Orthogonal features based classification of microcalcification in mammogram using Jacobi moments

Abstract

Objective: Breast calcifications can be present in mammograms which are one of the most important risk indicators of breast cancer. Digital mammography is a reliable tool to detect breast cancer at the early stage with no symptoms. The objective of this research work is to classify the microcalcification patterns into benign and malignant. Methods: In this research, a novel approach is proposed for classification of microcalcifications based on shift-invariant transform, Jacobi moments and Support Vector Machines (SVM). The Jacobi moments are used to extract orthogonal features from the location of microcalcifications based on orthogonal or weighted polynomial basis which uses recurrence relation to avoid the loss of precisions. The Jacobi features will be given as input to SVM classifier for classifying the mammogram images into normal and abnormal. The abnormality will be further classified into benign or malignant. Findings: The validity of the proposed approach is evaluated using MIAS database. In the process of mammogram enhancement, the experimental results shows that shift-invariant transform achieves better results than contourlet transform. The classification rate of proposed system is 98.65% for normal, 95.80% for benign and 93.35% for malignant accuracy. The performance of the proposed approaches measured by sensitivity, specificity and confusion matrix. The measurement of performance is 0.84 sensitivity and 0.95 specificity for stage1 and 0.85 sensitivity and 0.83 specificity for stage 2. Application/Improvement: The results show that our proposed approach gives high level of accuracy for classification of microcalcification. This approach is very useful to avoid unnecessary biopsy.