Pulmonary Nodule Recognition Based on Multiple Kernel Learning Support Vector Machine-PSO.

摘要: Pulmonary nodule recognition is the core module of lung CAD. The Support Vector Machine (SVM) algorithm has been widely used in pulmonary recognition, and Multiple Kernel Learning (MKL-SVM) achieved good results therein. Based on grid search, however, MKL-SVM needs long optimization time course parameter optimization; also its identification accuracy depends fineness grid. In paper, swarm intelligence introduced Particle Swarm Optimization (PSO) combined with to be MKL-SVM-PSO so as realize global parameters rapidly. order obtain optimal solution, different inertia weights such constant weight, linear nonlinear weight are applied nodules recognition. experimental show that model training proposed only 1/7 search algorithm, achieving better effect. Moreover, Euclidean norm normalized error vector measure proximity between average fitness curve after convergence. Through statistical analysis 20 times operation inertial weights, it can seen dynamic superior algorithm. shorter; value convergence much closer value, which than weight. Besides, a verified.