Introduction
In external radiotherapy for regions such as the head and neck, where air and bone heterogeneity are high, the uncertainty of the calculated dose distribution depends on the accuracy of the calculation algorithm as one factor [1–5]. To reduce this uncertainty, it is crucial that an accurate calculation algorithm is selected.
The Eclipse treatment planning system (Varian Medical Systems, Palo Alto, CA, United States) provides two algorithms to calculate dose: the anisotropic analytical algorithm (AAA) and Acuros XB (AXB). AAA applies simplified density scaling of the dose kernel to heterogeneous media using convolution and superposition [6]. AXB uses the Linear Boltzmann Transport Equation (LBTE) to calculate the behavior of radiation particles as they travel through and interact with matter [7, 8]. In air, AAA overestimates doses, whereas AXB doses are in good agreement with those of Monte Carlo simulations and actual measurements [9–12]. Additionally, in high-density metal structures, Pawalowski et al. reported that AAA overestimates doses by more than 10%, AXB doses are in good agreement with Monte Carlo simulations and actual measurements [12]. Therefore, AXB is recommended for accurate dose calculation of regions with heterogeneity, such as air and bone in the head and neck region [13–15]. However, Israngkul et al. reported that the dose delivered to 95% volume of targets (D95%) calculated by the AXB were reduced by approximately 28% compared with the AAA in the case of a small target and large air content for volumetric modulated arc therapy (VMAT) of pituitary carcinoma [14]. Therefore, using the same dose prescription method (i.e. normalizing at D95% ) may result in a dose escalation when the dose calculation algorithm for VMAT plan is changed from AAA to AXB for head and neck cancer (HNC). When changing the dose calculation algorithm used in clinical to AXB, it is important to ensure that the doses delivered to targets using AXB do not greatly deviate from using AAA. Because the doses to target reported in previous papers were calculated based on AAA or similar algorithms. For example, RTOG-0522, one of the evidence-based medicines for HNC, adopted the dose prescription for D95% of targets using AAA or similar algorithms [16]. In this study, we clarified the dose difference between AAA and AXB caused by the target’s air content in the VMAT planning of HNC. We also investigated that the dose parameters of targets using AXB were equivalent to the D95% of targets using AAA, even as the target’s air content increased.
Materials and methods
Virtual phantom with air structure for VMAT planning of the whole neck
For the head and neck region, a water-equivalent cylindrical virtual phantom was created in the treatment planning system, Eclipse ver. 15.6, with a diameter of 26 cm, length of 50 cm, and slice thickness of 2 mm. We transferred the plan and the structure, including the planning target volume (PTV) contoured by the radiation oncologist, of a patient who underwent whole neck VMAT at our institution to the center of this cylindrical phantom. Whole neck VMAT in conjunction with the simultaneous integrated boost (SIB) method for the prescribed PTV dose of 70 Gy (PTV70Gy) was delivered via 2.12 Gy per fraction. 95% volume of PTV70Gy was covered by the 100% of the prescribed dose. The cylindrical air structure (–1000 Hounsfield unit: HU) of 50 cm in length was placed at the center of the cylinder phantom [14, 17], and the diameter of the air structure was incremented by 1 cm from 0 to 6 cm (Fig. 1). The target’s air content was varied by changing the air in PTV70Gy divided by PTV70Gy in percent (air/PTV) as follows: 0%, 1.6%, 6.4%, 13.9%, 23.4%, 34.3%, and 46.2%. The isocenter was set at the center of PTV70Gy. The gantry rotation angle was set from 181° to 179° with collimator angles of 20° and 340°. The multi-leaf collimator (MLC) (Millennium 120) in TrueBeam (Varian Medical Systems, Palo Alto CA, United States) was employed. The X-ray energy was 6 MV, and the maximum dose rate was 600 monitor unit (MU)/min. Dose-to-medium was used for AXB dose calculations, and the grid size for dose calculations was 2 mm [18]. The accuracy of AAA and AXB dose calculations at our institution had been adjusted to agree with measured values within 1–2% in representative field sizes at the commissioning.
