Research on classification and recognition of the skin tumors by laser ultrasound using support vector machine based on particle swarm optimization

Abstract

As a common disease endangering human health, skin tumor causes millions of deaths every year. In order to effectively detect skin tumors, a laser ultrasonic detection method based on the PSO-SVM algorithm is proposed to identify and detect human skin tumors. The physical model of laser ultrasound in human skin tumors is established by FEM (finite element method), and the ultrasonic wave generated by laser in different tumor locations (the skin epidemis, the dermis and between the two layers) are analyzed in detail. The minimum value, the sample entropy, the permutation entropy and the shannon entropy of laser ultrasound are extracted as the sensitive features, and the three types of tumors are classified and identified by the SVM (support vector machine). The penalty factor C and kernel function parameters γ in SVM are optimized by the PSO(particle swarm optimization) algorithm. The simulation results show that the proposed PSO-SVM algorithm has a good classification effect on skin tumor prediction, which provides a new detection method for laser ultrasound to detect skin tumors.