Classification of ductal tree structures in galactograms

The objective of this study is the classification of galactograms, medical images which depict the ductal tree of human breast, in order to provide insight into the relationship between tree topology and radiological findings regarding breast cancer. We present two different descriptors for the classification of the ductal trees; the tree asymmetry index and the maximum common skeleton, both of which quantify the similarity between tree topologies. Experimental results demonstrate the effectiveness of the proposed approach, which reaches a classification accuracy of 83%, and also indicate that our method can potentially aid in early breast cancer diagnosis.