ESTRO 2024 - Abstract Book

S4474

Physics - Machine learning models and clinical applications

ESTRO 2024

Figure 1. An example of an equatorial CT slice (a) and the calculated (b) and predicted (c) dose distributions with the MultiResUnet. The horizontal lines that appear on the CT are the positions of line profiles (d-f). The relative difference (g) and gamma evaluation (h) are also shown.

Conclusion:

Deep learning models show great potential in predicting dose distributions for IORT, especially the MultiResUnet model. The use of a segmentation mask enhances the results of the training. The data augmentation also has an impact on the results. However, the results are still acceptable without it.

Keywords: IORT, Deep Learning, Dose prediction

References:

[1] P Ibáñez, et al. Med Phys 48 (12), 8089