In this study, we evaluated the performance of the PRO and PO algorithms for generating spine SABR VMAT plans by comparing the DV parameters of the target volume and surrounding normal tissues, total MU, and MCSv. To date, this study is the first attempt to assess the plan quality of VMATPRO and VMATPO in patients with cervical, thoracic, and lumbar spinal tumors. When comparing the DV parameters of the target volume and surrounding normal tissues, VMATPRO achieved better PTV coverage and dose uniformity while reducing the dose to the spinal cord or cauda equina and Ring1.5 cm than VMATPO. However, for VMATPRO, improvements in plan dosimetric quality can lead to increases in overall plan complexity and total MUs, which can compromise treatment deliverability and efficiency, respectively.
Similar studies have investigated the optimization algorithms in terms of dosimetric quality for various SABR treatment sites, such as the lungs and brain [13–15]. Visak et al. investigated the dosimetric quality of VMATPRO and VMATPO in 12 lung SABR patients with a single dose of 30 Gy [14]. They demonstrated that the PRO algorithm provided higher MUs and higher modulation of lung SABR VMAT plans, while the dose to normal tissues was reduced compared with the PO algorithm. They also reported that the PO algorithm increased the intermediate-dose spillage, which can result from more exposure to normal tissues [14]. In contrast, some institutions demonstrated that the plan quality between both algorithms was comparable with no statistical significance, although VMATPRO increased the total MUs and plan complexity [13, 15]. For the prostate, head and neck, and brain treatment sites, which do not involve SABR VMAT plans, there was an increase in the total MUs and the level of modulation when the PO algorithm was used, showing better OAR sparing and an opposite result to the SABR VMAT plans [12, 20]. Thus, the efficacy of the optimization algorithms may vary depending on the radiotherapy regimen or treatment site. Therefore, it is necessary to analyze the plan quality for each condition.
The PRO algorithm, which utilizes a point-cloud model, can have a high number of calculation points 1) inside small or narrow structures, such as lenses, optic nerves, and spinal cords, or 2) around the edge of irregular structures, such as vertebral bodies and head and neck nodes [21]. In addition, the grid size for the structure can be adjusted as much as the user wants, resulting in an increase in the calculation points inside the structure. Because this facilitates more degrees of freedom for the calculation grid size, a sophisticated modulation scheme of VMAT plans is possible using the PRO algorithm [12]. With these characteristics, the PRO algorithm can generate many small and irregular MLC openings compared with the PO algorithm. In contrast, the PO algorithm uses only one fixed grid size with a single matrix over the CT images during optimization, and the degrees of freedom for the calculation grid size are relatively small compared with the PRO algorithm. In addition, the PO algorithm tends to significantly remove small openings when compared with the PRO algorithm [13]. The MLC openings for randomly selected control points between VMATPRO and VMATPO for a representative spine SABR patient are shown in Fig. 3. It was observed that MLC shapes defined by the PRO algorithm were smaller and more irregular than those defined by the PO algorithm, and were associated with sparing critical normal organs during optimization. Spine SABR has the characteristic of having a long and narrow OAR including the spinal cord or cauda equina that must be protected within an irregular PTV of the vertebral body. Limiting the number of calculation points per volume leads to a potential loss of information that must be considered during optimization. To effectively reduce the dose to the spinal cord or cauda equina, which are widely recognized as critical organs, during optimization, the use of the PRO algorithm would be more advantageous for spine SABR VMAT plans because of the ability to generate small or irregular MLC shapes, and the greater degrees of freedom for the calculation grid size. For the dosimetric evaluation, the dose distributions in VMATPRO and VMATPO for spine SABR are shown in Fig. 1. A noticeable dose reduction in the spinal cord or cauda equina for VMATPRO was achieved compared with VMATPO that had good PTV coverage.
To improve the dosimetric plan quality, high modulation, implying complex MLC movements and the usage of small or irregular MLC apertures, is required; however, this leads to an increase in the number of total MUs and a decrease in plan delivery accuracy [12, 22]. Liu et al. compared PRO and PO algorithms in terms of plan quality and correlations between gamma passing rates and plan complexities for both lung SABR and brain SRS [13]. The criteria for the gamma analysis were 3%/3 mm and 2%/2 mm for lung SABR and 5%/1 mm and 3%/1 mm for brain SRS, with 10% as the threshold value. They reported less agreement between planned and delivered dose distributions when VMATPRO had higher MLC variability and total MUs. Although the overall gamma passing rates with all gamma criteria for VMATPRO decreased compared with those for VMATPO, the average gamma passing rates for lung SABR and brain SRS were above 90% and 95% under the criteria of 2%/2 mm and 3%/1 mm and 3%/3 mm and 5%/1 mm, respectively, and VMATPRO was considered clinically acceptable [13]. In this regard, our institution acquired the gamma passing rates of portal dosimetry with gamma criteria of 2%/1 mm, and all VMATPRO and VMATPO (> 90%) were found to be clinically acceptable. Additionally, our previous study investigated the correlation between gamma passing rates and the modulation degree of VMAT plans [23]. We utilized identical TrueBeam STx with a high-definition 120™ MLC and then generated 100 VMAT plans with various tumor sites, including the lung, spine, liver, brain, and head and neck, using the PRO algorithm. Measurements of the dose distributions for each VMAT plan were acquired using MapCHECK2™ and ArcCHECK™ (Sun Nuclear Corporation, Melbourne, FL, USA). As a result, the average gamma passing rates for all criteria were above 90%, which is regarded as clinically acceptable. It was found that there was less correlation between gamma passing rates with all criteria and MCSv, with no statistical significance, except for the correlation of the 3%/3 mm criterion with ArcCHECK™ (r = 0.210, p-value = 0.036) [23]. Thus, we can conclude that TrueBeam STx has a guaranteed performance with a high degree of agreement between the planned and actual delivered doses, regardless of plan complexity. For the TrueBeam STx used in this study, the modulation degree of the VMAT plans according to the optimization algorithm could be considered less important. Nevertheless, careful evaluation of its deliverability as well as that of other treatment machines is needed for the clinical implementation of the PRO algorithm.
Radiation myelopathy is a rare but catastrophic complication of radiation exposure to the spinal cord or cauda equine [1–3]. The RTOG 0631 guidelines recommend dose constraints to the spinal cord (D1.2 cc < 7 Gy, D0.35 cc < 10 Gy, and D0.035 cc < 14 Gy) and cauda equina (D5 cc < 14 Gy, and D0.035 cc < 16 Gy) for spine SABR [24]. However, according to the retrospective study by Sahgal et al., the maximum dose of the spinal cord in a single fraction was estimated, ranging from 9.20 Gy to 12.40 Gy which was associated with a 1–5% risk of radiation myelopathy [8, 25]. For the cervical and thoracic spine SABR in our study, the maximum dose to the spinal cord for VMATPRO (9.04 Gy) was approximately 2 Gy less than that for VMATPO (11.08 Gy), resulting in a 2% reduction in the risk of radiation myelopathy. The maximum dose of spinal cord PRV for VMATPO (13.83 Gy) exceeded 12.40 Gy while that for VMATPRO (12.25 Gy) did not. Since radiation myelopathy can lead to death, it is important to reduce the risk of this complication by sparing the spinal cord or cauda equina as much as possible. Therefore, patients treated with spine SABR may benefit from VMATPRO.