Fig. 1: ROI-based vs Random Field-based Spatial Inference.
From: A framework for focal and connectomic mapping of transiently disrupted brain function
a ROI-based spatial inference. Simulated stimulation points resulting in two hypothetical deficits—A (blue) and B (red)—are counted across pre-defined ROIs, here represented in 2D space for simplicity. A statistical test is then performed on the counts to infer the spatial distribution of function in terms determined by the ROI boundary. b Parcellation-induced mislocalisation. When there is a poor correspondence between the parcellation and the underlying functional substrate, the inference either fails completely (left) or is distorted (middle and right). c Random field-based spatial inference. Here each stimulation point is convolved with a predefined Gaussian kernel, so that each ___location is now supported across the entire ___domain, enabling the application of voxel-wise inference on a regular grid. A statistical test is then performed at each voxel to retrieve the substrates associated with the observed deficits A and B. The colourmap is the negative decimal log of each p-value. Thresholding following multiple comparisons correction is not shown here for simplicity.
