Abstract
Functional neuroimaging can be used to characterize two types of abnormality in patients with neurological deficits: abnormal functional segregation and abnormal functional integration. In this paper we consider the factors that influence the experimental design, analysis, and interpretation of such studies. With respect to experimental design, we emphasize that: 1) task selection is constrained to tasks the patient is able to perform correctly, and 2) the most sensitive designs entail presenting stimuli of the same type close together. In terms of data preprocessing, prior to statistical analysis, we note that structural pathology may call for constraints on nonlinear transformations, used by spatial normalization, to prevent distortion of intact tissue. This means that one may have to increase spatial smoothing to reduce the impact of inaccurate normalization. Important issues in statistical modeling concern the first level of analysis (estimation of activation within subject), which has to distinguish correct from incorrect responses. At the second level (between subjects), inference should be based on betweenβsubjects variance. Provided that these and other constraints are met, deficits in functional segregation are indicated when activation in one or a set of regions is higher or lower in patients relative to control subjects. In contrast, deficits in functional integration are implied when the influence of one brain region on another is stronger or weaker in patients relative to control subjects. J. Magn. Reson. Imaging 2006. Β© 2006 WileyβLiss, Inc.