Author ORCID Identifier

https://orcid.org/0000-0002-8674-5657

Year of Publication

2021

Degree Name

Doctor of Philosophy (PhD)

Document Type

Doctoral Dissertation

College

Medicine

Department

Radiation Medicine

First Advisor

Dr. Damodar Pokhrel

Abstract

Lung stereotactic body radiotherapy (SBRT) is a viable alternative to surgical intervention for the treatment of early-stage non-small-cell lung cancer (NSCLC) patients. This therapy achieves strong local control rates by delivering ultra-high, conformal radioablative doses in typically one to five fractions. Historically, lung SBRT plans are manually generated using 3D conformal radiation therapy, dynamic conformal arcs (DCA), intensity-modulated radiation therapy, and more recently via volumetric modulated arc therapy (VMAT) on a C-arm linear accelerator (linac). Manually planned VMAT is an advanced technique to deliver high-quality lung SBRT due to its dosimetric capabilities and utilization of flattening-filter free beams to improve patient compliance. However, there are limitations in manual treatment planning as the final plan quality heavily depends on a planner’s skill and available planning time. This could subject the plan quality to inter-planner variability from a single institution with multiple planners. Generally, the standard lung SBRT patient ‘simulation-to-treatment’ time is 7 working days. This delays clinic workflow and degrades the quality of treatment by eliminating adaptive re-planning capabilities. There is an ongoing effort to automate treatment planning by creating a model library of previously treated, high-quality plans and using it to prospectively generate new plans termed model-based knowledge-based planning (KBP). KBP aims to mitigate the previously mentioned limitations of manual planning and improve clinic workflow.

As part of this dissertation, lung SBRT KBP models were created using a commercially available KBP engine that was trained using non-coplanar VMAT lung SBRT plans with the final dose reported from an advanced Acuros-based algorithm. The dissertation begins with the development of a robust and adaptable lung SBRT KBP model for early-stage, centrally-located NSCLC tumors that is fully compliant with Radiation Therapy Oncology Group (RTOG)-0813 protocol’s requirements. This new model provided similar or better plan quality to clinical plans, however it significantly increased total monitor units and plan complexity. This prompted the development and validation of an automated KBP routine for SBRT of peripheral lung tumors via DCA-based VMAT per RTOG-0618 criteria. This planning routine helped incorporate a historical DCA-based treatment planning approach with a VMAT optimization automated KBP engine that helps reduce plan complexity. For both central and peripheral lung lesions, the validated models are able to generate high-quality, standardized plans in under 30 min with minimal planner effort compared to an estimated 129 ± 34 min of a dedicated SBRT planner’s time. In practice, planners are expected to meticulously work on multiple plans at once, significantly increasing manual planning time. Thus, these KBP models will shorten the ‘simulation-to-treatment’ time down to as few as 3 working days, reduce inter-planer variability and improve patient safety. This will help standardize clinics and enable offline adaptive re-planning of lung SBRT treatment to account for physiological changes errors resulting from improper patient set-up.

Lastly, this dissertation sought to further expand these KBP models to support delivering lung SBRT treatments on a new O-ring linac that was recently introduced to support underserved areas and fast patient throughput. Despite learning from a C-arm modality training dataset, these KBP models helped the O-ring linac to become a viable treatment modality for lung SBRT by providing an excellent plan quality similar to a C-arm linac in under 30 min. These KBP models will facilitate the easy transfer of patients across these diverse modalities and will provide a solution to unintended treatment course disruption due to lengthy machine downtime. Moreover, they will relieve the burden on a single machine in a high-volume lung SBRT clinic. Further adaptation and validation of these KBP models for large lung tumors (> 5 cm) with multi-level dosing scheme and synchronous multi-lesion lung SBRT is ongoing.

Digital Object Identifier (DOI)

https://doi.org/10.13023/etd.2021.038

Available for download on Thursday, October 14, 2021

Share

COinS