Concavity and convexity analysis of mammographic masses via an iterative boundary segmentation algorithm

The problem of computer-aided classification of benign and malignant breast masses as seen on mammograms using morphological features is addressed. We propose methods of shape analysis treating the object's boundary in terms of local details. We use an iterative boundary segmentation method to separate major portions of the boundary and label them as concave or convex segments. In order to analyze the shape information localized in each segment, we compute features through polygonal modeling of the tumor boundaries. The features developed in the present study when used in combination with the global shape feature of compactness resulted in a classification accuracy of 81% with a database of 28 benign masses and 25 malignant tumors.