Preliminary Study of an Intra-operative PET Imaging Probe System.

Abstract

PET imaging has gained widespread acceptance in cancer imaging because Positron Emission Tomography can identify physiological changes due to cancer. Nevertheless conventional PET imaging has difficulty detecting tumors less than 1cm in diameter in clinical use due mainly to background radiation, statistical noise, resolution loss due to lack of depth interaction resolution in detectors, and annihilation photon acolinearity. Conventionally if detected tumors are surgically removable, surgeons locate and remove the tumors during surgery based on the preoperative scans. One of the drawbacks of relying solely on preoperative imaging is that tumor locations could be displaced during surgery due to patient’s movement. In this dissertation, a preliminary study of an intra-operative PET imaging probe system is presented. The proposed probe system consists of a low resolution partial ring detector and a high resolution imaging probe that is equipped with a position tracker. The high resolution probe operates in coincidence with the partial ring detector. The high resolution imaging probe and its proximity to target lesions contribute to the localization of small tumors. In addition, the probe system can be used to detect occult tumors. The ultimate goal is to provide incremental 3-dimensional reconstructed images that are re-projected in real time onto a plane whose orientation is driven by the tracking device. A prototype of the PET imaging probe system was built to test the feasibility of the intra-operative PET imaging probe system. Coincidence detection efficiency of about 0.00012% was observed. A variant of 3-dimensional one-pass list-mode maximum likelihood algorithm (OP-LML) was developed to reconstruct images from the measured data. A row-action maximum likelihood algorithm was integrated with the OP-LML. To speed up image reconstruction by a factor of 30-40, the proposed algorithm was parallelized and was run on a graphics processing unit. Reconstructed images from simulated data with no intrinsic blurring showed resolution of 1.0mm~1.5mm FWHM. However as we expected, reconstructed images from the experimental set-up with limitations failed to separate two Na-22 point sources 1.5mm apart. Experimental resolutions of 4mm FWHM in the longitudinal direction and 2mm FWHM in the transverse direction were obtained for the two point sources.