Evaluation of an End-to-End Radiotherapy Treatment Planning Pipeline for Prostate Cancer

Advisor: Laurence Court, PhD

Radiation treatment planning is a crucial and time intensive process in radiation therapy. This planning involves the careful design of a treatment regimen tailored to a patient’s specific condition which includes type, location, and size of the tumor with reference to surrounding healthy tissues. For prostate cancer, this tumor may be either local, locally advanced with extracapsular involvement or extent into the pelvic lymph node chain. Automation of essential parts of this process would not only allow for rapid responses to tumor growth but the framework to allow for greater thought into enhancing tumor control for better outcomes.

The first objective into this works goal of automation of the treatment planning process is automatic segmentation of key structures. Delineation of both target and normal tissue structures is important as it sets the foundation for identifying where radiation must be delivered and what was be spared from excess radiation.

To accomplish this objective, deep learning segmentation models were developed from retrospective CT simulation imaging data and clinical contours to delineate both intact, postoperative, and nodal prostrate treatment structures. Quality contours were extracted in accordance to established contouring guidelines in literature. Model refinement on a holdout fine tune dataset was used to verify model contours before quantitative and qualitative evaluation. Predicted contours resulted in contours comparable in DSC and HD95 to proposed models in literature and clinically usable contours with no more than minor edits.

The second objective is the automation of VMAT planning for a breadth of prostate treatment scenarios. Development of VMAT plans for both intact, postoperative and nodal involvement treatment case necessary for both the sequence in daily treatment delivery but also the prospective distribution of radiation dose to target and normal tissues.

To accomplish this objective, Knowledge-Based planning models were developed to estimate patient specific DVH’s to guide plan optimization for delivery of radiation. Two models were developed for cases with and without lymph node involvement which includes if the prostate target is intact or postoperative with or without the presence of treatment devices such as hydrogel spacers and rectal balloons. A sequence of iterative optimization runs was created to ensure hotspot reduction and target conformality.

The findings showed that plans developed from automatically generated contours were clinically usable with minor edits for intact and postoperative treatments without lymph node involvement. For treatments with lymph node involvement, dose constraints were met for a select set of cases without excessive rectum curvature and or bladder shape. When comparing to clinical contours, clinical contours experienced the similar pass rates as those achieved by automated contours.

Advisory Committee:
Laurence Court, PhD, Chair
Carlos Cardenas, PhD
Steven Frank, MD
David Fuentes, PhD
Falk Poenisch, PhD
Julliane Pollard-Larkin, PhD
Zhiqian Yu, PhD

Attend via Zoom
Meeting ID: 836 9370 6869
Password: 984218