68-4

65/136

(grant number JP15K08702).script.The average processing times were 27 s and 41 s for the FIF and FIF-4SF plans, respectively. A longer time was required to create the FIF-4SF plan than the FIF plan. The average time of the 3 D dose calculation for the FIF plans was 19 s and the calculation time of the 3D dose calculation for the FIF-4SF plan was 28 s. The difference in the processing time was mainly explained by the time for the 3D dose calculation.We used WBI cases to demonstrate the feasi-bility of the proposed method. Because the proposed method only requires a conventional plan without sub-beams as the input, it can be easily applied to other treatment sites, such as the whole breast12) and esophagus.Kim et al. reported on FIF planning automation for whole-breast irradiation12). They used the ESAPI in their automatic technique. In their method, infor-mation of the original plan was exported to an executable program outside the TPS in the DICOM and DICOM-RT formats. In their method, the MLC shapes and beam weights for the sub-beams were calculated and automatically imported into the TPS. Using their method, FIF plans with the same quality as those manually created were obtained. We used almost the same approach as that of Kim et al., but our procedures can be performed within the TPS. Furthermore, we implemented several steps to personalize the treatment plan to the patient.Yu et al. implemented an automated MLC shaping technique for WBI using deep learning. They could produce dose distribution almost equivalent to manually produced treatment plans. They used the two-opposing-lateral-field technique and obtained relatively high maximum dose of approximately 110%. By combining their technique with ours, a fully automated treatment plan for WBI can be realized.By automating the sub-beam shape and adjusting the weight, these processes can be standardized17). The manual shaping of sub-beams and adjustment of beam weights depend on the skill and prefer-ence of the treatment planners. The automation of FIF planning can reduce the variability originated from the skill and preference of the treatment plan-ners and the possibility of human error. However, because the transition from manual to automated planning can potentially lead to systematic errors that are difficult to detect18), the final verification must be performed by humans.A limitation of the present study is all manual FIF plans were created by one physicist. The processing time and the quality of the FIF plans probably depend on the experience of treatment planners. In this study, the similar plans were obtained for the manual and automatic FIF plan-ning. One reason of this similarity is probably that a single person made the manual FIF plans. The manual plans would diverge from the automatic FIF plans if multiple persons created the manual plans. However, the indication that the automatic FIF creation reduces the burden of the treatment planners is still valid because the sub-beam creation and weight adjustment are automatically performed.We developed a semiautomatic FIF planning method and implemented it in a TPS. By applying it to WBI, we confirmed that the semiautomatic FIF technique could reduce hotspot regions with a slight change in the PTV coverage compared with the original plan. When combined with a selection of an FIF scheme individualized to each patient, its performance was equal to or better than the manual FIF plan.Not applicable.This study was funded by JSPS KAKENHI HW, SS, TK, HN, TI, CK, KU, JT and KS partici-pated in the study design and data interpretation. Satoru Sugimoto performed manual treatment planning. KK participated in treatment planning data collection. HW wrote the draft manuscript. SS, TI, JT and KS reviewed and revised the draft and final manuscript. All authors read and approved the final manuscript.There are no conflicts of interest in relation to this study. 1) Smith T, Batko-Yovino S: The current status of pallia-385AcknowledgmentsFundingAuthor contributionsConflicts of interest statementReferences