Assessing the impact of ultrasound image standardization in deep learning-based segmentation of carotid plaque types

Georgia D. Liapi, Christos P. Loizou, Constantinos S. Pattichis, Marios S. Pattichis, Andrew Nicolaides, Maura Griffin, Efthyvoulos Kyriacou

Abstract: Highlights•Image standardization is highly influential in the deep learning-based workflows for atherosclerotic plaque type segmentation in carotid ultrasound images.•Image standardization is a premise for valid statistical analysis, regardless of the deep learning models’ generalization capacity.•To the best of our knowledge, this is the first time the known CFPNet-M, a small deep learning segmentation model is trained and evaluated with carotid B-mode ultrasound images.•In this study we demonstrated a Deep-learning-based approach for automated segmentation of all carotid plaque types (I to V), such as heavily or uniformly echolucent plaques.•Predominantly or uniformly echolucent plaques or plaques with juxtaluminal black areas on the lumen side are immensely missing or are underrepresented in relevant published studies, although these carotid plaques are often detected in clinical routine.