Arterial Wall Image Texture Feature Extraction And Classification Algorithm Research And Implementation

With the rapid development of computer technology, image processing, medicalimage processing and recognition based on computer processing technique is one ofthe hot spots in the field of digital image processing, the use of computer technologyfor medical image processing and recognition and diagnosis, developed accuracy highand rapid diagnosis of pathological image classification as a medical technologyprovides a powerful tool. In this paper, the arterial wall ultrasound images of mice forthe study, the study analyzed the texture image feature extraction, focusing on the BPneural network, K-nearest neighbor algorithm, Bayesian algorithms, decision tree,support vector machine classification algorithm were achieved, and the use ofmultiple classifier based decision level fusion algorithm, greatly improving therecognition rate of the image, on atherosclerosis have better warning effect.The principal research of this article is organized as follows:(1) In this study, the mice for the study, the50mice were divided into twogroups, labeled as apoe group and normal group, each with25, then apoe group ofmice fed a high lipid content of food, normal group feeding ordinary food, threeweeks after the two groups of mice were placed in Visual Sonics Vevo2100real-timemolecular imaging ultrasound system and carotid artery wall extract image data atdifferent voltage, current, frequency conditions. In reading and summarizing a largenumber of domestic and foreign literature and information, find a single classificationalgorithm unavoidable defects, in order to compensate for this shortcoming, whilepreventing doctors in the clinical detection of atherosclerosis in patients withmisdiagnosis, based on five kinds of paper, classifier decision level fusion algorithmwarning system framework and design simulation warning system.(2) In order to follow-up with a better mouse arterial wall image featureextraction effect, we need to filter out the noise interference, the completion of theoriginal image pre-processing simulation experiments, including filtering,thresholding image sharpening and histogram equalization based on the provisionsdealing with the gray of image enhancement methods. (3) In order to make each classifier training and recognition results, the manualinterception of the arterial wall experimental image, the image texture featureextraction, each with10images for the standard, each figure is calculated20characteristic parameters exist for the vector and a large number of statistical texturefeatures of the image, and finally the texture parameters normalized, As a training andtest samples for each classifier.(4) In this paper, the BP neural network were realized, K-nearest neighboralgorithm, Bayesian algorithms, decision tree, support vector machine classifieralgorithm to complete the identification of the arterial wall of the image, and a largenumber of statistical image recognition rate, experimental results show that thisseveral algorithms have a defect can not be avoided, the final decision-making stagethe largest voting method using fusion algorithm, this algorithm is to make up fortheir lack of complementarity between the algorithm makes better improve therecognition rate of the arterial wall image Therefore, this system has a better warningeffect.