Application of attributed graphs in diagnostic pathology.

OBJECTIVE: To compute attributed graphs based upon calculation of the minimum spanning tree (MST) for various applications in diagnostic lung pathology. STUDY
DESIGN: The study design included assistance in histologic diagnosis, confirmation of the diagnosis in single cases, measurement of texture alterations after induction chemotherapy, estimation of prognosis of operated-on lung cancer patients and analysis of lung cancer cells in association with differentiation markers. The histologic slides were Feulgen stained, and features of the integrated optical density (IOD) were associated with the nodes of the MST. The same procedure was applied to immunohistochemically and ligand histochemically stained slides by calculation of the mean staining intensity of the cytoplasm of tumor cells. A measure for structural entropy was introduced by computing the relative differences in distance and IOD between neighboring tumor cells in a 1/r2 field of force. In addition, the current of entropy was computed.
RESULTS: Structural entropy reflects alterations in regular textures; the current of entropy is an especially good prognostic parameter in lung cancer. In immunochemistry and ligand histochemistry, construction of the attributed MST permits detailed insight into locally different staining behavior of tumor cells and immunocompetent cells.
CONCLUSION: Attributed graphs contain important information that can be used for the estimation of survival or for confirmation of diagnostic entities, such as tumor cell types.