Abstract: We present a geometry-driven graph learning framework for histopathology that converts Tissue Microarray (TMA) images into Delaunay triangulation graphs of superpixels. Unlike ...