Segmenting the carotid-artery wall in ultrasound image sequences with a dual-resolution U-net

Nolann Lainé (1, 2) , Guillaume Zahnd (3) , Herve Liebgott (2) , Maciej Orkisz (1)

1 MYRIAD - Modeling & analysis for medical imaging and Diagnosis
2 Imagerie Ultrasonore
3 Institute of Biological and Medical Imaging

Thickening of intima-media complex in the common carotid artery is a biomarker of atherosclerosis. To automatically measure this thickness, we propose a region-based segmentation method, involving a supervised deep-learning approach based on the dilated U-net architecture, named caroSegDeep. It was trained and evaluated using 5-fold cross-validation on two open-access databases containing a total of 2676 annotated images. Compared with the methods already evaluated on these databases, caroSegDeep established a new benchmark and achieved a mean absolute error twice smaller than the inter-observer variability.

Mots clés: Atherosclerosis; Carotid artery; Segmentation; Ultrasound; Deep learning

Domaines: Imagerie médicale; Intelligence artificielle [cs.AI]; Traitement du signal et de l'image [eess.SP]; Cardiologie et système cardiovasculaire

Nolann Lainé, Guillaume Zahnd, Herve Liebgott, Maciej Orkisz. Segmenting the carotid-artery wall in ultrasound image sequences with a dual-resolution U-net. 2022 IEEE International Ultrasonics Symposium (IUS), Oct 2022, Venise, Italy. pp.1-4, ⟨10.1109/IUS54386.2022.9957590⟩. ⟨hal-03897937⟩