Radiochromic film was first developed in 2004 and is the most commonly used dosimetric radiotherapy tool [1]. Gafchromic EBT3 films (Ashland Inc., Wayne, NJ, USA), currently the most widely used radiochromic films, feature high spatial resolution, tissue equivalence characteristics, and the advantage of two-dimensional (2D) measurements [2,3]. These films are also used for quality assurance and to determine in vivo dosimetry when administering advanced radiotherapy techniques like intensity-modulated radiation therapy and volumetric-modulated arc therapy [3-7].

Radiation doses to the EBT3 film surface are determined by obtaining scanner images of the film’s opacity before and after radiation exposure. Radiation dose values are then derived from red-green-blue (RGB) image values which mathematically represents color information using three or four different color parameters. This step is followed by color quantification standardization. However, procedural variables like active layer non-uniformities, scanning non-uniformities, and film scan angle can affect dosimetric accuracy [1,8-11].

Various institutions have evaluated the overall dosimetry uncertainty of EBT3 films according to RGB configurations within various measurement and scanning environments. These experiments revealed the dosimetric uncertainty estimates of 3.2% for R, 4.9% for G, and 5.2% for B channels [12]. Another study proposed a calibration curve that used the triple-channel analysis method to reduce dosimetric uncertainty and relative standard deviation and improve dosimetric accuracy [13,14]. These efforts underscore the importance of accurately measuring radiation doses to EBT3 film; relevant research is ongoing.

Texture analysis quantifies pixels within 2D images that cannot be recognized with the naked eye [15-17]. Texture analysis is used for radiomic analysis of medical images. For example, a patient’s prognosis can be determined using texture analysis. 2D images can be quantified through texture analysis and used for dose measurement [18]. Shih et al. obtained polymer gel surface images using scanning electron microscopy (SEM) after application of various (5, 10, 15, and 20 Gy) radiation doses. The researchers analyzed four textural features to determine potential correlations with the delivered doses [19]. Using the dose calibration curve, the authors found differences in delivered doses of −7.60% for 5 Gy, 5.80% for 10 Gy, 2.53% for 15 Gy, and −0.95% for 20 Gy [19].

Images of the film’s active layer were obtained using SEM after irradiation with various radiation doses. Each SEM image was subjected to texture analysis to determine its relevant features. We additionally determined the dose calibration curve between the calculated and delivered dose values to evaluate the dose calibration curve’s accuracy.

1. Radiochromic EBT3 film

The EBT3 film consists of a 28-μm-thick active layer that reacts to radiation and a 125-μm-thick transparent polyester substrate layer that protects the active layer. The substrate layer is positioned above and below the active layer to reduce damage and interference by external ultraviolet (UV) and normal lights. The active layer in the EBT3 film contains yellow-colored, radiation-sensitive chemicals that become opaque after polymerization from radiation exposure. The adjusted opacity value obtained from each of the RGB channels is used to calculate optical density, which strongly correlates with radiation dose. We purchased EBT3 film samples with the upper substrate layer removed to directly image the active layer’s surface (Fig. 1). Because EBT3 film is sensitive to UV and general light, the samples were stored in a dark room except during irradiation and the acquisition of SEM images.

Figure 1. Depiction of Gafchromic EBT3 film’s surface structure. (a) Laminated EBT3 film. (b) Unlaminated EBT3 film.

2. Irradiation

All EBT3 film samples were cut into 1×1 cm² pieces and placed on a 5-cm thick solid water phantom and covered with oil paper to prepare for irradiation and prevent scratching of the active layer’s surface. A 1.5-cm thick solid water phantom was placed on the EBT3 film’s surface. A 100-cm source-to-surface distance and 10×10 cm² irradiation field were established. After dose calidation for 6 MV photon rays was performed according to the American Association of Physicians in Medicine’s Task Group 51 protocol, the samples were irradiated using 6 MV photon rays produced by the TrueBeam STx (Varian Medical Systems, Palo Alto, CA, USA) [20] to ensure accurate dose delivery. Radiation doses of 0, 5, 10, 15, and 20 Gy were administered over five sets, using one set of four films. The irradiated films were placed in a dark room for more than 24 hours for self-development.

3. SEM scanning

To obtain the SEM images, we coated the 4-nm-thick EBT3 surfaces with iridium to amplify the electron signals. SEM SNE-4500M (SEC, Suwon, Korea) was for SEM image acquisition in environments of 10 kV and 100 μA. Four measurement points were designated uniformly on each film sample and each measurement point was examined under 100× and 3,000× magnification.

