Atherosclerotic Plaque Segmentation in Carotid Ultrasound Image using Hybrid-A-UNet Deep Learning Model

Cardiovascular diseases (CVD) and strokes are major health concerns worldwide, and early detection of these conditions can prevent mortality and expensive surgeries. However, conventional methods of severity detection and prediction of CVD and stroke in their early stages are not sufficiently accurate or automated. Therefore, there is a need for Artificial Intelligence (AI) based methods that can accurately and automatically detect these conditions.In this study, a deep learning (DL) model attention-channel-based UNet (A-UNet) was proposed to identify carotid plaques in images of the common (CCA). The accuracy of the model was determined using a cross-validation K5 protocol. The performance of the A-UNet model was compared with that of the UNet model.The plaque segmentation results showed that the A-UNet model outperformed the UNet model. The correlation coefficient (CC) value for A-UNet was 0.96, compared to 0.93 for the UNet model. Similarly, the area under the curve (AUC) value for A-UNet was 0.97, compared to 0.964 for the UNet model. The study concludes that the A-UNet model is useful for segmenting very high risk plaque images that are difficult to diagnose using other methods.