Author ORCID Identifier
0000-0001-9915-7443
Date of Graduation
12-2018
Document Type
Dissertation (PhD)
Program Affiliation
Medical Physics
Degree Name
Doctor of Philosophy (PhD)
Advisor/Committee Chair
John D Hazle
Committee Member
James Bankson
Committee Member
Robert Bast Jr
Committee Member
David Fuentes
Committee Member
Konstantin Sokolov
Abstract
Superparamagnetic relaxometry (SPMR) is an emerging technology that leverages the unique properties of biologically targeted superparamagnetic iron oxide nanoparticles to detect cancer. The use of ultra-sensitive sensors enables SPMR to detect tumors ten times smaller than current imaging methods. Reconstructing the distribution of cancer-bound nanoparticles from SPMR measurements is challenging because the inverse problem is ill posed. Current methods of source reconstruction rely on prior knowledge of the number of clusters of bound nanoparticles and their approximate locations, which is not known in clinical applications. In this work, we present a novel reconstruction algorithm based on compressed sensing methods that relies on only clinically feasible information. This approach is based on the hypothesis that the true distribution of cancer-bound nanoparticles consists of only a few highly-focal clusters around tumors and metastases, and is therefore the sparsest of all possible distributions with a similar SPMR signal. We tested this hypothesis through three specific aims. First, we calibrated the sensor locations used in the forward model to measured data, and found a 5% agreement between the forward model and the data. Next, we determined the optimal choice of the data fidelity parameter and investigated the effect of experimental factors on the reconstruction. Finally, we compared the compressed sensing-based algorithm with the current reconstruction method on SPMR measurements of phantoms. We found that when a multiple sources were reconstructed simultaneously, the compressed sensing approach was more frequently able to detect the second source. In a blinded user analysis, our compressed sensing-based reconstruction algorithm was able to correctly classify 80% of the test cases, whereas the current reconstruction method had an accuracy of 43%. Therefore, our algorithm has the potential to detect early stage tumors with higher accuracy, advancing the translation of SPMR as a clinical tool for early detection of cancer.
Keywords
superparamagnetic relaxometry, nanoparticles, early detection of cancer, compressed sensing, convex optimization