A Fluence Reconstruction Framework to Resolve Spectral Under-Response In Epid-Based SRS QA
Abstract
Purpose
To implement a correction methodology for amorphous silicon Electronic Portal Imaging Device (EPID) dosimetry that resolves systemic discrepancies due to beam hardening, aperture-specific response, and tongue-and-groove (TG) effects for Varian MLCs. Traditional EPID-based quality assurance deviates from ionization chamber (IC) measurements due to the detector’s under-response to hardened leakage and optical scatter in the phospor layer that causes a loss of signal in small fields.
Methods
The RT-Plan is used to calculate correction maps by separately calculating transmission fraction, instantaneous physical gaps, and TG edge components. This separation allows independent application of: (1) spectral correction, (2) aperture-dependent output factor correction, and (3) an absolute TG correction to the measurement. A simultaneous solver optimized three parameters: a leaf-tip offset and an exponential aperture correction function to match IC-measured sweeping gap dose ratios (5–30 mm). These corrections are relevant for HyperArc plans, where the projection of off-axis targets on the EPID create dose-spreading effects with large contributions from the transmitted beam and small apertures.
Results
For a 6FFF beam on a TrueBeam with Millennium MLC, the spectral factor was U=0.56, indicating EPID-measured transmission is nearly half that of an IC. Optimization yielded a 0.8 mm geometric leaf-tip offset, with aperture correction factors ranging from 1.05 (5 mm gap) to 1.01 (≥20 mm). In a proof-of-concept analysis of two HyperArc plans, the methodology significantly improved gamma passing rates (3%/1mm): from 80.6±10.5% to 95.7±1.7%.
Conclusion
Accurate EPID dosimetry requires accounting for detector-specific responses to beam characteristics. By incorporating the proposed spectral, aperture, and spatial TG corrections, the EPID response was successfully adjusted to that of an ionization chamber. No adjustments of beam model parameters should be made based on EPID signals without properly modeling these detector effects. Furthermore, similar response analyses should be extended to other detectors used in patient-specific QA programs.