Tree-based modelling for the classification of mammographic benign and malignant micro-calcification clusters

Computer Aided Detection (CAD) systems are being developed toassist radiologists in diagnosis. For breast cancer the emphasis is shifting fromdetection to classication of abnormalities. The presented work concentrateson the benign versus malignant classication of micro-calcication clusters,which are a specic type of mammographic abnormality associated with theearly development of breast cancer. After segmentation (automatic or man-ual), tree-based representations were used to distinguish between benign andmalignant clusters, which takes into account clinical criteria such as the num-ber of micro-calcications in the clusters and their distribution and is basedon the topology of the trees and the connectivity of the micro-calcications.The idea of using tree structure based on the distance of individual calcica-tions for the classication of benign and malignant micro-calcication clustersis novel and closely related to clinical perception. Tree structures used in thisstudy are distinct from decision trees classiers being used in many machinelearning approaches. Initial evaluation on the Digital Database for ScreeningMammography (DDSM) data shows promising results, with an accuracy equalto 91%, which is comparable to state of the art CAD systems and is in linewith clinical perception of the morphology and appearance of benign and ma-lignant micro-calcication clusters