An automatic region of interest technique for the evaluation of PET brain section images

Transverse section images obtained from F-18 FDG cerebral metabolic studies exhibit an approximate bilateral symmetry. One way of evaluating region function is to compare glucose utilization in one region of a parametric image with a corresponding region on the contralateral side. Manual positioning of ROI markers is tedious and is subject to a large error from operator bias. To overcome these problems, the authors have developed an automatic method for constructing an objective ROI map which allows comparisons between structures on the right and left sides. The first part of the method is the generation of an accurate mask of gray matter structure. This is obtained by subtracting a heavily smoothed image from the contrast enhanced image. The mask is formed by dividing this resultant image into itself. It is then made symmetric by rotating it about a line passing AP through the image centroid and adding it to the unrotated mask. The mask is sectioned into an ROI map which is overlayed on the original data