Room: AAPM ePoster Library
Purpose:
Patient-specific VMAT QA using EPID-based portal dosimetry can potentially mask errors within a treatment plan due to the rotation of EPID with gantry. The purpose of this work was to investigate the feasibility of splitting VMAT plans into angular segments to better detect and quantify potential plan errors.
Methods:
VMAT plans were anonymized and a commercial treatment planning system was used to divide full arc plans into segments of 5°, 10°, 20°, 45°, 90°, and 180°. Development mode on a Varian TrueBeam was used for automated acquisition of portal dosimetry. Composite angle-resolved measurements were compared to full-arc measurements using portal dosimetry and a cylindrical diode array phantom. Portal dosimetry measurements were analyzed using a gamma-index with 3%/2 mm 1 cm MLC CIAO, while phantom measurements were assessing using 3%/2 mm 10% threshold.
Results:
Python scripts were used to automate segment acquisition and image collection within the same amount of time as a standard full-arc plan. Dummy control points were used between segments to ensure proper image acquisition, resulting in all segments = 90° passing. Without these control points built in, 63% of segments failed for divisions up to 90°. Standard VMAT plans showed 42% of segments failed with divisions = 45°, while SBRT plans showed failures at 5° only (1%). Passing results for composite images matched passing results for standard portal and phantom measurements.
Conclusion:
Gamma-index failure rates increased with decreasing arc-segment size in angle-resolved portal dosimetry of VMAT plans, despite consistent results between composite segment images and standard portal dosimetry/phantom measurements. Based on SBRT results, small errors in dose delivery appear to lead to large gamma errors when small doses are delivered. Further investigation is needed to quantify the nature and impact of the observed discrepancies, as well as the feasibility of angle-resolved portal dosimetry for VMAT QA.
Not Applicable / None Entered.