Comparative analysis of dose verification between computed tomography scan phantom and virtual digital phantom of Delta4

This study compared the dose verification results between plans created based on computed tomography (CT) scan and digital virtual phantom.


INTRODUCTION
Intensity-modulated radiation therapy (IMRT) planning is important in cancer therapy owing to the precision and complexity of treatment delivery. IMRT has more advantages than the traditional 3-D conformal radiotherapy, especially for patients with complex target tumors. 1 Therefore, the treatment plan needs an accurate dose verification system. A proper quality assurance (QA) procedure is essential to verify the accuracy of dose delivery before IMRT  can be carried out. There are many types of dosimeters on the market for ensuring QA, such as ion chambers, film, 2-D diodes, ion chamber matrix detectors, electronic portal imaging devices, gel dosimetry, and 3-D diode arrays. 2 Delta4 is one of the most widely used tools for 3-D dose verification. Many published studies showed that the Delta4 with the 3-D diode arrays is stable and reliable, making it an ideal tool for dose verification. [3][4][5][6] However, there are some limitations that should be considered when silicon diode detectors, such as Delta4, are used as radiation detectors. It might cause X-ray attenuation on the phantom, resulting in unwanted artifacts, inaccuracy in the low-energy photon fields, and slight distortion in images. 7 For clinical QA plans, the establishment of a verification plan is based on the digital phantom provided by the manufacturer, which is an ideal model and reduces the error introduced by the detectors. Alternatively, the verification plan is based on the reconstruction images acquired by computed tomography (CT) scanning. Although it is similar to the real state of Delta4, it might show some differences with the clinical measurement results.
Few studies have reported the effects of detector scattering on the Delta4 devices using low-dose CT scanning. The present study aimed to compare differences between CT-scan and digital virtual phantom (Uniform) in the Delta4 equipment that can influence the accuracy and reliability of dose verification, and provide an accurate QA plan.

QA planning and testing
Two QA plans were developed for each patient as follows: CT-scan group as the reference verification plan, which scanned and imported the Delta4 phantom into the planning system to establish the model; Uniform group as the test verification plan, which was generated by the artificial dataset of the uniform polymethyl methacrylate cylinder provided by the manufacturer (its relative density is 1.147). During the planning process, the outer contour of the two phantoms (named as Body) was manually drawn for statistical analysis of the dosimetric indices. Rigid registration was carried out to ensure that the outer contours of the two bodies were the same. The two sets of QA plans were transmitted to the accelerator control room and Delta4 analysis software for preparation. The investigated dose values were the minimum (D min ), maximum (D max ), mean (D mean ), and isocenter dose (D iso ) to the body obtained from dose-volume histograms generated by the treatment planning system.

Verification equipment
All QA plans were carried out on the Varian 600CD linear accelerator with 6-MV X-rays. Delta4, a 3-D dose verification system from ScandiDos, Uppsala, Sweden, was used as the verification system. It

Collection of verification data
The 3-D dose distribution was measured for each patient. The dose differences between the CT-scan group and the Uniform group were compared by the Delta4 analysis software. Gamma index analysis with global dose error normalization, such as distance to agreement (DTA), dose deviation (DD), and gamma passing rate, was carried out to compare the calculated and delivered dose distributions at 1%/1 mm, 2%/2 mm, and 3%/3 mm thresholds for each set of DD/DTA. The default dose threshold used in the analysis is 20%, and voxels below the threshold will not be considered in the gamma-pass rate analysis.

Statistical analysis
The Statistical Package of Social Sciences program version 19.0 (Armonk, NY: IBM Corp, USA) was used to carry out statistical analysis. A paired t-test was used to evaluate the association between the passing rates and the different modeling patterns. All statistical data are expressed as the mean ± standard deviation, and P < 0.05 is considered statistically significant.   Total n = 10. D iso , reference dosimeter of the isocenter point; D max , the maximum dose; D mean , the average dose; D min , the minimum dose.