Dose difference with increasing target’s air content
For each target’s air content, the dose differences between AAA and AXB were obtained in the cylindrical virtual phantom (Section Virtual phantom with air structure for VMAT planning of the whole neck). The whole neck VMAT plans transferred to the cylindrical virtual phantom were recalculated by AAA and AXB with the same MU and MLC motions [13] while changing the air/PTV (0%, 1.6%, 6.4%, 13.9%, 23.4%, 34.3%, and 46.2%). D98%, D95%, D50%, and D2% of PTV70Gy were compared between AXB and AAA. Dx is the dose covering x% of the volume.
Differences of dose indices between AAA and AXB
For each target’s air content, the dose differences between D95% of PTV70Gy using AAA and D95%, D50% and D100%-air/PTV of PTV70Gy using AXB were obtained in the cylindrical virtual phantom (see Virtual phantom with air structure for VMAT planning of the whole neck). The whole neck VMAT plans transferred to the cylindrical virtual phantom were calculated by the D95% prescription using AXB for the PTV70Gy, and recalculated by AAA with the same MLC motions [13] while changing the air/PTV (0%, 1.6%, 6.4%, 13.9%, 23.4%, 34.3%, and 46.2%). For each target’s air content, the MUs calculated by the D95% prescription using AXB were 829.6, 830.3, 838.2, 880.0, 931.4, 969.2 and 1001.1 MU. D100%-air/PTV indicates doses of 100% of PTV volume minus the air/PTV. D100%-air/PTV was developed from the results of Section Dose difference with increasing target's air content, where AAA produced greater values than AXB with D98%, D95%, D50%, and D2% for PTV70Gy, and the difference increased as the target’s air content increased.
Dose escalation or underdose with air for each prescription method: clinical cases
Three clinical VMAT plans with the target’s air contents of 5.61% (whole neck), 17.02% (larynx), and 28.19% (nasopharynx) were used to evaluate MU for each dose prescription calculated by AAA and AXB. There were differences at the dose to water and the dose to medium between AAA and AXB, and the dose distributions could not be simply compared on the treatment planning systems [17]. Therefore, MUs were calculated for each prescription method using AAA and AXB under the same optimizations and MLC motions at each clinical case because MU differences were used to evaluate dose escalation or underdose for each dose prescription using AXB compared to the conventional AAA_D95%. The same MU leads to the same doses and, thus, a high MU indicates a dose escalation and a low MU indicates an underdose compared to the conventional AAA_D95% [13]. MUs were obtained with the D95% dose prescription method for PTV70Gy using AAA (AAA_D95%) and with the following dose prescription methods for PTV70Gy using AXB: (1) D95% dose prescription (AXB_D95%), (2) D50% dose prescription (AXB_D50%), and (3) the 100% dose prescription method to 100% of PTV volume minus the air/PTV (AXB_D100%-air/PTV). AXB_D100%-air/PTV was developed from the relationship between the target’s air content and dose calculated by the algorithms (see Differences of dose indices between AAA and AXB). The PTV volume in clinical cases were 419.1 cm3 (whole neck), 102.8 cm3 (larynx), and 192.6 cm3 (nasopharynx). The air region was automatically delineated from –1024 to –150 HU [19]. This study was conducted in accordance with the Declaration of Helsinki (as revised in 2013). We had received informed consent from all participants in the study. The study was approved by the ethics committee of the Kobe City Nishi-Kobe Medical Center (institutional review board number: 2022–23).
