J Zhang¹*, Y Yang², Y Chen², J Zhang³, M Chen², (1) The Hong Kong Polytechnic University, Hung Hom, Kowloon, HKSAR, (2) Zhejiang Cancer Hospital, University Of Chinese Academy Of Sciences, Hangzhou, Zhejiang, CHN, (3) Duke University Medical Center, Durham, NC

• 
  J Zhang

(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

Purpose: To propose the new angle descriptor-projection overlap volume (POV)-and evaluate its performance in improving the accuracy of similar plan referencing and DVH prediction used in our in-house developed plan quality assurance pipeline for Lung cases.
Methods: Our previously proposed plan similarity metric was adapted by adding the contribution from POV. We also included knowledge-based DVH prediction in our plan quality assurance pipeline in addition to statistical inferencing of similar plans. Thirty treatment plans were randomly selected from each of the 470 VMAT and 334 IMRT Lung databases and used for independent validations. Improvement of plan similarity was validated by statistically comparing clinical used dose objectives similarity of reference plans selected by the plan similarity metric with and without POV for Lungs, Heart, and SpinalCord. DVH prediction root mean square errors were also compared for the three OARs between POV included and POV free models.
Results: The mean absolute difference of Heart mean dose between referenced similar plans and target plan has been reduced by 1.33Gy (p-value < 0.01) in IMRT and 1.31Gy (p-value < 0.01) in VMAT on average. Limited improvement of dosimetric similarity has been observed for Lungs and no improvement for SpinalCord maximum dose. DVH prediction has seen 40.8% mean decrease (p-value < 0.001) in VMAT and 34.6% (p-value < 0.009) in IMRT for Heart. There has been limited DVH prediction accuracy improvement for SpinalCord and no improvement for Lungs.
Conclusion: POV quantifies the angular relationship between the target volume and OAR. It significantly increases the accuracy of reference-plan based plan quality evaluation and improves DVH prediction accuracy, especially for Heart. More evaluations on the plan quality assurance pipeline, such as replan of the identified low-quality plans, are warranted in future studies.

Keywords

Quality Assurance, Lung, Treatment Planning

Taxonomy

TH- External Beam- Photons: Treatment planning using machine learning/Knowledge Based Planning/automation

Contact Email