Statistical performance evaluation, system modeling, distributed computation and signal pattern matching for a Compton medical imaging system

Radionuclide cancer therapy requires imaging radiotracers that concentrate in tumors and emit high energy charged particles that kill tumor cells. These tracers, such as 131I, generally emit high energy gamma photons that need to be imaged to estimate tumor dose and changes in size during treatment.;This research describes the performance of a dual-planar silicon-based Compton imaging system and compares it to a conventional parallel-hole collimated Anger camera with high energy general purpose lead collimator for imaging higher energy photons emitted from 131I.;The collimated Anger camera imposes a tradeoff between resolution and sensitivity because of the physical constraints resulting from the mechanical collimation. As the gamma-ray photons exceed 364keV, increased septal penetration and scattering further degrade the imaging performance. Simulations of the Anger camera and the Compton imaging system demonstrate a 20-fold advantage in detection efficiency and higher spatial resolution for detecting high energy photons by the Compton camera since it decouples the resolution-sensitivity tradeoff.;The system performance and comparison are also analyzed using the modified uniform Cramer-Rao bound algorithms we developed and verified along with the Monte Carlo calculations and system modeling. The bound calculations show that the effect of Doppler broadening is the limiting factor for Compton camera performance for imaging 364keV photons. Performance of the two systems was compared and analyzed by simulating a 2D disk with uniform activities. For the case in which the two imaging systems detected the same number of events, the proposed Compton imaging system has lower image variance than the Anger camera with HEGP when the FWHM of the desired point source response is less than 1.2 cm. This advantage was also demonstrated by imaging and reconstructing a 2D hot spot phantom.;In addition to the performance analysis, the distributed and parallel Maximum Likelihood Maximization Expectation algorithm with chessboard data partition was evaluated for speeding up image reconstruction for the Compton imaging system. A 1 x 64 distributed computing system speeded computation by about a factor of 22 compared to a single processor. Finally, a real time signal processing and pattern matching system employing state-of-the-art digital electronics is described for solving problems of event pile-up raised by high photon count rate in the second detector.