ESTRO 2024 - Abstract Book

S4473

Physics - Machine learning models and clinical applications

ESTRO 2024

Purpose/Objective:

Intraoperative radiation therapy (IORT) requires fast and accurate treatment planning, as these procedures involve tumor removal surgeries preceding irradiation which may alter the original treatment plan. While Monte Carlo (MC) calculations are considered the gold standard, they require significant computational resources and time. The use of deep learning algorithms using modern GPUs may provide a faster alternative. In this study, we trained and tested two deep learning models to predict 3D dose distributions for INTRABEAM spherical and needle applicators.

Material/Methods:

Two deep learning models, U-Net and MultiResUnet, were trained to predict dose distributions of the spherical and needle applicators of the INTRABEAM system. A cohort of 800 paired CTs and dose distributions (720 training, 80 validation) of partial breast irradiations with a spherical applicator and 41 (36 training, 5 validation) of kyphoplasty using the needle applicator were used separately for training and validation. Doses were calculated using a fast MC algorithm [1], serving as the ground truth. The training set was normalized to improve dose prediction. Additional tissue material information was incorporated into the input channel as a segmentation mask, and an image augmentation technique was introduced to produce an image by taking different parts from other images in the dataset. Gamma analysis was performed to compare the predicted dose to the simulated dose.

Results:

The MultiResUnet outperformed the Unet with a 97.6% average gamma passing rate over the 96.5% of the Unet for a 5%/0.5 mm gamma criterion for the spherical applicators. A similar performance was observed for the needle applicator. An example of a predicted dose distribution with the MultiResUnet is shown in figure 1. The inference time of the network is a factor of 875 times faster than the dose calculation code used.