RESULTS
of the transverse plane, sagittal plane, and coronal plane are shown from left to right. Although, in the dose visualization, the curves of the CT-scan group were close to those of the Uniform group, the CT-scan group had slightly elevated doses. The comparison between the two groups based on the dose-volume histogram of the body dose is shown in Figure 2a. As there were slight differences between the two groups, we subtracted the two curves to acquire detailed information on the distribution differences, as shown in Figure 2b. The Uniform group had a slightly larger irradiation volume in the low-dose area, but lower irradiation volume in the medium-dose area and the high-dose area than the CT-scan group, and this is consistent with the results in Table 1.
Meanwhile, we compared the dose differences between the dosevolume histogram and the measured dose distributions by the Delta4 matrix. The QA results were evaluated by analyzing the passing rate of DTA, DD, and gamma in the two groups. The results are shown in Figure 3 and Table 2. Gamma index analysis at thresholds 1%/1 mm, 2%/2 mm, and 3%/3 mm was carried out to compare the delivered dose distributions. The Uniform group had a higher gamma passing rate than the CT-scan group. There were significant differences between the two groups (P < 0.01). With an increase in the parameter value,

DISCUSSION
The characteristics of fast processing speed and high sensitivity of diodes make the measurement of external irradiation dose increasingly popular. Silicon semiconductor diodes, which are easy to measure, small in size, and have good mechanical durability, are commonly used as radiation detectors. A study showed that the p-type silicon diode is more conducive to radiation therapy than other types of diodes, 8 and there are 1069 p-type silicon semiconductor diode array detectors distributed on the Delta4 plates. Studies have shown that low-energy X-ray is more susceptible to harden in material attenuation than high-energy X-ray. 9 Errors caused by scattering might lead to CT artifacts and inaccurate CT numbers. 10 Artifacts can degrade CT simulation imaging and impair accurate delineation of tumors for radiation treatment planning purposes. 11 Even though many artifacts from the early days of CT are now substantially reduced, some artifacts remain, and new technologies might introduce incompletely characterized artifacts. Modern techniques for artifact reduction were studied and described. 12 In IMRT verification planning, it is necessary to establish a verification phantom on the planning system that simulates the CT scanning of the patient. Huang et al. used a human-like phantom reconstructed from a 3-D diode array to assess the patient's CT dose. 13 Density scaling artifacts were reported in the treatment planning system 14 : A small change in the density could imply a systematic error of 1-2% in the calculated dose. This systematic error could be significant for QA calculations carried out for non-unity density phantom materials.
Tani et al. proposed an optimum density scaling factor for phantom materials for a commercially available 3-D dose verification system (Delta4) to improve the accuracy of the calculated dose distributions in the phantom material. 15 Due to the characteristics of a semiconductor probe, the dose verification needs to be carefully designed. 16 Therefore, the artifacts of the Delta4 phantom need to be further optimized to ensure the accuracy of modeling verification in an effective way.  Therefore, the low-dose area was slightly higher, whereas the highdose area was somewhat lower in the Uniform group than in the CT-Scan group. The details are explained in Figure 4. To simplify the problem, we assumed a pencil beam on the phantom. Figure 4 shows a typical build-up curve for a 6-MV X-ray beam at a 0 • angle of incidence.
CT-Scan phantom has the same percentage depth dose as the uniform phantom before the beam is incident on the high-density area. The dose deposition of CT-Scan phantom will increase to a certain extent in high-density regions, resulting in a decrease in photon fluence with the exception of high-density area; therefore, the percentage depth dose curve after the high-density region will be lower than the Uniform phantom. Thus, the body of the percentage depth dose curve in the CTscan group will be low in the low-dose area, but elevated in the highdose area.
There were certain limitations to the present study. First, we used the Anisotropic Analytical Algorithm in the Treatment Planning System, which has some shortcomings dealing with particle transport in heterogeneous materials. Acuros XB algorithms could be a promising algorithm in the latest version of Eclipse for providing an accurate assurance. Second, the QA results did not consider factors, such as accelerator output and mechanical errors in the gamma analysis; therefore, the quantitative analysis was not accurate. Third, there might be some differences between the two phantoms compared with the actual measurement conditions. Therefore, it is suggested that a precise digital virtual phantom should be designed based on the actual physical parameters of the Delta4 phantom to further improve the reliability and accuracy of the verification results.
In the present study, we compared the verification results and quantitatively evaluated the differences between the CT-scan group and digital virtual phantom (Uniform) group in the QA of IMRT patients, and found that the volume of the CT-scan group was higher than that of the Uniform group in both the intermediate measurement area and the high-dose area, but a slightly lower volume in the low-dose area.
Data analysis under different parameter settings showed that the Uniform group had a significantly higher passing rate than the CT-scan group. Therefore, the use of digital virtual modules to generate validation plans is recommended in clinical practice, and the design of this type of phantom requires further improvement. However, there are still some limitations for plan verification based on uniform phantom, in which the fine structure in the Delta4 phantom was not taken into account. Therefore, in clinical practice, the design of the phantom should be further refined to ensure more accurate dose verification.