Date of Award
Summer 8-15-2022
Degree Name
Doctor of Philosophy (PhD)
Degree Type
Dissertation
Abstract
Proton therapy can achieve better sparing of normal tissues than the conventional photon radiation therapy due to proton’s Bragg Peak property. However, to unlock the full potential of protons, accurate prediction of in vivo proton stopping power ratio (SPR) is required for proton therapy treatment planning. The current standard practice is to map SPR from Hounsfield Unit (HU) values of a single-energy computed tomography (SECT) scan through a stoichiometric calibration technique. This technique is subjected to a variety of factors that congregate on the uncertainties in SPR estimation, including the calibration uncertainty (up to 0.5% to 1.8% of the total proton beam range), SECT uncertainty (beam hardening, reconstruction artifacts, etc.), and patient positioning uncertainty (misalignment, motions, and anatomical changes). Two emerging techniques have been proposed to improve proton SRP estimation accuracy in proton therapy: dual-energy computed tomography (DECT) and proton computed tomography (pCT). The former attempts to achieve better material differentiation than SECT by scanning the patient at two different photon energies. The latter aims to avoid sources of uncertainties in HU-to-SPR conversion by using protons directly as the imaging particle. A previously proved highly accurate DECT-based SPR mapping technique using a joint statistical image reconstruction method with a linear basis vector model (JSIR-BVM) was integrated with a clinical Monte Carlo-based treatment planning system (TPS) for dose prediction comparison with the standard stoichiometric SECT method. Percentage deviation from the ground-truth volume receiving 80% of the prescription dose within a 5 mm distal-ring region around the planning target volume was 2.6% for JSIR-BVM and 6.8% for SECT in the simulated case, showing a nontrivial risk of underdosing to the tumor region if planned with SECT. For the clinical head-and-neck cancer patient case, the percentage difference between JSIR-BVM and SECT in the mean dose and the volume receiving 80% of the prescription dose in a similarly defined ROI was 2.35% and 13.86%, respectively. The results demonstrate that our JSIR-BVM method provides more accurate and less variable mass-density maps than SECT for a simulation case with known ground truth, resulting in noticeable improvements in dose-calculation accuracy. Hence, this work constitutes an important transitional step towards realizing the clinical benefits of more accurate imaging of radiological quantities by JSIR-BVM. The clinical impact of the DECT-based JSIR-BVM SPR mapping technique was evaluated based on dose-volume histograms (DVHs), the mean dose in clinical target volume (CTV), and maximum dose within serial organs at risk (OARs). No recalculated DVH metric differed by more than 0.37% in 2 of the 3 cases. However, in the third case with the brainstem overlapping the CTV, when recalculated on the DECT SPR map, the mean dose to the CTV and the maximum dose in the brainstem increased from 54 Gy to 56 Gy and 55.1 Gy to 57.7 Gy, respectively, indicating a nontrivial risk in treatment toxicity associated with inaccurate prediction of proton beam range. The results validate that a methodology for evaluating the clinical impact of highly accurate DECT SPR maps has been developed. The differences between SECT and DECT dose distributions were clinically meaningful in one of the three evaluated patient cases. On the other hand, a novel pCT system has been proposed and developed as discussed in this dissertation. We first demonstrated the clinical feasibility through Monte Carlo simulation, then expanded the generality and compatibility of this technology for various beam characteristics with a model-based reconstruction explicitly developed for the system. The prototype of the pCT detector is composed of two strip ionization chambers measuring locations and lateral profiles of the exiting beam and a multi-layer ionization chamber (MLIC) measuring the integral depth doses (IDDs), which can be translated to residual energies of the exiting proton beams. A collimator with a round slit of 1 mm in diameter was placed in the central beam axis upstream from steering magnets to collimate the spot size down to 1 mm. The maximum deviation in reconstructed proton SPR from the ground truths was reported to be 1.02% in one of the 13 inserts when the number of protons per beamlet passing through the slit dropped to 103. The imaging dose was correlated linearly to incident protons and was determined to be 0.94 cGy if 103 protons per beamlet were used. Imaging quality was acceptable for planning purposes and held consistently through all levels of imaging dose. Spatial resolution was measured as 5 lp/cm in all simulations, varying imaging dose. The results prove the clinical feasibility of the pCT system with an imaging dose lower than kV cone-beam computed tomography (CBCT), making it potentially an excellent tool for localization and plan adaption in proton therapy. A reconstruction approach was developed to eliminate the use of a collimator by modeling the IDD of an uncollimated proton beam as a weighted sum of percentage depth doses (PDDs) of constituent narrow beamlets separated by 1mm. The beamlets' water equivalent path lengths (WEPLs) were determined by iteratively minimizing the squared L2-norm of the forward projected and simulated IDDs. The final WEPL values were reconstructed into pCT images, i.e., proton SPR maps, through simultaneous algebraic reconstruction technique with total variation regularization (SART-TV). When the proposed reconstruction approach was applied, the percentage deviations from reference SPR were within ±1% in all selected ROIs. The mean absolute error of the reconstructed SPR was 0.33%, 0.19%, and 0.27% for the cylindrical phantom, and the adult phantom at the head and lung region, respectively. The frequency at 10% of the modulation transfer function (MTF) was 6.38 cm-1. The mean signal-to-noise ratio (SNR) of all the inserts was 2.45. The mean imaging dose was 0.29 cGy and 0.25 cGy at the head and lung region of the adult phantom, respectively. The results suggest that with the proposed reconstruction approach, the pCT system can achieve similar SPR accuracy and spatial resolution as the pCT system with an additional collimator while avoiding the potential side effects caused by extra neutron dose generated by collimating proton beams. Finally, the possibility of using the pCT system to extract proton scattering information was explored. Two forward models of predicting integrated transverse dose distribution of the exiting proton beam were implemented and compared. Moreover, the differential Molière model was utilized to reconstruct the scattering length of the imaging object. The scattering length map achieved 0.83% mean absolute deviation from the reference values when reconstructed through a modified simultaneous algebraic reconstruction technique (SART) algorithm and can be used as a correction for SPR estimation or to provide additional information in proton treatment planning. In summary, an evaluation study of dose prediction and clinical impact of the DECT-based JSIR-BVM SPR mapping technique was conducted. The transition of this highly accurate technique toward clinical application was established. Furthermore, a novel pCT system incorporated with a PBS facility and detected with an MLIC detector was proposed and developed. The feasibility of the system was proved through Monte Carlo simulation. Moreover, a reconstruction approach modeling the IDDs of the exiting proton beam was developed to further improve the system design by eliminating the additional hardware that may cause extra neutron dose and unnecessary quality assurance. Finally, proton scattering information was reconstructed using simulated data based on the pCT design, which can further improve SPR accuracy or provide additional patient anatomic information for proton treatment planning.
Language
English (en)
Chair
Tianyu Zhao
Committee Members
Hong Chen, Tianyu Zhao, Tiezhi Zhang, Deshan Yang,