Analyzing anatomical structures of branching topology through elastic matching of tree encodings

Automated techniques for the analysis of anatomical tree-shape structures provide insight into the association between topology and underlying pathology. In this paper we present a framework to analyze tree topologies using representative encodings of parent-child node relationships and sequence matching techniques. Our methodology introduces the notion of elastic matching between tree encodings which results in revealing corresponding groups of nodes between the compared trees. Experimental evaluation using a dataset of breast ductal trees regarding radiological findings of breast cancer demonstrates the effectiveness of the proposed method which outperforms state-of-the-art tree characterization techniques in terms of sensitivity, specificity and accuracy.