AUTHOR=Cheon Wonjoong , Ahn Sang Hee , Jeong Seonghoon , Lee Se Byeong , Shin Dongho , Lim Young Kyung , Jeong Jong Hwi , Youn Sang Hee , Lee Sung Uk , Moon Sung Ho , Kim Tae Hyun , Kim Haksoo 

TITLE=Beam Angle Optimization for Double-Scattering Proton Delivery Technique Using an Eclipse Application Programming Interface and Convolutional Neural Network

JOURNAL=Frontiers in Oncology

VOLUME=Volume 11 - 2021

YEAR=2021

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2021.707464

DOI=10.3389/fonc.2021.707464

ISSN=2234-943X

ABSTRACT=<p>To automatically identify optimal beam angles for proton therapy configured with the double-scattering delivery technique, a beam angle optimization method based on a convolutional neural network (BAODS-Net) is proposed. Fifty liver plans were used for training in BAODS-Net. To generate a sequence of input data, 25 rays on the eye view of the beam were determined per angle. Each ray collects nine features, including the normalized Hounsfield unit and the position information of eight structures per 2° of gantry angle. The outputs are a set of beam angle ranking scores (<italic>S</italic><sub>beam</sub>) ranging from 0° to 359°, with a step size of 1°. Based on these input and output designs, BAODS-Net consists of eight convolution layers and four fully connected layers. To evaluate the plan qualities of deep-learning, equi-spaced, and clinical plans, we compared the performances of three types of loss functions and performed <italic>K</italic>-fold cross-validation (<italic>K</italic> = 5). For statistical analysis, the volumes V<sub>27Gy</sub> and V<sub>30Gy</sub> as well as the mean, minimum, and maximum doses were calculated for organs-at-risk by using a paired-samples <italic>t</italic>-test. As a result, smooth-L1 loss showed the best optimization performance. At the end of the training procedure, the mean squared errors between the reference and predicted <italic>S</italic><sub>beam</sub> were 0.031, 0.011, and 0.004 for L1, L2, and smooth-L1 loss, respectively. In terms of the plan quality, statistically, Plan<sub>BAO</sub> has no significant difference from Plan<sub>Clinic</sub> (<italic>P</italic> &gt;.05). In our test, a deep-learning based beam angle optimization method for proton double-scattering treatments was developed and verified. Using Eclipse API and BAODS-Net, a plan with clinically acceptable quality was created within 5 min.</p>