Voxelwise Bayesian lesion-deficit analysis

Relating cognitive deficits to the presence of lesions has been an important means of delineating structure-function associations in the human brain. We propose a voxel-based Bayesian method for lesion-deficit analysis, which identifies complex linear or nonlinear associations among brain-lesion locations, and neurological status. We validated this method using a simulated data set, and we applied this algorithm to data obtained from an acute-stroke study to identify associations among voxels with infarct or hypoperfusion, and impaired word reading. We found that a distributed region involving Brodmann areas (BA) 22, 37, 39, and 40 was implicated in word reading.