AN ASSESSMENT OF 2 METHODS FOR GENERATING AUTOMATIC REGIONS OF INTEREST

Two fully automatic methods for generating regions of interest (ROIs) for nuclear medicine images are described and assessed. One of these, involving registration of a previously defined ROI onto a new image, u ses spatial information and is appropriate for two- and three-dimensio nal images which may be static or dynamic. The other method is based o n artificial neural networks and uses temporal information. It is appr opriate for dynamic images only. The registration method has been test ed using 10 pairs of stress and redistribution images obtained from ca rdiac perfusion SPET Regions of interest of the left ventricular muscl e, defined on the stress images, were registered onto the redistributi on images, where they were compared with reproducibility of manually d rawn ROIs. Both methods were tested on 17 Tc-99(m)-MAG3 kidney dynamic studies, where the original ROIs corresponding to both kidneys and th e bladder were defined using the COST B2 hybrid phantom. Our results i ndicate that neither method is as reliable as having ROIs redrawn by t he operator, although there are indications that an artificial neural network which combines the use of the spatial and temporal information could prove useful for dynamic studies. ((C) 1998 Lippincott Williams & Wilkins).