4. Calculation of textural parameters

Texture analysis used a gray level co-occurrence matrix (GLCM) to determine between-pixel relationships. When determining GLCM, the between-pixel distance “d” was changed from 1 to 30, and we used between-pixel angles “θ” of 0°, 45°, 90°, and 135°. The quantization level (N_g) was 64, the level most frequently used for texture analysis. Using GLCM, the four most representative textural features—Entropy, Contrast, Energy, and Homogeneity—were calculated according to the following equations:

here, i and j refer to the horizontal and vertical axis coordinates of GLCM, respectively. The spatial dependence matrix of GLCM is denoted by p(i, j), indicating the probability of a between-pixel relationship for each (i, j).

Four textural features were obtained to determine the distances and angles between pixels. The average between-pixel angle was chosen as the representative value. In other words, for each surface image, we obtained representative values for each of the four textural analysis parameters.

5. Determination of dose calibration curves and statistical analysis

We performed a linear regression analysis to determine the correlation between the textural features and delivered doses and calculated maximum R-squared values to evaluate the correlations between doses and each textural parameter.

Fig. 2 depicts a surface EBT3 film image obtained using SEM. The images that were magnified 3,000× represented the active layer’s structural molecules more clearly than those magnified 100×. As the delivered dose increased, the active layer’s surface became smoother through molecular polymerization.

Figure 2. An EBT3 film surface image obtained using scanning electron microscopic (SEM). Each measurement point is visualized under 100× and 3,000× magnification and absolute radiation doses of 0, 5, 10, 15, and 20 Gy.

Fig. 3 depicts the correlation between textural features and radiation doses as a function of pixel distance. Under 100× magnification, Contrast and Homogeneity had R-squared values of ≥0.5 for all between-pixel distances. For Homogeneity (d=1), the maximum R-squared value was 0.761. Using 3,000× magnification, the R-squared value was ≥0.5 for all between-pixel distances except for the correlation with a d=1–4. Correlation (d=9) produced an R-squared value of 0.964.

Figure 3. The maximum coefficient of determination value based on the between-pixel distance and the textural features obtained following irradiation and under (a) 100× and (b) 3,000× surface magnification.

Fig. 4 depicts the dose calibration curve obtained using correlation (d=9) under 3,000× magnification and with the highest maximum coefficient of determination value. When the radiation dose was calculated based on the dose correction curve of the correlation (d=9), the differences between the actual delivered radiation dose values and calculated dose values were 0.09, 1.96, −2.29, 0.17, and 0.08 Gy for 0, 5, 10, 15, and 20 Gy, respectively.

Figure 4. Correlation curve of the parameter of Correlation (d=9) for the EBT3 film’s surface under 3,000× magnification and various radiation doses.

We performed dosimetry on EBT3 film samples using SEM and texture analysis of the samples’ surfaces. SEM was used to visualize the microscopic molecular structure of the film’s active layer. Texture analysis was used to quantify changes in molecular structure caused by radiation. Changes to the film’s surface texture were highly correlated with delivered radiation doses.

Shin et al. [19] reported gradual surface smoothing with delivery of increasingly high radiation doses. These results are consistent with our findings. Arjomandy et al. [21] found that delivering 2–10 Gy doses to EBT2 film surfaces reduced surface roughness at higher doses. Microscopically, the reaction between the radiation particles and the active layer’s chemical molecules elicits polymerization. Volotskova et al. found changes to molecular structure after EBT3 and EBT-XD films were irradiated and then analyzed using SEM surface imaging [22]. The authors found no reduction in surface roughness, unlike our results. Polymerization caused the active surface layer’s needle-shaped structure to elongate as a result of radiation exposure [22].

Because we only used SEM to analyze the EBT3 surface images, surface impurities or defects could have affected our results. However, we avoided using film samples with impurities or defects at the measurement points when we acquired our SEM images. Additionally, SEM uses high-voltage, accelerated electrons that might deliver excess radiation to an object’s surface, even if very few electrons were accelerated onto the film’s active layer. Importantly, the active layer’s ability to absorb radiation doses can affect the delivered radiation dose. When SEM images are obtained through electrons accelerated at low voltages, the image noise increases, potentially affecting measurement accuracy. We acquired our SEM images using an optimized level of 10 kV, exposing the film to as few electron rays as possible. Consequently, our maximum exposure time was 10 seconds or less.

We investigated the feasibility of measuring EBT3 film microscopically. We visually determined the effects of radiation by observing changes to the active layer’s molecular structure following irradiation. These changes were quantified using texture analysis and were highly correlated with the delivered dose. These results confirm that irradiated film radiation levels can be accurately determined using a radiation dose calibration curve.

We confirmed a high correlation between EBT3 film surface texture and delivered radiation doses. Our methods allowed us to analyze microscopic changes the film’s active layer that can subsequently be used for dosimetry.

This study was supported by a VHS Medical Center Research Grant, Republic of Korea (grant number: VHSMC 22016).

The author has nothing to disclose.

The data that support the findings of this study are available on request from the corresponding author.