被视为经典立体定向平均法的替代方法，但这种新方法旨在成为单受试者PET激活研究分析的第一步。
