Spatial low frequency pattern analysis in positron emission tomography: a study between normals and schizophrenics.
Using the two-dimensional Fourier transform and the brain's centroidal principal axis, a method is developed for the analysis of PET metabolic brain images without the use of predefined anatomic regions of interest. We applied the method to images from a group of 11 normal and 12 medicated schizophrenics tested under resting conditions and under a visual task. A cortical/subcortical spatial pattern was found to be significant in two directions; anterior/posterior and chiasmatic (left-anterior/right-posterior). The best individual clinical classification (Jackknife classification) occurred under visual task at two axial brain levels: at the basal ganglia with correct classification rates of 91% and 84%, while the cerebellum had rates of 82% and 92%. These high classification rates were obtained using only the four coefficients of the lowest spatial frequency. These results point to the generalized brain dysfunction of regional glucose metabolism in chronic medicated schizophrenics both at rest and at a visual image-tracking task.