EUVIMEDEUVIMED entdecken
EvaluationsstudieOpen Access · CC BY 4.0

Evaluation of a prostate SBRT planning workflow using auto‐segmentation and knowledge‐based planning

Trisha Jones, Kirk Luca iD, Mingdong Fan, Pretesh R Patel, Ashesh B Jani, Xiaofeng Yang iD, Justin Roper, Jie Ding iD

Journal of Applied Clinical Medical Physics · 2025

DOI: 10.1002/acm2.70295 ↗

Kurz gefasst

Evaluationsstudie in der Zeitschrift Journal of Applied Clinical Medical Physics, 2025. Verfasst von Trisha Jones und 7 weiteren Autor:innen. Beteiligte Einrichtung: Department of Radiation Oncology, Emory University, Atlanta, Georgia, USA. Hauptthemen laut MEDLINE: Prostatic Neoplasms, Radiotherapy Planning, Computer-Assisted, Radiosurgery, Artificial Intelligence. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 51 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: NIH HHS, National Institutes of Health.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Purpose

This study investigates a prostate stereotactic body radiotherapy (SBRT) planning workflow integrating commercial artificial intelligence (AI) contouring and knowledge‐based planning (KBP). The purpose is to determine whether AI‐generated contours are comparable to physician‐delineated contours in achieving dosimetric goals.

Methods

In this retrospective study, 20 patients with intact prostate cancer were included. A commercial AI contouring software was applied to computed tomography (CT) scans. Contouring accuracy for the prostate, rectum, and bladder was assessed against clinical contours using geometric metrics including Dice similarity coefficient (DSC), surface DSC (sDSC), and added path length (APL). Volumetric modulated arc therapy plans were generated using an in‐house prostate SBRT KBP model without user interaction. The prescribed dose was 36.25 Gy in 5 fractions to the planning target volume with a 40 Gy simultaneous integrated boost to the prostate. All plans were calculated using AcurosXB and normalized to cover 98% of the prostate. Given the high precision required for prostate SBRT, all plans used the physician‐delineated (clinical) prostate contours. Three plans were generated for each patient: (1) a reference plan using clinical contours, (2) a plan using clinical prostate contour and AI organs at risk (OARs), and (3) a plan using clinical prostate contour and post‐processed AI OARs that removed any overlap with the clinical prostate contour. The latter two plans were recalculated on clinical contours with fixed monitor units to evaluate the dosimetric impact of AI contouring. Plan quality was evaluated using NRG‐GU013 criteria.

Results

The average DSC values were 0.83, 0.86, and 0.94 for prostate, rectum, and bladder, respectively, and the average sDSC values were 0.62, 0.81, and 0.85. AI prostate contours were clinically unacceptable. AI rectum and bladder contours overlapped the clinical prostate contour in 15 and 20 cases, respectively. All reference plans using clinical contours met NRG criteria. Using AI OARs or post‐processed AI OARs, only one case exceeded rectum V36Gy limits due to over‐contouring, but it became clinically acceptable after recalculation on clinical contours. Plans using post‐processed AI OARs yielded dosimetric results more comparable to reference plans for rectum and bladder sparing.

Conclusions

This study investigates a prostate SBRT treatment planning workflow by leveraging AI contouring and KBP. Although a fully automated workflow is not yet feasible, results are encouraging when physician‐delineated target volumes are combined with post‐processed AI contours.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Prostatic NeoplasmsRadiotherapy Planning, Computer-AssistedRadiosurgeryArtificial IntelligenceImage Processing, Computer-AssistedHumansMaleOrgans at RiskRadiotherapy, Intensity-ModulatedRetrospective StudiesRadiotherapy DosageWorkflowTomography, X-Ray ComputedKnowledge BasesAlgorithms

Stichwörter der Autor:innen

AI contouringProstate SBRTknowledge‐based planning

Autor:innen und Institutionen

  • Department of Radiation Oncology, Emory University, Atlanta, Georgia, USA.
  • Medical Dosimetry Program, School of Health Sciences, Southern Illinois University, Carbondale, Illinois, USA.

ORCID-Profile: Kirk Luca, Xiaofeng Yang, Jie Ding

Förderung

  • NIH HHS
  • National Institutes of Health

Zitieren

APA

Jones, T., Luca, K., Fan, M., Patel, P. R., Jani, A. B., Yang, X., Roper, J., & Ding, J. (2025). Evaluation of a prostate SBRT planning workflow using auto‐segmentation and knowledge‐based planning. Journal of Applied Clinical Medical Physics. https://doi.org/10.1002/acm2.70295

Vancouver

Jones T, Luca K, Fan M, Patel PR, Jani AB, Yang X, et al. Evaluation of a prostate SBRT planning workflow using auto‐segmentation and knowledge‐based planning. Journal of Applied Clinical Medical Physics. 2025. doi:10.1002/acm2.70295

BibTeX
@article{Jones2025,
  author = {Jones, Trisha and Luca, Kirk and Fan, Mingdong and Patel, Pretesh R and Jani, Ashesh B and Yang, Xiaofeng and Roper, Justin and Ding, Jie},
  title = {Evaluation of a prostate SBRT planning workflow using auto‐segmentation and knowledge‐based planning},
  journal = {Journal of Applied Clinical Medical Physics},
  year = {2025},
  doi = {10.1002/acm2.70295}
}
Kostenlos starten →