ESTRO 2024 - Abstract Book

S3098

Physics - Autosegmentation

ESTRO 2024

The model achieves prostate and urethra segmentations with DSC scores of 0.907 ± 0.027 and 0.427 ± 0.097 when compared to Reader 1, and 0.911 ± 0.023 and 0.452 ± 0.141 when compared to Reader 2. These results are comparable to the inter-reader variability of 0.913 ± 0.027 and 0.349 ± 0.143, for the same structures.

The segmentations of the training and test datasets will be made available with a later publication.

Conclusion:

In this work we show an automatic method capable of segmenting the prostate, urethra, and prostatic zones at a comparable level as two experienced radiologists.

Keywords: Automatic Segmentations, Urethra, Prostatic Zones

References:

[1] – Groen, V.H. et al. (2022) ‘Urethral and bladder dose–effect relations for late genitourinary toxicity following external beam radiotherapy for prostate cancer in the flame trial’, Radiotherapy and Oncology, 167, pp. 127–132. doi:10.1016/j.radonc.2021.12.027.

[2] – American College of Radiology. Prostate Imaging – Reporting and Data System. 2019. Version 2.1.

[3] – Geert Litjens, Oscar Debats, Jelle Barentsz, Nico Karssemeijer, and Henkjan Huisman. "ProstateX Challenge data", The Cancer Imaging Archive (2017). DOI: 10.7937/K9TCIA.2017.MURS5CL

[4] – Isensee, F., Jaeger, P. F., Kohl, S. A., Petersen, J., & Maier-Hein, K. H. (2021). nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nature methods, 18(2), 203-211.

[5] – Dice, L.R. (1945). Ecology, 26(3).