Clinical Commissioning and Implementation of An In-House Artificial Intelligence (AI) Tool for Automatic Head-And-Neck Intensity Modulated Radiation Therapy (IMRT) Treatment Planning
X Li1*, Y Sheng1, QJ Wu1, Y Ge2, C Wang1, D Brizel1, Y Mowery1, J Lee1, W Wang1, H Stephens1, F Yin1, Q Wu1, (1) Duke University Medical Center, Durham, NC, (2) University of North Carolina at Charlotte, Charlotte, NC
Presentations
(Saturday, 3/26/2022) [Central Time (GMT-5)]
Purpose: Radiation treatment planning for head and neck (HN) cancers is generally considered to be challenging and labor-intensive. With recent advancements in deep learning, we developed an Artificial Intelligence (AI) planning tool for HN IMRT that predicts fluence maps from patient anatomy. The AI tool demonstrated promising plan quality and efficiency in retrospective validation and testing. In this study, we describe the clinical implementation and commissioning process of this in-house tool.
Methods: The AI tool was packaged with a graphical user interface (GUI) that was interfaced to the commercial treatment planning system (TPS) in the clinical environment. A designated workstation was configured to operate the AI algorithms to minimize the load and effect on the clinical environment. The AI-generated fluence map was imported into TPS for dose calculation, followed by an optional automatic fine-tuning step to finalize the plan. Planners could visually examine the AI plan’s dose distribution and make further adjustment as needed clinically. The AI tool’s workflow and performance were evaluated and validated by physicians and physicists specialized in HN treatment. Training, validation, and testing datasets of the AI tool consist of 200, 16, and 15 retrospective cases. The commissioning dataset of the AI tool employed 28 more recent cases. The commissioning AI plans were retrospectively evaluated based on the corresponding clinic plans regarding critical dose-volume endpoints and 3D dose distributions.
Results: The average plan generation time including manual operations was about 10~15min/case. The AI dose distributions were reasonable, with slightly better dose conformity (1.12±0.03 versus 1.17±0.07 in clinical plans). OAR sparing was comparable with the clinical plans.
Conclusion: The in-house AI IMRT treatment planning tool was commissioned for HN in our clinic. The commissioning process demonstrates outstanding performance and robustness of the AI tool, and the established commissioning workflow provides sufficient validation and documentation for clinical use.
ePosters
Keywords
Radiation Therapy, Treatment Planning, Commissioning
Taxonomy
TH- External Beam- Photons: treatment planning/virtual clinical studies
Contact Email