Resumen
We have developed a graph- and attention-based A.I. model (PathoGraph) that predicts survival and identifies tissue areas of special diagnostic interest in whole slide images (WSI) of glioblastoma. CD276, an important immune checkpoint molecule, was confirmed to be a marker of malignant glioma cells/putative glioma stem cells (GSCs). The previously developed PathoFusion framework was used to selectively detect these cells. The graphs and attention scores that were obtained on the basis of this selection provide information that is hidden from conventional microscopic observation.