MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Towards a Treatment Planning Optimization Framework Utilizing Predicted Quality Assurance Outcomes From a Machine Learning Model to Maximize Plan Quality and Deliverability

P Wall1*, J Fontenot1,2, (1) Louisiana State University, Baton Rouge, LA, (2) Mary Bird Perkins Cancer Center, Baton Rouge, LA

Presentations

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

Room: AAPM ePoster Library

Purpose: Current inverse planning algorithms incorporate specific mechanical restrictions – such as constraining leaf motion – that are designed to reduce the complexity of the treatment plan. However, mechanical constraints do not guarantee a corresponding improvement in measured quality assurance (QA) outcomes. Therefore, this work explored the feasibility of an optimization framework for directly maximizing predicted QA outcomes of plans without compromising dosimetric quality.

Methods: VMAT plans were retrospectively designed for ten prostate patients. A support vector machine (SVM) was developed – using a database of 500 previous VMAT plans – to predict gamma passing rates (GPRs; 3%/3mm percent dose-difference/distance-to-agreement with local normalization) based on selected complexity features. An optimization algorithm was devised by utilizing the SVM model to iteratively modify mechanical treatment features most commonly associated with suboptimal GPRs. Specifically, leaf gaps (LGs) less than 5 cm were widened by random amounts, which impacts several complexity features such as small aperture scores and aperture area uniformity. The original ten plans were optimized with this QA-based algorithm using maximum LG displacements of 1, 3, and 5 mm before corresponding changes in predicted GPRs and dose were assessed.

Results: Predicted GPRs increased by an average of 0.30, 1.25, and 1.69% after QA-based optimization for 1, 3, and 5 mm maximum random LG displacements, respectively. Differences in dose were minimal, resulting in negligible changes in tumor control probability (maximum increase = 0.1%) and normal tissue complication probability (maximum decrease = 0.3% among bladder, rectum, and femoral heads).

Conclusion: A novel algorithm for optimizing predicted GPRs was developed and shown to increase predicted deliverability without degrading dosimetric quality of given plans. This method for integrating QA outcomes directly into planning optimization could help improve the probability and efficiency of arriving at a truly optimal treatment in terms of both dosimetric quality and deliverability.

Funding Support, Disclosures, and Conflict of Interest: This work was supported in part by a grant through the Mary Bird Perkins Cancer Foundation.

Keywords

Optimization, Quality Assurance, Treatment Planning

Taxonomy

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

Contact Email