ESTRO 2024 - Abstract Book

S3135

Physics - Autosegmentation

ESTRO 2024

[1] X. Ye et al., “Comprehensive and clinically accurate head and neck cancer organs-at-risk delineation on a multi-institutional study,” Nature Communications 2022 13:1, vol. 13, no. 1, pp. 1–15, Oct. 2022, doi: 10.1038/s41467 022-33178-z.

[2] V. Oreiller et al., “Head and neck tumor segmentation in PET/CT: The HECKTOR challenge,” Med Image Anal, vol. 77, p. 102336, Apr. 2022, doi: https://doi.org/10.1016/J.MEDIA.2021.102336.

[3]

“SegRap Challenge,” 2023. [Online]: https://segrap2023.grand-challenge.org/

[4] F. Isensee, P. F. Jaeger, S. A. A. Kohl, J. Petersen, and K. H. Maier-Hein, “nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation,” Nature Methods 2020 18:2, vol. 18, no. 2, pp. 203–211, Dec. 2020, doi: https://doi.org/10.1038/s41592-020-01008-z.

2794

Digital Poster

New DL approach for dysphagia analysis in videofluoroscopy swallowing studies in HN cancer patients

Lucía Cubero 1,2 , Christophe Tessier 1 , Joël Castelli 1 , Renaud de Crevoisier 1 , Franck Jégoux 3 , Javier Pascau 2,4 , Oscar Acosta 1 1 Université de Rennes, CLCC Eugène Marquis, Inserm, LTSI – UMR 1099, Rennes, France. 2 Universidad Carlos III de Madrid, Departamento de Bioingeniería, Madrid, Spain. 3 PU-PH ORL, Service ORL et Chirurgie Maxillo-faciale, CHU Pontchaillou, Rennes, France. 4 Hospital Gregorio Marañón, Instituto de Investigación Sanitaria Gregorio Marañón, Madrid, Spain

Purpose/Objective:

Dysphagia, defined as difficulty in swallowing, is one of the main secondary effects of radiation therapy in head and neck cancer (HNC) patients. Videofluoroscopy swallowing studies (VFSS) are considered the gold standard for dysphagia evaluation. A VFSS is obtained by acquiring continuous X-ray images during swallowing, providing structural and functional information about the process. However, assessing dysphagia from VFSS is subjective and time-consuming. To overcome this limitation, we present a deep learning (DL)-based approach to automatically label ten regions of interest for dysphagia evaluation in VFSS. Moreover, we extract seven different dynamic parameters from the VFSS to describe and compare the swallowing process between healthy and pathological subjects.

Material/Methods:

This study included 15 patients diagnosed with head and neck cancer of the oropharynx and apparent signs of dysphagia, and 15 subjects with a healthy larynx and no signs of dysphagia. VFSS of the 30 participants, acquired between 2016 and 2022 in the same medical center at a frame rate of either four or seven frames per second with an image size of 1112x890 pixels, were available. The first objective was to automatically label ten regions of interest in the VFSS. With this aim, an expert speech pathologist manually labeled six anatomical landmarks and segmented four different regions of interest throughout an isolated swallowing event in each VFSS. The training database included 750 frames extracted randomly from the