Development and Application of Simulation Tools for Virtual Imaging Trials in Computed Tomography

Limited Access
This item is unavailable until:
2024-05-26

Date

2022

Journal Title

Journal ISSN

Volume Title

Abstract

Virtual imaging trials (VITs) utilize anatomically accurate computational models of human subjects and validated imaging simulators for evaluating and optimizing existing and upcoming clinical imaging technologies. By eliminating the logistics of organizing clinical trials, providing full access to ground-truth information, and mitigating ethical considerations relating to exposure of real human subjects to ionizing radiation, VITs significantly improve upon the time- and cost- effectiveness of traditional trials. For VITs in computed tomography (CT), hybrid simulations approaches combining ray-tracing for simulating the primary and Monte Carlo (MC) simulations for the simulating the scattered component of the signal with radiation dose have proven effective for simulating realistic images. Although an effective methodology for simulating images, it requires obtaining realistic scanner-specific estimates of radiation dose and projection-specific scatter in a time-efficient manner, which existing solutions fail to provide.In awareness of these limitations, the purpose of this Ph.D. dissertation was to develop, validate, and apply simulation tools for enabling hybrid imaging simulations for VITs in CT. The research introduced in this dissertation included the following parts: (1) development and validation of a GPU-accelerated Monte Carlo tool for rapid dose and scatter estimation in conventional single source energy-integrating CT, (2) extension of the GPU-accelerated MC tool to model advanced CT systems – dual source CT and photon-counting CT, and (3) demonstrating the utility of the developed simulation tools for specific VIT applications. The dissertation also demonstrated the potential of these tools developed specifically for CT imaging to enable VITs for other x-ray-based imaging modalities. In part 1, a GPU-accelerated Monte Carlo (MC) tool for rapid estimation of dose and scatter estimation in energy-integrating CT was developed. The dose estimates from the tool were validated against measurements of absorbed dose with thermoluminescent dosimeters (TLDs) in anthropomorphic phantoms while the scatter estimates were validated against scatter-to-primary ratios measured using the single-blocker method implemented with a lead cylinder and a CTDI phantom. The developed MC tool combined with kernel- and CNN-based denoising methods was integrated with a ray-tracing imaging simulator to establish a framework for simulating realistic scanner-specific images and organ doses in energy-integrating CT. In part 2, the GPU-accelerated MC tool was extended to model advanced CT systems such as dual-source and photon-counting CT. The modeling of dual-source CT involved the incorporation of a second source-detector pair enabling estimation of cross-scatter. For modeling the physics of image acquisition in photon-counting CT (PCCT), a modular detector response model accounting for the physics of signal generation and effects of non-idealities such as x-ray crosstalk, charge sharing, and pulse pileup in a variety of CdTe- and Si-based photon-counting detectors was developed and validated. The detector response model was integrated with the imaging framework developed in part 1 using spatio-energetic mean and covariance matrices for computing the mean signal and noise in PCCT images. In part 3, the simulation tools developed in parts 1 and 2 were applied to specific VIT applications. The first application involved the assessment of task-specific benefits from improved spatial resolution of a prototype silicon-based PCCT system in context of perceptual benefits across a variety of anatomies (lung, liver, head-and-neck, and inner auditory canal) and radiomics estimation for lung lesions. The second application involved evaluating the impact of low kV imaging on skin dose in contrast-enhanced CT. In addition to CT, the simulation tools developed in this framework were also extended to digital tomosynthesis, where a comprehensive database of organ dose coefficients for adult and pediatric patients across multiple exam protocols was developed for prospective and retrospective clinical organ dosimetry. In conclusion, the research introduced and presented in this dissertation successfully accomplished the development, validation, and application of simulation tools for hybrid imaging simulations for VITs in CT while also demonstrating the potential of the tools developed for enabling VITs in other x-ray-based imaging modalities.

Department

Description

Provenance

Citation

Citation

Sharma, Shobhit (2022). Development and Application of Simulation Tools for Virtual Imaging Trials in Computed Tomography. Dissertation, Duke University. Retrieved from https://hdl.handle.net/10161/25154.

Collections


Dukes student scholarship is made available to the public using a Creative Commons Attribution / Non-commercial / No derivative (CC-BY-NC-ND) license.