2456 - Initial Assessment of Organ-at-Risk Variance Using Hounsfield Unit-Accurate Cone Beam CT and Auto-Segmentation for External Beam Radiation Therapy for Prostate Cancer

D. Zaks, R. J. Lalonde, M. S. S. Huq, and A. C. Olson; Department of Radiation Oncology, UPMC Hillman Cancer Center, Pittsburgh, PA

Purpose/Objective(s): Recent advancements in cone beam CT (CBCT) imaging include Hounsfield Unit (HU) accuracy, enabling calculation of dose distribution directly on the CBCT image. This advance affords the opportunity to calculate accumulated radiation dose to critical organs-at-risk (OARs) on a per-fraction basis. We report our initial feasibility assessment of combining HU-accurate CBCT with AI-generated auto-segmented OARs for prostate radiation therapy as we formulate an offline adaptive workflow. Materials/

Methods: A single patient receiving prostate radiotherapy on a linear accelerator with a 6-DoF couch was imaged daily with HU-accurate CBCT for the first five fractions of a planned 25 fraction course. HU-accurate CBCT images were exported, critical OARs (bladder, rectum, penile bulb, bowel bag) were auto-contoured using commercially available software with physician verification for accuracy, and then re-imported back into our treatment planning system. The planned fields were copied over to the auto-contoured image set, shifts were applied based on the alignment performed at the console, and dose was computed without altering the fluence pattern or monitor units. DVHs for OARs were reviewed.
Results: DVHs from the bladder, rectum, and bowel bag were found to match closely to prescription dose. Bladder, rectum, and penile bulb delivered doses were all below the per-fraction constraints from the prescription, while the accumulated dose for the bowel bag was greater than prescription dose. Penile bulb dose for fractions 2 and 3 were above the per-fraction constraint, but after five fractions total delivered dose was below the expected dose sum based on the prescription. The percent volume changes from CT sim to the median CBCT volumes were -41.5% (bladder), 1.4% (rectum), 25.5% (penile bulb), and -41.0% (bowel bag). See Table 1.
Conclusion: Measuring OAR dose with daily DVHs using HU-accurate CBCT and auto-segmented OARs is a clinically feasible workflow. Accumulated OAR dose, rather than prescription dose, could potentially improve normal tissue complication probability estimation for OARs that deform and change shape between fractions. Future directions include expanding automation to improve efficiency, comparing different auto-segmentation tools for OAR delineation, and integrating deformable registration of planning target volumes into an offline adaptive workflow.

	Volume on CT simulation (cc)	Volume on CBCT, median (IQR) (cc)	OAR Constraint	Expected Dose Sum, 5 Fractions (Gy)	Summated Dose, 5 fractions (Gy)
Bladder	351.6	497.4 (361.5-633.4)	V48 < 25%	9.6	8.0
Rectum	44.0	43.4 (40.0-46.8)	V52 < 30%	10.4	8.3
Penile Bulb	4.7	3.5 (2.1-4.9)	Mean < 40 Gy	8	7.3
Bowel Bag	1932.8	2726.0 (2413.7-3038.3)	V55 < 5 cc	11	12.8