Title: A hybrid kernel principal component analysis and support vector machine model for analysing sonographic features of parotid glands in Sjogren's syndrome

Authors: Ping-Feng Pai, Ya-Hsin Chang, Ming-Fu Hsu, Jachih Fu, Hsin-Hua Chen

Addresses: Department of Information Management, National Chi Nan University, 470 University Road, Puli, Nantou, Taiwan. ' Department of Information Management, National Chi Nan University, 470 University Road, Puli, Nantou, Taiwan. ' Department of International Business Studies, National Chi Nan University, 470 University Road, Puli, Nantou, Taiwan. ' Department of Industrial Engineering and Management, National Yunlin University of Science and Technology, 123 University Road, Section 3, Douliou, Yunlin, Taiwan. ' Division of Allergy Immunology Rheumatology, Taichung Veterans General Hospital, No. 160, Sec. 3, Chung-Kang Rd., Taichung, Taiwan

Abstract: Sjogren|s syndrome is a chronic inflammatory process that primarily involves the exocrine glands. Clinical manifestations range from autoimmune exocrinopathy to extraglandular involvement affecting the lungs, kidneys, blood vessels and muscles; Sjogren|s syndrome can occur along with, or accompany, other autoimmune diseases. Moreover, diagnosing Sjogren|s syndrome saves lives. The term support vector machine (SVM) refers to an emerging machine learning technique based on statistical learning theory which can solve classification problems using small sampling, non-linearity and high dimensions. However, both dimension reduction and parameter determination greatly influence the performance of the SVM technique. Therefore, this study develops a SVM with an immune algorithm (IA) and kernel principal component analysis (KSVMIA) model to diagnose Sjogren|s syndrome effectively and efficiently. In the proposed KSVMIA model, the kernel-based principal component analysis (KPCA) technique is used to reduce the dimension of Sjogren|s syndrome data; the IA is employed to determine the SVM models. Experimental results reveal that the developed model can classify Sjogren|s syndrome data obtained from ultrasound in terms of efficiency and accuracy, showing that the presented KSVMIA model is a promising alternative for diagnosing Sjogren|s syndrome.

Keywords: Sjogren|s syndrome; kernel PCA; principal component analysis; KPCA; support vector machines; SVM; immune algorithms; chronic inflammatory disease; exocrine glands; ultrasound; medical diagnosis; sonography; parotid glands.

DOI: 10.1504/IJMMNO.2011.037202

International Journal of Mathematical Modelling and Numerical Optimisation, 2011 Vol.2 No.1, pp.98 - 108

Published online: 30 Nov 2010 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article