Convolutional neural network-based prediction of gamma pass rate in volumetric modulated arc therapy using 3d fluence maps and complexity metrics.

This study aims to develop and validate a deep learning-based model to accurately predict gamma pass rate (GPR) in treatment plans for volumetric modulated arc therapy (VMAT). A dataset comprising 360 VMAT treatment plans was utilized to train and validate the deep learning models. The 3D fluence maps, along with complexity metrics of the treatment planning, were extracted from each treatment plan as input features. Convolutional neural networks were utilized to establish the relationship between input features and GPR. The performance of the 3D fluence-based model (3D Fluence model) was compared with that of models that utilized the 2D Fluence model, dose distributions, and complexity metrics. The 3D Fluence model demonstrated superior performance, achieving a mean absolute error of 0.58 ± 0.21% for the 3%/2 mm gamma criterion and 1.09 ± 0.17% for the 2%/2 mm criterion. This model outperformed other models, including those that incorporate complexity metrics. The proposed 3D Fluence model utilizing deep learning offered a promising approach for accurately and efficiently predicting GPR in VMAT. This method can streamline the patient-specific quality assurance process, resulting in enhanced treatment quality and patient safety.