Research on Carotid Ultrasonic Image Detection Based on Convolutional Neural Network

Carotid ultrasound is a main and convenient method for diagnosing plaque, Therefore, accurately obtaining plaques information from ultrasound images is essential for further clinical diagnosis. Due to the interference of noise and the differences in artificial technical operations, the missed inspection of the plaques is likely to cause missed inspection. Therefore, a comparative experiment based on the constructing algorithm based on convolutional neural networks is performed to achieve more accurate detection and identification of cervical arterial tube cavity and plaques. First of all, we constructed the carotid artery data set, and then conducted the classification test of the carotid lumen and plaques through the YOLOv5 network based on migration and learning, and used the Faster R-CNN and SSD network for comparison experiments. Experiments show that the average accuracy obtained by YOLOv5 network reaches 0.928 when the IOU value is 0.5, and 0.659 when the IOU value is 0.75, and the average recall rate reaches 0.673, which are higher than the Faster R-CNN and SSD networks; The experiment shows that the average precision of the comprehensive comparison is also better than the other two comparison networks. At the same time, the model calculation speed meets the real-time needs. Therefore, the YOLOv5 network can improve the correctness and practical significance of the detection of the lumen and plaques in terms of carotid image detection.