Visualizing Complex Patterns in the Spread of Head and Neck Cancers

Abstract

A predictable pattern in the spread of head and neck cancers has been derived from multidimensional scal ing (MDS). MDS places points in an abstract space such that distances between points are proportional to differences in the occurrence of tumor in different clus ters of lymph nodes. The purpose of the present work is to visualize the predictions of this model. Ellipses are expanded in the abstract space to represent the spreading cancer. Lymph nodes on an image of a real head are made to change color as their corresponding points are encountered by the expanding ellipse in the abstract space. Correlations between the actual and predicted patterns of metastases are high (r = .963 to .999), and experienced surgeons believe these depic tions of metastatic spread are accurate. Such transfor mations of pathological processes between real and abstract spaces could assist surgeons in locating tu mors and in other medical decisions.