Automatic classification of cervical cells using a binary tree classifier

Abstract

A k-means clustering algorithm for designing binary tree classifiers is introduced for the classification of cervical cells. At each nonterminal node of the designed binary tree classifier, two sets of effective feature are selected: one is based on the Bhattacharyya distance, a measure of separability between two classes; the other is based on the merits of classification accuracy. The classification result has shown the effectiveness of the features and the binary tree classifier used.