Yang, Yuechen.; Guo, Junlin.; Zhu, Yanfan.; Yue, Jialin.; Zhu, Junchao.; Wang, Yu.; Zhao, Shilin.; Yang, Haichun.; Guo, Xingyi.; Tanevski, Jovan.; Barisoni, Laura.; Rosenberg, Avi Z.; Huo, Yuankai. (2026).Ìý.ÌýIS and T International Symposium on Electronic Imaging Science and Technology, 38(11), 1981–1989.Ìý
High-throughput analysis of whole slide images, which are extremely large microscope images of tissue, is opening new ways to study how tissues are organized and to discover biomarkers, meaning measurable signs of disease. But it has been hard to capture how individual tissue structures relate to one another and how those spatial patterns connect to clinical outcomes. To address this, the authors developed HistoWAS, short for Histology-Wide Association Study, a computational framework that links tissue organization to patient outcomes. HistoWAS combines two parts: first, it measures tissue structure using standard features plus 30 additional spatial and topological features, which describe how cells and structures are arranged and clustered, based on methods originally used in geographic information systems; second, it uses an association-analysis approach similar to Phenome-Wide Association Studies, which systematically tests many features one by one for links to clinical variables while correcting for multiple statistical comparisons. As a proof of concept, the team applied HistoWAS to 385 tissue slides stained with PAS, a common staining method that highlights kidney structures, from 206 participants in the Kidney Precision Medicine Project. They analyzed 102 total features and released both the code and data publicly.
