MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Towards Automation of Treatment Planning System Quality Control

J A Lovis1*, L Van Dyke1, M Roumeliotis1,2, K Thind1,2, S Quirk1,2, (1) Tom Baker Cancer Centre, Calgary, AB, CA, (2) University of Calgary, Calgary, AB, CA

Presentations

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

Room: AAPM ePoster Library

Purpose: To develop and validate a suite of automated software modules designed to streamline robust, and efficient quality control (QC) of the Varian Eclipse external-beam treatment planning system (TPS).

Methods: Automated software modules were developed using the Eclipse Scripting API to implement recommended tests in the Medical Physics Practice Guideline 5a (MPPG5a). Modules were generated to: (1) run batch analysis (all energies and field sizes) of profile comparison and percent depth dose, (2) visualize individual comparisons, (3) patient-specific plan comparison between dose calculation algorithms, (4) assess integrity of the TPS dose calculation framework, and (5) generate PDF reports. Each software module is designed to evaluate differences between two calculation algorithms, and validate with measured data comparison. Analysis was completed for: 6MV, 10MV, 15MV, 6FFF, and 10FFF and field sizes of 3x3cm² up to 40x40cm². Evaluation were completed with tolerances: 2% infield and 3% out-of-field/penumbra (MPPG5a).

Results: Comparison of 1100 PDDs, profiles, and over 50 patient plans calculated using the Eclipse Analytics Anisotropic Algorithm (AAA_13623 and AAA_15606) showed all field sizes and energies were within tolerance for all regions, except for select small field metrics. For 3x3cm² and 4x4cm² the average out-of-field deviation increased up to 21.3% for all photon energies. These profiles evaluated with the single profile analysis module, allowing for visualization of dose profiles generated from both calculation models and validated with Varian reference data. This demonstrated that the relative increase to out of-field-dose was large between algorithms; however, absolute difference was less than 1%. Plan comparisons using the patient duplication module showed changes of less than 1% to MUs, dose volumes (V30%, V50%, V70%, V90%, and V95%), maximum, minimum, and mean doses.

Conclusion: The integration of a suite automated software modules into routine QC workflow enables repeatable, robust, and reportable characterization of tests outlined in MPPG 5a.

Keywords

Acceptance Testing, Computer Software, Quality Assurance

Taxonomy

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

Contact Email