Carotid Atherosclerotic Plaque Diagnosis Based On Data Mining Technology

In our country, the incidence of carotid artery disease is as high as 70% in the crowd over 60 years old. And the age has advanced from 60 to 45. So it has become one of diseases which threaten to the health of the elderly. It has the tendency of young. Currently the diagnosis of the disease is mainly done by doctors, which is not only inefficient, but also difficult for the human brain to analyze a large number of high-dimensional data. To solve these problems, we use Weka to analyze the hemodynamic information from 311 cases of electronic medical records of carotid atherosclerosis patients from Xi’an Tangdu Hospital. We use BP algorithm, C4.5 algorithm and support vector machine to establish carotid atherosclerotic plaque diagnostic classifier. By comparing three algorithms, support vector machine classifier has optimum performance. We optimize support vector machine classifier. Its performance is greatly improved. Doctors can make scientific diagnosis decision of carotid atherosclerotic plaque. The main works are summarized as follows:1) To fully understand the clinical diagnostic indexes of carotid atherosclerosis plaque. Through the analysis, we can know that there is a close relationship between the carotid artery hemodynamic information and the formation of carotid atherosclerosis plaque, extracting the portion of the data and taking all the attributes in carotid artery hemodynamic information.2) Preprocessing carotid artery hemodynamics information data set. Firstly, the data format is adjusted to ARPF file. Secondly, we normalize these attributes in the hemodynamic information according to the different needs of data mining algorithms. Finally, we use different attribute selection methods to filter the most relevant attributes in order to filter out subset of the most relevant attributes to establish classifier.3) Using BP algorithm, C4.5 algorithm and support vector machine to build classifiers. Different algorithms have different parameters, so different methods are used to choose different parameters.The principle of selection is to select the parameters combination within a certain range that makes the accuracy of classifiers be optimal. We use this combination to model.4) Evaluating the three classifiers and optimizing the best performance classifier in these three. From the modeling time, the interpretation, the error and the cost we analyze the three classifiers and select the best performance classifier among them. By comparing, the support vector machine classifier has the best performance. But there is still a lot of room for improvement, so the AdaBoost algorithm of ensemble learning is used to optimize it.The most relevant hemodynamic attributes that format the carotid atherosclerotic plaque are obtained by the experimental results. Among the three kinds of carotid atherosclerosis plaque classifiers, the support vector machine classifier has the best performance. Through the optimization, the carotid atherosclerosis plaque diagnostic classifier’s performance has been increased. The final classification accuracy is 75.5%. The optimized classifier can help doctors make scientific decisions on carotid atherosclerotic plaque diagnosis. Although it can not be diagnosed only by hemodynamic information, it can not be ignored.