Results
Dose differences with increasing target’s air content
The dose differences between the two calculation algorithms as a function of the target’s air content are shown in Figure 2. AAA produced greater values than AXB with D98%, D95%, D50%, and D2% for PTV70Gy, and the difference increased as the target’s air content increased. The dose differences between AXB and AAA with D98%, D95%, D50%, and D2% for PTV70Gy were 3.08%–15.72%, 2.35%–13.92%, 0.63%–4.59%, and 0.14%–6.44% as the target’s air content increased from 5% to 30%. The dose differences were larger at near minimum doses (D98%) and D95% than at D50% and near maximum dose (D2%) with small increases in the target’s air content. Furthermore, the dose differences at D98% and D95% increased linearly, while those at D50% and D2% increased logarithmically with increasing target’s air content.
Differences of dose indices between AAA and AXB
The dose differences between D95% of PTV70Gy using AAA and D95%, D50% and D100%-air/PTV of PTV70Gy using AXB as a function of the target’s air content were shown in Figure 3. D95%, D50% and D100%-air/PTV were the dose parameters in section Dose escalation or underdose with air for each prescription method: clinical cases to use as the dose prescription volume. Compared to the D95% of PTV70Gy using AAA at the target’s air content of 5%, the corresponding D95% of PTV70Gy using AXB was –1.04%, the corresponding D50% of PTV70Gy using AXB was 6.11%, and the corresponding D100%–air/PTV of PTV70Gy using AXB was –1.12%. Compared to the D95% of PTV70Gy using AAA at the target’s air content of 30%, the corresponding D95% of PTV70Gy using AXB was –16.09%, the corresponding D50% of PTV70Gy using AXB was 3.58%, and the corresponding D100%-air/PTV of PTV70Gy using AXB was –0.97%.
The dose differences between D95% of PTV70Gy using AAA and D100%-air/PTV of PTV70Gy using AXB were within ±1%, when the target’s air content was ≥ 5%. In contrast, the D95% of PTV70Gy using AXB was lower than that of D95% of PTV70Gy using AAA caused by increasing target’s air content > 5%. Additionally, the D50% of PTV70Gy using AXB was higher than D95% of PTV70Gy using AAA, especially as the target’s air content decreased.
Dose escalation or underdose with air for each prescription method: clinical cases
The MUs of AAA_D95%, AXB_D95%, AXB_D50%, and AXB_D100%-air/PTV for three clinical cases with different target’s air contents are shown in Table 1. In all cases, the MU of AXB_D95% was higher than that of AAA_D95%, and the difference increased as the target’s air content increased (maximum of 6.74% for the nasopharynx). In contrast, the MU of AXB_D50% was lower than that of AAA_D95%, and the difference increased as the target’s air content decreased (minimum of -1.69% for the whole neck). The MU of AXB_D100%-air/PTV was in agreement with that of AAA_D95% (±2%) in all cases, and it was unaffected by the target’s air content ≥ 5%.
|
|
Air/PTV (%) |
AAA |
AXB |
||
|
D95% |
D95% |
D50% |
D100%-air/PTV |
||
|
Whole neck |
5.61 |
654.78 MU |
668.09 MU (+2.03%) |
643.70 MU (–1.69%) |
666.59 MU (+1.80%) |
|
Larynx |
17.02 |
600.22 MU |
629.13 MU (+4.82%) |
596.16 MU (–0.68%) |
604.50 MU (+0.71%) |
|
Nasopharynx |
28.19 |
751.82 MU |
802.52 MU (+6.74%) |
746.12 MU (–0.76%) |
755.57 MU (+0.50%) |
Discussion
In this study, we clarified the dose difference between AAA and AXB with increasing target’s air content using a virtual phantom and clinical cases (Fig. 2, Fig. 3, and Tab. 1). We found that the D100%-air/PTV of PTV using AXB was comparable to the D95% of PTV using AAA when the target’s air content was ≥ 5%.
The dose difference increased as the target’s air content increased (Fig. 2) owing to the lack of scattering with AAA and occurrence of scattering with AXB [8, 20]. The correlation shown in Figure 2 can be used to estimate the dose difference between calculation algorithms. In particular, the differences at near minimum dose (D98%) and D95% were larger than those at D50% and near maximum dose (D2%), similar to the results reported by Israngkul et al. [14]. It means that the dose difference between the D95% prescription and others becomes more pronounced. Although the dose differences between AAA and AXB for IMRT and VMAT decrease because of dose compensation from out of field [9, 17], the dose differences between calculation algorithms are large when the target’s air content is large.
