The superposition and registration of differing tomographic views is a difficult problem for investigators attempting to correlate brain form (structure), derived from x-ray computed tomography (CT) images, and brain function (metabolism), revealed by nuclear medicine positron emission tomography (PET) images. Our approach to this problem is based upon a three-stage strategy. First, we are developing practical methods for the accurate and reproducible placement of the head within a tomographic scanner's aperture. Second, we are developing techniques for monitoring head position during the image acquisition process, so that corrections may be made for head movements before the image is generated. Third, we are developing simplified algorithms for the scaling and registration on a digital display subsystem of digitized images from different scanners. Precise orientation of the subject's skull within the scanner's aperture is monitored and recorded through the use of a Polhemus position/orientation measurement subsystem connected to an IBM PC-XT, allowing simultaneous use of two independent sensors. The development of two inexpensive custom-molded oral appliances allows the Polhemus subsystem's sensor to be fixed to the subject's skull. A novel targeting algorithm was derived to provide visual cues related to head position within a scanner's imaging volume to the system operator. Two-sensor software was completed, and extensive evaluation has begun prior to its experimental use with test subjects. Future efforts will continue to center testing of the accuracy and repeatability of skull placement in tomographic scanners, as well as in refinement of algorithms for removing motion artifacts during the scanning process. The position/orientation measurement subsystem will also be interfaced to the Real-Time Gamma Camera Image Correction System, so that motion artifacts may be eliminated from the small-field-of-view gamma camera served by the system.
