Abstract
This paper presents a classification method for normal and abnormal region of interests (ROls) in breast cancer using the chi-square distance on the texture features obtained from the local binary pattern (LBP) technique. The LBP features are calculated independently at different radii pixels in the varying neighborhood pixels using two different LBP variant i.e., uniform and non-uniform invariant rotational features. The chi-square distance is applied on the feature matrices to find the similar features belonging to a particular class. K-nearest-neighbors (KNN) classifier is used to classify features belonging to normal and abnormal class using the chi-square similarity distance matrices. The method achieved maximum accuracy of 95.3% with sensitivity and specificity values of 93.8% and 97.8% respectively for one of the best scenarios in non-uniform rotationally invariant features.