We determined the relationship between the dose differences of D95%, D50% and D100%-air/PTV of PTV70Gy using AXB compared to D95% of PTV70Gy using AAA and the target’s air content, as shown in Figure 3. The dose differences of PTV70Gy between D95% using AAA and D100%-air/PTV using AXB were within ±1%, even as the target’s air content was ≥ 5%. The relationship shown in Figure 3 can be used to estimate the occurrence of a dose escalation or underdose. The MUs of AAA_D95%, AXB_D95%, AXB_D50%, and AXB_D100%–air/PTV for three clinical cases with different target’s air contents were shown in Table 1. The MU differences between AXB_D100%–air/PTV and AAA_D95% were within ±2%, even as the target’s air content increased. These results, similar to those of Section Differences of dose indices between AAA and AXB, indicate that AXB_D100%–air/PTV can be adapted not only for virtual phantoms but also for clinical cases with target’s air content ≥ 5%.
The difference in dosage between AAA and AXB caused by an air cavity present within the PTV depends on various factors such as the size and distance of the cavity, field size, and the density of the surrounding medium [8, 13, 14, 21]. Rana et al. reported that smaller field sizes and longer cavity distances result in larger AAA and AXB dose differences in the air regions for beam from a gantry angle of 0 degree [21]. The dose index D100-air/PTV of PTV using AXB evaluates the dose of the region excluding air cavities from the PTV. As a result, it showed agreement within 1% of the D95% of PTV using AAA, even in cases of smaller field sizes for laryngeal cancer. Therefore, only the target’s air content needs to be checked and small target sizes, such as the larynx and nasopharynx, tend to have a larger target’s air content. It should be noted that it is not dose escalation compared to the conventional AAA_D95%. Other methods have been reported to improve the accuracy of dose calculation in HNC [22–24]. Asher et al. proposed removing the air from the PTV as the air cavity within the larynx presents a challenge for inverse planning software. The software tries to “push” dose into the air to achieve adequate target coverage, especially when using more complex dose calculation algorithms [22]. However, removing the air cavity may lead to underdosing of the treatment volumes at the interface between air and tissue [25], which may increase the risk of cancer recurrence in the normal tissue adjacent to the air-tissue interface [26]. Moreover, the procedure of omitting the air cavities from the targets would be burdensome to the planner.
When changing from the conventional method of AAA to AXB for targets with large air contents, such as the head and neck, evaluations similar to our study must be performed and discussed with the radiation oncologist, radiation technologists, and medical physicists to prevent a dose escalation or underdose. Although the AXB algorithm is known to be more accurate, this study compared the doses and MU calculated by AXB to those calculated by the AAA algorithm used as a reference. Our study focuses on the transitional period when the dose calculation algorithm was changed from AAA to AXB. In clinical practice, AXB would be the ideal algorithm to use as a reference.
Conclusions
In HNC VMAT, the dose difference between AAA and AXB increased as the target’s air content increased, and changing from conventional AAA_D95% to AXB_D95% had the risk of dose escalation. The D100%-air/PTV of PTV using AXB was comparable to the D95% of PTV using AAA when the target’s air content is ≥ 5%.
Acknowledgments
We thank Edanz (https://jp.edanz.com/ac) for editing a draft of this manuscript.
Conflicts of interest
None declared.
Funding
This work was supported partly by Japan Society for the Promotion of Science (JSPS) KAKENHI (grant number: 20K08093).
Author contribution
T.I., H.M. were associated with concept and design; T.I., Y.Y. and Y.S. took the measurements. T.I., H.K. and K.K. analyzed the data. T.I., H.M., M.I., H.D. and Y.N. prepared the manuscript. All authors read and approved the final manuscript.