Carotid Intima-media Thickness Measurement In Ultrasound Image Based On SVM

The common carotid artery(CCA)intima media thickness(IMT)is a widely accepted and important marker of early atherosclerosis.Ultrasound imaging is widely applied in clinical environment for visualization of carotid artery,in which the traditional measurement of IMT is based on manual tracing.The major drawback of traditional measurement is highly user dependent and time consuming.The purpose of this study is to propose a fully automatic segmentation(AS)method for the IMT measurement to solve the drawbacks of traditional measurement.Since intima media complex segmentation can be considered as a classification of pixels,a support vector machine is carried out to solve the issue.The approach can be divided into three stages:(1)the region of interest(ROI)detection;(2)extract the statistical features extraction and classification;(3)post-processing.In the stage of ROI detection,firstly,K-means is used to cluster the pixels based on the intensity,considering the results of clustering,the arguments of linearly adjusting and threshold are chosen to normalize image and transform it into binary image.The detection of region of interest(ROI)is operated based on the binary image.In the stage of classification,a support vector machine(SVM)which previously trained by statistical intensity features is used to classify the pixels in ROI into “IMT Boundary pixels” and “Non-IMT boundary pixels”.Then at the last stage,a heuristic searching method of column-by-column was processed to debug the result of classification.The thesis improved the approach then,the new approach classify the pixels into LII and MAI directly.After this,a set of 80 longitudinal ultrasound(US)images of the CCA were used to test the proposed approaches.The performance of AS was assessed by comparing with the manual segmentation ground truth(GT).The mean total CPU time per image spent is 0.88 s.The proposed fully automatic method is reliable for the IMT measurement and shows high efficiency and accuracy,which satisfies the basic demand of clinical application.