For the Cone-PB 60° plan of the EVHP with a larger DCAI (such as 3° or 5°), the adjacent beams showed little overlaps at the sagittal measuring plane of the exit dose, and an independent circular or elliptical dose distribution area was predicted. In the meantime, the planned off-axis ratio at the central axis resembled a sine curve and was much lower than the measured. This is mainly because the Pinnacle3 SBRT dose algorithm (PB) computes the dose to the target based on interpolation of the absorbed dose as measured in water, without density correction. In the current study, the air gap between the phantom surface and SMC was about 8–10 cm, and this was taken as the equivalent tissue when the SBRT PB algorithm was adopted to calculate the exit-dose distribution. The calculated equivalent depth of beam penetration was thus far more significant than the actual depth, leading to a large deviation between the planned and measured results. The minimum RD of the CPD for Cone–PB 60° plans was > 16%, with a maximum of 50% and average of about 30%. The PB algorithm is therefore not suitable for calculating the exit dose and cannot be used with IVD verification.
The most commonly employed dose algorithms in TPS are the Acuros XB (AXB), analytical anisotropic algorithm (AAA), CCC, PB, and MC. The CCC and PB algorithms were usually used in early TPS, while the AAA and AXB algorithms are used in current Eclipse systems [9]. A previous report showed that the calculation results of the AXB and AAA algorithms were largely consistent with the MC simulation results in uniform media, while the difference between the MC simulation and AAA algorithm was more significant than that for the AXB algorithm in heterogeneous media [10]. Similarly, another study [11] compared and analyzed the dose-calculation accuracy of the AAA and AXB algorithms in lung SBRT planning, and found a larger systematic dose error with the AAA compared with the AXB algorithm. Another previous study [12] indicated that, although the AAA and AXB calculations differed in terms of the D95, Dmax, TV95, and HI of PTV in patients with left lung cancer, the magnitude of the differences was too small to have any clinically observable effect. Alghamdi et al. [13] compared the dose calculation accuracies of the AXB, AAA, and PB algorithms in four different-density media, and found that the AXB algorithm results were closest to the actual measured dose.
In the current study, the exit dose could be predicted precisely when the CCC algorithm was used to calculate the dose distribution of arc therapy in SBRT with a DCAI of 1°, with a maximum RD between the measured and calculated results of < 3%. The planned off-axis ratio curve on the central axis was consistent with the measurements, and its difference was minimal. The average passing rate of 2D gamma analysis (3%/3 mm) between the predicted exit doses and the in vivo measurements was 99.63% ± 0.55%, indicating that the CCC algorithm and 1° DCAI were suitable for calculating the exit dose and IVD verification.
Most routine radiotherapy plans are currently verified using the pretreatment dose in a homogeneous phantom. However, this can only detect the difference between the calculated and measured doses in the phantom, and fails to detect differences between the planned dose and the dose actually received by the patient. Some studies have shown minimal significance of pretreatment verification with a homogeneous phantom. However, even if the calculated results are consistent with the measurements in the phantom, this cannot guarantee that the dose received by the patient is consistent with the planned dose [14]. Many factors might affect the dose received by the patient, including setup error, body changes, gastrointestinal filling, respiratory movements, and the stability of the radiotherapy equipment. The true accuracy of the delivered dose can only be reflected by real-time online measurement of the received dose in the patient’s body during radiotherapy, or by measuring the exit dose outside the patient.
Mijnheer et al. [15] suggested that online IVD measurements should be conducted for patients receiving radical radiotherapy, to detect dose errors caused by various factors in the overall treatment procedure. Pretreatment dose verification alone fails to detect > 50% of severe dose errors during the actual therapy process. Furthermore, in vivo measurements will identify potential errors in dose calculation, data transfer, patient setup, dose delivery, and changes in patient anatomy, suggesting that all treatments with curative intent should be verified through IVD measurements combined with pretreatment checks [16]. Another study reported that IVD monitoring detected > 74% of errors caused by equipment faults or human error, suggesting that real-time IVD verification should be used for first radiotherapy treatments [17].
Compared with pretreatment dose verification with a homogeneous phantom, IVD verification can evaluate the results of plan implementation more directly and accurately. The main IVD verification methods currently include the following: (1) point dosimetry, including semiconductor detectors, optically stimulated luminescent dosimeters, thermoluminescence dosimeters (TLD), metal-oxide-semiconductor field-effect transistors (MOSFET), and plastic scintillation detectors; (2) transmission dose detection methods, such as Delta4 Discover (transmission type) and integrated quality control monitors, with the main advantages of high resolution, large-size field measurement, non-coplanar irradiation measurement, and real-time measurement during treatment, without affecting the treatment process; (3) log file analysis; and (4) EPID real-time monitoring [18]. In contrast, the main disadvantages of these IVD methods include lower spatial differentiation rate, fewer measurement points, and poor positioning accuracy of point dosimetry; lack of sensitivity to setup error and intrafraction motion for EPID; only detecting the incoming dose and lack of sensitivity to errors caused by patient body changes, respiratory movement and setup error for transmission dose detection; and log file analysis is not an independent measurement method. At least two monitoring methods are thus recommended [19]. In the present study, we only explored the impacts of three dose algorithms and DCAIs on IVD verification of small field partial arc SBRT with a phantom study. We aim to investigate full arc treatment of small fields, volumetric modulated arc therapy and intensity-modulated radiation therapy, and more algorithms in subsequent research. We will then carry out IVD verification of actual patients based on the phantom study.
The commonly used 2D and 3D dose verification systems, such as MatriXX, ArcCHECK, Delta4, and MapCHECK, cannot perform direct real-time IVD verification. Based on SMC and a self-made EDPD, we therefore investigated the impact of different dose algorithms and DCAIs on IVD verification of SBRT. The results indicated that, for IVD verification, the DCAI of SBRT arc therapy should be 1° when the CCC algorithm was used. In addition, the MC algorithm was the best of the three tested algorithms, followed by the CCC algorithm, both of which could calculate the exit dose accurately, while the PB algorithm was the worst and was not suitable for exit-dose calculation or IVD verification. SMC can thus be used to conduct IVD verification of SBRT partial arc therapy.