A single-pair antenna microwave medical detection system based on unsupervised feature learning

Abstract

The microwave medical detection method is an emerging non-invasive technology, which starts showing great potential in microwave biomedical applications. However, the practical application of it still faces challenges such as the detection equipment is complicated and difficult to control, and various interferences in the empirical situation. The difference between the microwave signal of healthy organs and that of the patients with stroke is sometimes too subtle to be detected when there are various noises within the detecting environment. This paper designed a single-pair antenna microwave medical detection system based on unsupervised feature learning for stroke detection. The system uses unsupervised feature learning, principal component analysis (PCA), to extract features, and then uses support vector machine (SVM) to classify whether there is a stroke. The use of a single-pair antenna greatly reduces the dimensionality of the sample features and also eliminates the interference between antenna arrays. This paper also optimized the detection position of the single-pair antenna. The performance of the detection system was verified by simulation and experiment. The results show that in the case of random interference, the detection system will also achieve better results, and when the antenna is placed in the left and right of the brain, the best performance will be achieved.