Three‐dimensional analysis of clustered voxels in 15O‐butanol brain activation images

Abstract Changes in the regional cerebral blood flow (rCBF) are often used as an indicator of changes in local neuronal activity of the brain. We present here quantitative measurements with positron emission tomography (PET) of rCBF with a freely diffusible flow tracer 15 O‐butanol in control and activation states, and a pixel‐by‐pixel statistical parametric analysis of the rCBF changes combined with a cluster analysis. Anatomically standardized rCBF activation data from 39 normal subjects were analyzed for the occurrence of clustered voxels. Noise data was obtained from repeat measurements of rCBF with the brain in the activated state and from simulations. The variance in test‐control images was largest outside the skull, and large in soft tissue regions around the brain. It was moderate but inhomogenous in gray matter, and low and homogenous in white matter. A special picture was generated of the conformation of sampled data with a normal distribution. In the gray and white matter, the pixel values were fount to conform to a normal distribution, permitting calculation of Student's t‐images. In the cluster analysis, activations are detected as clusters of voxels with high t‐values. The clusters in activated regions, however, were considerably larger than the full width half‐maximum of spatial autocorrelation function and were easily detected. Tables of the empirical Poisson‐like distributions of the number of clusters of different sizes are provided, from which P values of the significance of the occurrence of clusters of different sizes can be estimated. © 1993 Wiley‐Liss, Inc.