AUTOMATIC DETECTION OF HYPOPERFUSED AREAS IN SPECT BRAIN-SCANS

We describe a method for automatic identification of areas with perfus ion changes in SPECT brain images. An intersubject registration techni que is used to stereotactically register images from a selected contro l group allowing for reference images to be created by averaging the s ubjects image data. An individual SPECT brain scan can be brought into registration with the reference image and comparison to the normal re ference group can be made by subtracting the two volumes. Furthermore, since the variance in the reference group is known, a z-score image o r an image coded in standard deviations, can be computed. The SPECT re ference volume is defined in the same coordinate system as a brain atl as, and anatomical labeling of areas of interest is possible. We show results from the creation of an average image based on ii individuals and from its comparison with pathological cases.