Performance analysis of different wavelet feature vectors in quantification of oral precancerous condition

Abstract

This paper presents an automatic method for classification of progressive stages of oral precancerous conditions like oral submucous fibrosis (OSF). The classifier used is a three-layered feed-forward neural network and the feature vector, is formed by calculating the wavelet coefficients. Four wavelet decomposition functions, namely GABOR, HAAR, DB2 and DB4 have been used to extract the feature vector set and their performance has been compared. The samples used are transmission electron microscopic (TEM) images of collagen fibers from oral subepithelial region of normal and OSF patients. The trained network could classify normal fibers from less advanced and advanced stages of OSF successfully.