This work demonstrated that the Pinnacle3 Auto-Planning system was able to produce comparable quality VMAT plans by different delivery systems, and was therefore able to successfully handle the geometric and dosimetric variations between NPC and BC sites. The significant differences in modulation complexity of VMAT plans between Trilogy and Synergy were found.
The dosimetric studies of auto plan showed excellent target volume coverage and OARs sparing to various target paradigms [13–16]. Hazell et al.  compared auto plans with manual head and neck cancer IMRT treatment plans, they found that target coverages in the auto plans were similar to manual plans, but automatically generated plans had less irradiation of healthy tissue. In contrast to other studies, the primary goal in this study was to generate “same” plans.
There were several major differences between Varian Trilogy and Elekta Synergy systems used for delivering VMAT plans. To reduce the variability in the quality of treatment plans, dose objectives in Auto-Planning can be automatically generated to drive the optimization of the new plan. What is noteworthy is that it required further user intervention to manually set optimization objectives at the end of the auto optimization process. Allowing for a manual post optimization, the coverage for PTVs and dose constrains for OARs were as same as possible.
In summary, as shown in Fig. 1, the Pinnacle Auto-Planning was able to produce VMAT comparable plans using Trilogy and Synergy delivery systems for more complex cases (breast and NPC tumor regions). The only differences occurred about brainstem and spinal cord because the concern was mostly focused to constraint the max dose of these two organs during optimization. However, the values were still within the tolerance range. Both delivery techniques, if considered from a clinical perspective appear to be equivalent.
Regarding the treatment MU, the corresponding data listed in Table 1 suggested that although the tumor volumes were similar for NPC and BC, the BC plans had lower MUs than NPC plans by applying approximately tangential arcs. The Synergy required less MU to treat the same volume tumor, even there was not drastic variation in the MU value among the two systems for either site. This difference of MU was largely attributed to the fact that large area segments were more often in Synergy plans than in Trilogy plans. Similar study in terms of IMRT plans was reported by Qi et al. . As observed from their published study, high MU efficiency was observed in direct aperture optimization (DAO) plans than direct machine parameter optimization (DMPO) plans because large area segments were used often in DAO plans.
Both Synergy and Trilogy plans had clinically acceptable plan quality, but we observed that the SA was different. The breast plans used larger SA than NPC plans. Synergy used larger SA than Trilogy for the same site. The Synergy plans were manifested in the larger segment area and lower MU relative to the Trilogy plans. Furthermore, VMAT auto plans were more sensitive to MLC errors .
To assess the beam complexity, as stated in the results, we found significant differences. In this study, the MCS of NPC and BC plans were 0.14 and 0.42 for Trilogy,0.12 and 0.3 for Synergy. Dhanabalan et al.  reported that average MCS for the head and neck VMAT plans were 0.2224. The MCS for auto plans were smaller than manual plans, which suggested the beams were more complex. The autoplans were more modulated as illustrated by Hansen et al. , which might cause the slightly lower pass rate of 97.7% in the ArcCHECK measurements. The question of the relationship of plan complexity and gamma index analysis of delivery accuracy was not included in this study.
However, much effort has been devoted to analysis that correlation using different patient-specific QA phantom. Crowe et al.  found that the ‘small aperture score’ provided threshold values which successfully distinguished deliverable treatment plans from plans that did not pass QA using MapCheck2 diode array. Li et al.  found that the planning parameters such as the average leaf travel and average field indicating a correlation between the plan complexity and the passing rate using ArcCHECK diode array. Dhanabalan et al.  studied the correlation results between the MCS and gamma analysis results indicate that MCS of a plan has a weak correlation with the planar as well as volumetric gamma analysis passing rates using Octavius4D phantom. Knowledge of this relationship will be further accumulated using our MatrixX and ArcCHECK phantom. This study demonstrates that treatment techniques differ in terms of treatment MU and MCS. Mcgarry et al.  suggested MCS was most suitable for inclusion within the cost function to limit complexity in IMRT optimization. VMAT plans,which are less complex, also have higher probabilities of yielding accurate dosimetric results.