Classification Of Stroke Based On Microwave Experiment Platform

Stroke is a kind of disease which is not only easy to get ill but also the mortality is high. There are about 1.3 million people died of a stroke each year in our country, which means stroke has become the main cause of residents′ death in our country. At present, it is more important to prevent it rather than treat it. Medical studies have demonstrated that stroke can increase the rate of cure and reduce the damage in the brain by timely detection and discovery. Because microwave nondestructive testing has the advantages of safety, low cost, easy to survey, so more and more researchers pay attention to it and would like to use it to detect a stroke.Research of using Microwave to detect stroke by the difference of the dielectric properties. Nowadays, microwave stroke detection methods are mostly based on microwave imaging. It is needed to use ultra-wideband microwave signal scanning and the ultra-wideband(UWB) antenna. In this paper, the machine learning method is applied in the judgment of the stroke. The microwave scans signals of the brain were acquired. And the support vector machine classifier was used to predict stroke.Firstly, this paper studies the theoretical basis of microwave stroke detection, including microwave principle, brain tissue structure and dielectric properties, and support vector machine based knowledge. Secondly, on the basis of theory research is proposed in the framework of the classification algorithm, microwave signal acquisition, preprocessing, selection of the training set and test set, is used to optimize the support vector machine(SVM) classifier that of SVM kernel function optimization, and puts forward the three optimization methods are particle swarm optimization, genetic algorithm optimization, grid search method of optimization, in order to establish the best classification model.We have to design the experimental platform of microwave detection in for stroke. The design of ultra-wideband antenna module is mainly included, and its return loss is tested to verify its availability. According to the dielectric properties of the brain tissue of the stroke of the head model design, there are two kinds of brain model of liquid and solid brain model. A method for microwave measurement using a two-port vector network analyzer(VNA). In this paper the S parameters, impedance parameters, admittance parameters have been researched. By these modules to form a complete stroke microwave detection experiment system, the data model of measuring brain stroke.In this paper, the microwave experiment platform has been used for detecting the stroke. The main related parameters were set up, such as band setting, bandwidth setting, sampling point setting, power setting and so on. Then, the accuracy of the support vector machine classifier was verified. The optimization time, the optimal parameter and the classification accuracy of the three optimization methods were compared. At the same time the effect of the blood clot position, size, dielectric properties for classification have been verified. Finally, it was proved SVM classifier for stroke prejudgment can be used in microwave detection.Finally, describing the direction of development and great advantage of microwave human body biological detection in the future, and the way to combine with the methods of machine learning. What’s more it will play an important role in the field of stroke diagnosis.