Integrating PSONN and Boltzmann function for feature selection and classification of lymph nodes in ultrasound images

Lymph nodes (LNs), part of the lymphatic system, are important in the proper functioning of the immune system. LN metastasis is an important index for staging malignant tumors. The present study proposes a system that classifies lymph nodes according to pathological change from ultrasound (US) images. Features are selected and extracted from the US images. A feature selection method that integrates the particle swarm optimization neural network (PSONN) with the Boltzmann function is proposed to select significant features. A multi-class support vector machine (SVM) is adopted to classify diseases of the LN in the region of interests (ROIs) of US images into six categories. The experimental results show that the proposed approach decreases the number of selected features and that its classification is highly accurate.

► Seventy eight features are defined and the SVM is adopted to classify ultrasound images.
► A novel feature selection method integrates the PSONN with the Boltzmann function.
► The feature selection method is used to select important features of lymph nodes.