Background and Aims: Atherosclerotic carotid plaques are an important cause of ischemic stroke. B-Mode ultrasound imaging is commonly used for detection and measurement of these plaques, yet image interpretation is difficult and subjective due to low contrast, shadows, speckle noise, and ultrasound artefacts. Computer-assisted methods can increase the accuracy and reduce the subjectivity in detection and measurement of carotid plaques. Methods: We propose an automated method using computer vision to detect and measure atherosclerotic plaques in B-Mode vascular ultrasound images. In this method, we first pre-process and normalize the images and track the outer vessel wall automatically. Segmentation algorithms are then used to separate the lumen from the plaque, both in still longitudinal images and in transverse cineclips. In cineclips, the method analyses all image slices automatically and detects atherosclerotic plaques in the entire carotid region, from common carotid artery to bifurcation and internal carotid artery. Results: Preliminary results are obtained by analysing carotid ultrasound scans of 58 patients, approximately 2000 images that include longitudinal stills and transverse cineclips. Manual segmentations by an independent blinded expert agree well in comparison with the results of automated segmentation from the proposed method. Conclusions: Preliminary results show that the proposed method can improve the accuracy of reporting carotid ultrasound scan in a clinical setting and significantly reduce the reporting time. Additionally, the method provides a platform for scaling of machine learning efforts on quantification of plaque vulnerability. These methods may improve risk stratification and treatment planning of carotid disease.
No datasets are available for this submission.
No license information is available for this submission.