Author: Shifeng Chen, Erica Fisler, Arun Gopal, Mariana Guerrero, Jason Hendershot, Kai Huang, Eric Kusmaul, Adam Schrum, Megan Steinberg, Kai Wang 👨🔬
Affiliation: University of Maryland School of Medicine, Department Radiation Oncology, Kaufman Cancer Center, Upper Chesapeake Medical Center, University of Maryland Medical System Central Maryland Radiation Oncology, Department of Radiation Oncology, University of Maryland Medical Center, Baltimore Washington Medical Center Radiation Oncology, Department of Radiation Oncology, University of Maryland School of Medicine 🌍
Purpose:
Our goal is to develop and clinically deploy an automated planning tool using multicriteria optimization (MCO) for VMAT in prostate only, whole pelvis radiotherapy treatments.
Methods:
An in-house Python-based tool was developed for RayStation 11B to automate the creation of VMAT plans with two full arcs using the MCO technique. The tool was initially validated on 10 retrospective prostate only cancer patients with the dosimetric comparison made between the automated plans and the corresponding clinical plans. The automated plans were scored by dosimetrists on a four-point scale: use-as is, acceptable needing physician preferred changes, needs further optimization, and needs replan. The tool was later expanded to generate whole pelvis plans. Three dosimetrists were recruited to pilot the tool clinically, scoring each generated plan using the four-point scale. Additionally, the extra time needed to optimize and finalize the plan were recorded.
Results:
Plan comparisons of the MCO-generated prostate only plan and the clinical plans showed statistically significantly reduced doses to OARs for the MCO plans. Dosimetrists found 82.5% of the prostate only plans to be acceptable needing physician preferred changes, while 17.5% required further optimization, and no plan necessitated replan. The tool was clinically implemented with 3 dosimetrist users as pilot test. Of the 16 recorded clinical use, 13 plans were scored as acceptable needing physician preferred changes, one plan each was rated to be use as is, unacceptable requiring further optimization, and needs replan, respectively. All the whole pelvis plans were scored as acceptable needing physician preferred changes. On average dosimetrists spent 10 minutes to further optimize the plans.
Conclusion:
This study demonstrated the process of developing and implementing an MCO-based auto-planning tool in the clinical workflow, with the goal of standardizing processes, reducing workload, and continuously improving the tool based on the feedback from dosimetrists.