Multicluster class-based classification for the diagnosis of suspicious areas in digital mammograms

This book chapter presents a multi-cluster class based classification approach for the classification of suspicious areas extracted from digital mammograms into benign and malignant classes. The approach creates multiple clusters and selects strong clusters for each class. The created strong clusters are used to form subclasses within benign and malignant classes and training of a classifier. The creation of strong multiple clusters during the classification process can improve the accuracy of the classification system. The experiments using multi-cluster class based approach and a standard classifier with a single cluster per class have been conducted on a benchmark database of digital mammograms. The results have shown that the multi-cluster class based approach makes a significant impact on improving the classification accuracy.