Luke Polson
Doctor of Philosophy in Medical Physics (PhD)
Research Topic
Enhancement and quantification in nuclear medicine imaging using artificial intelligence
Theses completed in 2010 or later are listed below. Please note that there is a 6-12 month delay to add the latest theses.
In this thesis, we developed a computational framework using physiologically based pharmacokinetic (PBPK) models to study radiopharmaceutical kinetics in the body. To handle the complexity of PBPK models (over 100 differential equations), a scalable modeling notation called the "reaction graph" is introduced. Implemented in Matlab's Simbiology module and shared in the systems biology markup language (SBML) format, this notation enables easy model reproduction. Referred to as the physiologically based radiopharmacokinetic model (PBRPK), it is fine-tuned specifically for radiopharmaceuticals.Using the PBRPK model and literature-based parameters, we addressed three key questions in radiopharmaceutical therapy. First, we systematically studied the interaction between hot and cold species and its impact on absorbed dose in tumors and organs at risk (OARs) and we found out that tumor receptor density and volume influence the degree of competition and the average absorbed dose in tumors and OARs in a way that lower receptor density (or lower tumor volume) leads to more dominant competition between hot and cold species that is the absorbed dose by tumor highly depends on the number of cold ligands injected to the patient.Next, the effect of injecting radiopharmaceuticals in smaller portions and with specific timing is investigated. We found that bolus injection leads to non-efficient receptor binding and sub-optimal delivered doses to the tumor and organs at risk. However, we found that injecting in several smaller portions leads to a higher absorbed dose by tumors and organs at risk. So multi-bolus injection can be thought of as a tool that increases the delivered dose to the tumor and OAR without any differential effect (delivering mode dose to the tumor and sparing OAR)Finally, the PBRPK model is used to simulate the impact of albumin affinity on radiopharmaceutical kinetics. Varying albumin affinity of the radiopharmaceuticals reveals changes in blood residence time, with lower dissociation rates resulting in longer residence times. Generally, lower dissociation rates increase tumor absorbed dose and decrease OAR dose, which makes the albumin binding a potential tool to achieve differential enhancement in targeting tumors and sparing OAR.
View record
The full abstract for this thesis is available in the body of the thesis, and will be available when the embargo expires.
View record
For whole-body (WB) kinetic modeling based on a typical Positron Emission Tomography (PET) scanner, a multi-pass multi-bed scanning protocol is necessary because of the limited axial field of view. Such a protocol introduces loss of early dynamics in the time-activity curves (TACs) and sparsity in TAC measurements; thus inducing uncertainty in parameter estimation when using least squares estimation (LSE) (i.e., common standard), especially for kinetic micro-parameters. To address the issue above, this thesis proposes a method that can estimate micro-parameters by building a reference TAC database, and selecting optimal parameters based on analysis of multiple aspects of the TACs, while in our assessment of performance compared to conventional methods we focus on general image qualities, overall visibility, and tumor detectability.To achieve the research goal above, we developed a novel parameter estimation method called parameter combination-driven estimation (PCDE), which has two distinctive characteristics:1) improved capability of finding a correct correlation between early and late TACs at the cost of the resolution of the estimated parameter, and 2) exploitation of multiple aspects of TAC. To compare the general image quality between the two methods, we plotted tradeoff curves for the normalized bias (NBias) and the normalized standard deviation (NSD). We also evaluated the impact of different iteration numbers of the ordered-subset expectation maximization (OSEM) reconstruction algorithm on the tradeoff curves. In addition, for overall visibility, which is a measure of the capability of identifying suspicious lesions in WB (i.e., global inspection), the overall signal-to-noise ratio (SNR) and spatial noise (NSDspatial) were calculated and compared. Furthermore, the contrast-to-noise ratio (CNR) and relative error of the tumor-to-background ratio (RETBR) were calculated to compare tumor detectability within a specific organ (i.e., local inspection). Through the proposed method, the improved general image quality, overall visibility, and tumor detectability were verified in micro-parametric images with OSEM reconstructions. We expect our work to contribute to opening the door to use of a typical PET scanner to reliably estimate kinetic micro-parameters in WB imaging, which has been so far very challenging owing to significant uncertainties in estimates when using LSE methods.
View record
If this is your researcher profile you can log in to the Faculty & Staff portal to update your details and provide recruitment preferences.