Material and Methods

a) Description of the dose verification system

The schematic of the dose verification system is described in Figure 1. A thin cylindrical PS is inserted in a central position of a cylindrical tank. This phantom is filled with water to maintain the electron equilibrium condition. This phantom should be placed along the rotation axis of the linac gantry so that the X-ray is incident on the side of PS. A CCD camera was placed away from the water tank to avoid radiation damage. It measures the light distribution of PS from the transparent front part of phantom.

b) Reduction of Cherenkov radiation

When X-ray is irradiated to the phantom, not only the PS emits scintillation light but also the Cherenkov radiation is generated in the water. Thus, the obtained CCD images contain both the scintillation light and Cherenkov radiation. By performing image subtraction, component of Cherenkov radiation is reduced. To this end, we also measured the light distribution using the phantom without PS under the same conditions as the phantom irradiation with PS.

c) Correction for diffusion effects of scintillation light 

Owing to the process of light propagation and photo detection, measured scintillation light distribution is more widespread than the original distribution. We represent this effect as point spread function (PSF). To determine the PSF, we make a point light source at the center of the PS using ultraviolet light with a pinhole collimeter and measure its distribution by the CCD camera.

d) Experimental condition of radiation irradiation 

10 MV X-ray beams at dose rate of 600 MU/min was used in this study. The field size was changed from 4×4 to 10×10 cm². The frame rate of the CCD camera was set to 4 fps.

e) Analysis

The accuracy of our dose verification system was evaluated via the mean relative error (MRE) between the scintillation light distribution and the convolved dose distribution. To evaluate high dose values, which is the most important in clinical use, top 20 % dose area was set to region of interest. While the scintillation light distribution was obtained through Cherenkov radiation subtraction process, the convolved dose distribution was obtained by convolving the calculated dose distribution of treatment planning system with the PSF of diffusion effects.

Results and discussions

a) Reduction of Cherenkov radiation

The difference of pixel values of CCD images under the two conditions, phantom irradiation with PS and phantom irradiation without PS, indicates the contribution ratio of Cherenkov radiation. The calculated results are listed in Table 1. The contribution of the Cherenkov radiation increases according to the radiation field size, which was up to 14.288%. That is because the irradiation volume to water increases with the radiation field size. By using image subtraction technique, this effect was easily corrected.

b) Correction for diffusion effects of scintillation light

The measured 2D distribution of the point light source was well fitted by single Gaussian distribution. However, the PSF should form double Gaussian distribution because it would consist of 2 components: propagation light from a point light source (narrow Gaussian distribution) and diffusion light in PS (wide Gaussian distribution). This discrepancy is probably because the intensity of diffusion light in PS is much smaller than that of propagation light from a point light source. Thus, the second component of double Gaussian distribution could not be obtained from this experimental data. Therefore, we determined the second component from the scintillation light distribution and the calculated dose distribution with field size of 10×10 cm² using an optimization technique for now. The final double Gaussian distribution is shown in Figure 2.

c) Accuracy of our dose verification system 

Comparison between the light distributions and convolved dose distributions are shown in Figure 3, and the calculated MREs are listed on the Table 2. Considering the diffusion effects of scintillation light, the MREs of correction results were under 1 % for each field size.