MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Detection of Subtle Non-Gross VMAT Plan Delivery Errors Via Real-Time EPID Imaging

V Leandro Alves1*, E Aliotta2 , H Nourzadeh3 , M Ahmed4 , J Siebers5 , (1) University of Virginia Health System, Charlottesville, VA, (2) University of Virginia, Charlottesville, VA, (3) University of Virginia Health Systems, Charlottesville, VA, (4) Vanderbilt University Medical Center, TN, (5) University of Virginia Health System, Charlottesville, VA

Presentations

(Tuesday, 7/16/2019) 1:15 PM - 1:45 PM

Room: Exhibit Hall | Forum 3

Purpose: To evaluate the ability of the Swiss cheese error detection (SCED) metric to detect subtle, non-gross VMAT plan delivery errors via real-time monitoring of cine of cine EPID imaging compared with Gamma analysis.

Methods: intrEPID, an in house software developed in C++ and Varian iTools interface, which interrogates the on-board EPID aS1200 imager with cine (~10 Hz) acquisition, was used to measure cine-EPID images of two slightly different VMAT auto plans of the same head and neck case, both planned as 1 full arc, 442 MU and varied only by ten iterations of optimization with the same objectives. The dose-coverage of the plans differed slightly as they were exported from the same plan optimization session, at differing number of iterations. SCED utilizes a dynamic aperture masking algorithm, which uses just-in-time interpolation of intended MLC positions to predict each frame’s beam apertures to ascertain aperture delivery errors. For gamma, a 3%/3 mm, 10% dose threshold and >90% passing rate was used to identify delivery errors.

Results: The two treatment plans are highly similar, yet have distinct modulation. In acquisition, each plan resulted in 688 acquired EPID frames. SCED just-in-time masking detected 11.8% of frames with than a >20 mm2 area mismatch. Gamma analysis indicated that 8.4% of the frames have <90% of points with γ<1. During-delivery gamma analysis on the composite of the delivered frames had >90% of points with γ<1 through-out the delivery. Mimicking clinical practice, the composite at the conclusion of the delivery indicated a 99.7% passing rate.

Conclusion: Real-time monitoring, which initially aimed to provided gross error detection, can detect subtle, non-gross errors such as a plan which differs slightly from the intended delivery. Detection requires frame-by-frame analysis, and is trivial when the SCED metric is used.

Funding Support, Disclosures, and Conflict of Interest: This work was funded by Varian Medical Systems.

Keywords

Quality Control, Treatment Verification

Taxonomy

Not Applicable / None Entered.

Contact Email