3394 - Effects of Prostate Auto-Planner on Plan Quality, Dosimetric Constraints, and Clinic Workflow: A Real World Comparison

Purpose/Objective(s):

We developed an automated approach for prostate cancer radiation treatment planning and validated its performance in the preclinical setting. The objective of this study was to conduct an inter-era comparison of real-world treatment plan quality metrics and workflow benchmarks before and after implementation of the autoplanner. The automated planning process is initiated after the treating radiation oncologist approves the structure set; the autoplanner then performs end-to-end plan generation without further manual intervention.

Materials/Methods:

We identified all prostate cancer treatment plans since January 1, 2024 which contained a single planning target volume (PTV) containing the prostate gland and optionally a portion of the seminal vesicles prescribed 70 Gy in 28 fractions. Cases were classified into before or after prostate autoplanner implementation in March 2025. PTV coverage metrics and organ-at-risk (OAR) dose metrics were exported and analyzed. Workflow timestamps were extracted from our clinical database to quantify contour approval, autoplan creation, and plan approval times. were used to assess intergroup differences in dosimetry and workflow times.

Results:

We identified 89 prostate treatment plans that met our inclusion criteria, 57 plans were prior to the implementation of our auto-planner, and 32 plans after. There were no statistical differences in PTV coverage or OAR dose metrics (Table). Although not statistically significant, all evaluated rectal and bladder metrics trended lower post-autoplanner, with mean reductions of approximately 3–12% across V40Gy–V70Gy. Prior to autoplanner implementation, the mean time from structure approval to plan approval was 5.2 days (SD: 3.2) as compared to 4.1 days (SD: 3.1 days) following autoplanner implementation (p=0.037).

Conclusion:

We implemented an automated prostate cancer treatment planning tool into clinical use at an academic radiation oncology center. In this real-world implementation analysis, autoplanner was associated with a shorter interval from contour generation to plan approval. Dose-volume metrics were comparable between pre- and post-implementation plans, though directional improvements in bladder and rectal metrics were observed. Larger cohorts are needed to confirm these trends.

Structure / Metric	Pre-/Post-Autoplanner	N	Mean	p-value
Rectum V70Gy(cc)	Pre	57	2.10	0.532
	Post	31	1.97
Rectum V50Gy(cc)	Pre	57	7.95	0.827
	Post	32	7.69
Rectum V40Gy(cc)	Pre	57	10.74	0.834
	Post	32	10.45
Bladder V70Gy(cc)	Pre	57	11.19	0.175
	Post	32	9.92
Bladder V50Gy(cc)	Pre	57	26.34	0.166
	Post	32	23.81
Bladder V40Gy(cc)	Pre	57	36.32	0.113
	Post	32	32.05