Ultrasound Image Analysis for Automated Classification of Atherosclerotic Plaque in the Human Carotid Artery
Cardiovascular diseases, caused by atherosclerosis, is one of the most common causes of death in the industrialized world. Two risk factors, that both can be assessed with ultrasound, are the Intima-Media Thickness (IMT) in the arterial wall and the composition of the atherosclerotic plaque. Plaques can be classied according to the subjective Gray-Weale scale. Its four classes are based on the echogenicity of the plaque, where class 1 is the most echolucent class and class 4 the most echogenic. In this thesis class 1 and 2 are merged into an echolucent class, and class 3 and 4 into an echogenic class. The motivation for this grouping is the fact that echolucent plaques are more unstable, and therefore more dangerous, than echogenic plaques. An automatic classication based on thresholding the plaque with an adaptive threshold has been developed. The threshold is a weighted sum of the echogenecity and the noise in the image, and it divides the plaque into black (echolucent) and white
(echogenic) regions. The classication rate on the training set was 95% (92 of 97) when using leave-one-out crossvalidation, while the performance on the validation set was 91% (81 of 89). This thesis also contains a description of a computer program for automatic measurements of the IMT. An automatic system decreases both the analysis time and the interobserver variability compared to a manual system.