Estimation and correction of produced light from prompt gamma photons on luminescence imaging of water for proton therapy dosimetry

Abstract

Although the luminescence images of water during proton-beam irradiation using a cooled charge-coupled device camera showed almost the same ranges of proton beams as those measured by an ionization chamber, the depth profiles showed lower Bragg peak intensities than those measured by an ionization chamber. In addition, a broad optical baseline signal was observed in depths that exceed the depth of the Bragg peak. We hypothesize that this broad baseline signal originates from the interaction of proton-induced prompt gamma photons with water. These prompt gamma photons interact with water to form high-energy Compton electrons, which may cause luminescence or Cherenkov emission from depths exceeding the location of the Bragg peak. To clarify this idea, we measured the luminescence images of water during the irradiations of protons in water with minimized parallax errors, and also simulated the produced light by the interactions of prompt gamma photons with water. We corrected the measured depth profiles of the luminescence images by subtracting the simulated distributions of the produced light by the interactions of prompt gamma photons in water. Corrections were also conducted using the estimated depth profiles of the light of the prompt gamma photons, as obtained from the off-beam areas of the luminescence images of water. With these corrections, we successfully obtained depth profiles that have almost identical distributions as the simulated dose distributions for protons. The percentage relative height of the Bragg peak with corrections to that of the simulation data increased to 94% from 80% without correction. Also, the percentage relative offset heights of the deeper part of the Bragg peak with corrections decreased to 0.2%-0.4% from 4% without correction. These results indicate that the luminescence imaging of water has potential for the dose distribution measurements for proton therapy dosimetry.

Prediction of dose distribution from luminescence image of water using a deep convolutional neural network for particle therapy

“…15 As a result, we obtained almost the same depth profiles as the dose for protons. 15 However, this correction method needs to estimate the Cerenkov light distribution by Monte Carlo (MC) simulation, which takes a long time to produce images. More than one day was required to produce the distribution for single energy by MC simulation.…”

“…For the luminescence imaging simulation, we calculated the images that had almost the same distributions as the measured luminescence images of water during proton and carbon-ion irradiation using the previously developed MC simulation. 15 We calculated the luminescence of water images for different irradiated energies and positions using the MC simulation. We used the scintillation photon-generation process as the method of light production from the luminescence of water at a lower energy than the Cerenkov light threshold.…”

“…For the refractive index of water and the spectrum of luminescence of water, we used the same values as those in our previous study. 15 In the optical photon simulation, we scored the optical photons generated from the Cerenkov light generation process in the water phantom during the irradiation of the protons or carbon ions. The camera's lens (30 mm 9 30 mm) was located 40 cm from the phantom surface.…”

“…Cerenkov light produced from the secondary electrons generated by the interaction of the prompt gamma photons with water caused this phenomenon. 15 For carbon ions, the luminescence image's depth profiles showed high intensity in the shallow area. 12 This high intensity was thought to be mainly due to the Cerenkov light emitted by secondary electrons whose energy was higher than the Cerenkov light threshold in the shallow area.…”

“…This proposed method has two advantages. One is that we can predict the dose distribution from the measured luminescence images faster and simpler than with our previous method 15 because we do not need MC simulation to predict the images after training the neural network. The other is that we can derive 2D dose distributions as well as 1D-depth doses using DCNN.…”