A Robust Technique Using Pruned Association Rule to Diagnose Breast Cancer from Mammograms

Abstract

Many researches have been carried out to diagnose early breast cancer and many authors have suggested many methodologies but have not been widely accepted either by the radiologist or oncologist so far due to many reasons. A robust technique for breast tumor classification using pruned association rule PARM algorithm is presented in this paper believed to be acceptable by the physician. The method proposed makes use of association rule mining technique to classify the mammogram breast images into two categories namely benign and malign. It combines the GLCM features extracted from images and high level knowledge from specialists. The developed algorithm can assist the Radiologist for effectively classify with multiple keywords per image to improve the accuracy. Here using Association rule mining performs benign-malignant classification on region of interest (ROI) that contains mass. Texture is one of the important characteristic for features to classify both from film reading and machine learning. ARM exploits this important factor to classify the mass into benign or malignant. The statistical textural features used in characterizing the masses are mean, standard deviation, entropy, skewness, kurtosis and uniformity which are standard features for classification purpose to avoid under fitting or over fitting for the classifier. The main aim of the method is to increase the effectiveness and efficiency of the classification process with an objective to reduce the numbers of false-positive and increasing the sensitivity of malignancies. Association rule mining was proposed for classifying the marked regions into malignant as we as benign are 92.30% sensitivity and 95.23% specificity which is very much encouraging in compare to the radiologist's sensitivity 80%.