Computer aided diagnosis system to detect breast cancer pathological lesions

Abstract

Breast cancer is one of the most frequent forms of women's cancer over the world. Studies of the World Health Organization (WHO) reported 1,151,298 cases in 2002. A reliable Computer-Aided-Diagnosis (CAD) system for automated detection/classification of pathological lesions is very useful and helpful, providing a valuable "second opinion" to medical personnel. In this work, we describe a new CAD system to diagnose six mammography pathological lesions classes (calcifications, well-defined/circumscribed masses, spiculated masses, ill-defined masses, architectural distortions and asymmetries) as benign or malignant tissues. Two different Artificial Neural Networks models: Feedforward Backpropagation and Generalized Regression were tested statistically with a precision of 94.0% and 80.0% of true positives, respectively. This CAD system was validated successfully on the MiniMammographic Image Analysis Society (MiniMIAS) database, with a dataset formed by 100 images. The CAD system performance shows similar or better classification results compared with others available methods. © 2008 Springer-Verlag Berlin Heidelberg.