Abstract

Nasopharynx segmentation in MR images based on one-class immune feature weighted support vector machines

Author(s): Lei Guo, Ming Hu, Lei Zhao, Ying Li, Guizhi Xu

In the brain Magnetic Resonance (MR) images, the nasopharynx part is highly irregular. It is difficult to accurately segment this part. Owing to its powerful capacity in solving non-linearity problems, One-class Support Vector Machine (SVM) method has been widely used as a segmentation tool. However, the conventional one-class SVMs assume that each feature of the samples has the same importance degree for the segmentation result, which is not necessarily true in real applications. In addition, oneclass SVM parameters also affect the segmentation result. In this study, ImmuneAlgorithm(IA)was introduced in searching for the optimal feature weights and the parameters simultaneously.An Immune FeatureWeighted SVM (IFWSVM) method was used to segment the nasopharynx in MR images. Theoretical analysis and experimental results showed that the IFWSVM had better performance than the conventional methods.