Modeling linear accelerator (Linac) beam data by implicit neural representation learning for commissioning and...

Area of Science:

• Medical Physics
• Computational Modeling
• Radiation Oncology

Background:

• Accurate radiation therapy relies on meticulous linear accelerator (Linac) beam data commissioning and quality assurance (QA).
• Current QA processes require extensive measurements across various field sizes, posing optimization challenges.
• Efficient data acquisition methods are crucial for maintaining high-quality medical physics practice.

Purpose of the Study:

• To develop an implicit neural representation (NeRP) model for linear accelerator (Linac) beam data.
• To assess the NeRP model's capability in predicting beam data from limited measurements.
• To evaluate the NeRP model's potential for verifying data collection accuracy and streamlining commissioning/QA procedures.

Main Methods:

• Utilized multilayer perceptron (MLP)-parameterized NeRP models to represent percentage depth dose (PDD) and profile data for 6 MV Linac beams.
• Embedded prior knowledge by training the NeRP on vendor-provided "golden" datasets.
• Trained the model on clinical data from one field size and predicted data for other sizes, validating against water tank measurements.
• Introduced intentional errors into datasets to test the NeRP model's error detection capabilities.

Main Results:

• NeRP-predicted beam data showed strong agreement with water tank measurements, achieving >95% Gamma passing rates (1%/1 mm) and <0.6% mean absolute errors.
• The model successfully identified erroneous beam data samples, indicated by inconsistent predictions and Gamma passing rates below 90% when trained on incorrect data.
• NeRP modeling demonstrated robustness in predicting accurate beam data even with sparse input measurements.

Conclusions:

• Established a novel NeRP-based technique for modeling and predicting Linac beam characteristics from sparse data.
• The NeRP model serves as a powerful tool for verifying the accuracy of beam data collection.
• This approach promises to simplify Linac commissioning and QA, reducing measurement requirements without sacrificing medical physics service quality.