Chapter 36 - Measuring Activation Pattern Reproducibility Using Resampling Techniques

Abstract

The chapter focuses that the central problem in brain activation experiments is the reproducibility of the resulting functional activation patterns, if the experiments were to be repeated with an independent group of subjects. This is an important issue that is influenced by data analysis model biases, resulting from normalization procedures, definitions of noise, and the setting of activation thresholds, and by the small sample sizes that are typically analyzed. Earlier work examined the reproducibility of suprathreshold, spatially localized activation foci. To address activation pattern reproducibility while avoiding model dependent noise definitions and thresholding biases, the use of scatter plots has been proposed. The chapter explains that with these plots, signal-level reproducibility is visualized in all pairs of Talairach-aligned voxels for pairs of activation images from independent groups of subjects.