Date of Award
Doctor of Philosophy (PhD)
Positron emission tomography (PET) is a highly sensitive imaging modality that can provide in vivo quantitative information of biological processes at a biochemical level. The majority of PET systems have a ring geometry where the detectors fully encircle an axial section of a patient to provide whole-body imaging capability using multiple bed positions. However, these clinical whole-body PET systems are limited in terms of both image resolution and system sensitivity. To achieve a high resolution, customized geometries which trade global utility for application specific performance (breast, small animals etc.) imaging probes may be utilized. In some cases, an application-specific PET system can also significantly improve the system sensitivity, while in other cases new imaging capabilities are the main focus of a system design. Since these systems may have non-cylindrical geometries, they require a generalized reconstruction framework which is system geometry independent.To support the implementation of such PET systems with unconventional geometries, we have developed a fully 3D list-mode GPU based image reconstruction framework using maximum-likelihood expectation-maximization (ML-EM) algorithm. We have implemented correction procedures like normalization, attenuation correction and scatter correction for this image reconstruction framework to support arbitrary system geometry and interactive imaging using the PET/US system. We compare the performance of our framework with a clinical reconstruction framework. For all the systems investigated in our work, we have used this reconstruction framework. We also modified the framework to support continuous patient bed motion acquisition. Appropriate correction techniques (attenuation, scatter, normalization) for continuous bed motion need to be explored in future. We propose a high-performance dedicated breast PET imaging device concept which will scan both breasts simultaneously and have high sensitivity and resolution; fine temporal resolution; complete visualization of both breasts, mediastinum and axilla; and a modular design that can be inserted into a body MR scanner to provide dual-modality imaging (PET and MRI) allowing for MRI guided biopsy access. The volumetric geometry is illustrated along with GATE Monte Carlo simulations of various sized lesions (4-6mm) of differing lesion: background ratios (6:1 to 4:1) mimicking different biological uptake. We finally compare this dedicated breast PET imager to a state-of-the-art clinical PET/CT scanner. In another direction, we propose a novel PET/Ultrasound imaging system that will bring both these modalities to patient bedside to support point-of-care applications with near real-time interactive imaging capability. It consists of a movable hybrid imaging probe consisting of PET detector arrays and an ultrasound transducer. The PET detectors are in coincidence with a detector array behind a patient. The movable detectors make it possible for the operator to control the scanning trajectory freely to achieve optimal coverage and sensitivity for patient specific imaging tasks. The initial application is to detect vulnerable plaque in carotid artery using [64Cu]DOTA-ECL1i, a PET tracer targeting C-C chemokine receptor type 2 (CCR2) developed at Washington University which may identify high-risk plaques with inflammation. We present Monte Carlo simulation studies to test the feasibility of the system. We further optimize the system design, develop time-of-flight (TOF) PET detectors and develop a proof-of-concept POC PET/Ultrasound prototype. We perform phantom studies using this prototype and demonstrate the feasibility of this system to provide adequate anatomic, functional and molecular imaging information for carotid artery imaging. We also investigate another application of the POC PET technology to image lungs. Since, this is a stand-alone system targeted for imaging larger patient volumes, non-uniform tissue attenuation is a challenge. We implement a transmission-scan based attenuation correction and perform simulations to test this correction method with a known analytical µ-map based correction. We also report sensitivity and reconstructed phantom images as well as demonstrate an interactive scanning strategy. Finally, we discuss additional works needed to extend this list-mode reconstruction framework to support continuous-bed-motion acquisition in the future.
Dr. Joseph O'Sullivan
Dr. Yuan-Chuan Tai
Available for download on Thursday, April 27, 2023