Cherenkov emission‐based external radiotherapy dosimetry: II. Electron beam quality specification and uncertainties

Purpose Cherenkov emission (CE) is ubiquitous in external radiotherapy. It is also unique in that it carries the promise of 3D, micrometer‐resolution, perturbation‐free, in‐water dosimetry with a beam quality‐independent detector response calibration. Our aim is to bring CE‐based dosimetry into the clinic and we motivate this here with electron beams. We Monte Carlo (MC) calculate and characterize broad‐beam CE‐to‐dose conversion factors in water for a clinically representative library of electron beam qualities, address beam quality specification and reference depth selection, and develop a preliminary uncertainty budget based on our MC results and relative experimental work of a companion study (Paper I). Methods Broad electron beam CE‐to‐dose conversion factors kCθ±δθ include CE generated at polar angles θ ± δθ on beam axis in water. With modifications to the EGSnrc code SPRRZnrc, kCθ±δθ factors are calculated for a total of 20 electron beam qualities from four BEAMnrc models (Varian Clinac 2100C/D, Clinac 21EX, TrueBeam, and Elekta Precise). We examine beam quality, depth, and detection angle dependence for θ±δθ=90∘±90∘ (4π detection), 90∘±5∘, 45∘±45∘, and 90∘±45∘. As discussed in Paper I, 4π detection offers the strongest CE‐dose correlation and θ=90∘ with small δθ is most practical. The two additional configurations are considered as a compromise between these two extremes. We address beam quality specification and reference depth selection in terms of the electron beam quality specifier R50, obtained from the depth of 50% CE C50, and derive a best‐case uncertainty budget for the CE‐based dosimetry formalism proposed in Paper I at each detection configuration. Results The kCθ±δθ factor was demonstrated to capture variations in the beam spectrum, angle, photon contamination, and electron fluence below the CE threshold (∼260 keV in the visible) in accordance with theory. The root‐mean‐square deviation and maximum deviation of a second‐order polynomial fit of simulated R50 values in terms of C50 were 0.05 and 0.11 mm at 4π and 0.20 and 0.33 mm at 90∘±5∘ detection, respectively. The fit performance on experimental data in Paper I was in agreement with these values within experimental uncertainties (±1.5 mm, 95% CI). A two‐term power function fit of kCθ±δθ in terms of R50 at a reference depth dref=aR50+b resulted in total dref‐dependent dose uncertainty contribution estimate of 0.8% and 1.1% and preliminary best‐case estimate of the combined standard dose