Learning from deep learning and pathomics

Abstract

The suffix -omics has been applied to a variety of approaches with the overarching aim to identify, describe, and quantify processes that contribute to the form and function of cells and tissues. Historically, histopathologic approaches have relied on assessment of lesions to describe a pattern of injury and when possible, give a specific etiologic diagnosis. Broad pathologic entities have been split into smaller groups to provide more individualized diagnostic, prognostic, and etiologic information as a step towards precision medicine. More recently, pathomics has emerged as a complement to traditional histopathology, applying deep learning algorithms to digital histopathology images for detailed quantitative mining of visible or subvisual features that correlate with specific diagnoses, disease trajectories, or shared injury patterns. In a recent publication in Nature Communications, Hölscher and colleagues describe this exciting new approach of pathomics, with the potential to provide insights into kidney disease diagnosis, prognosis, and mechanisms ( 1 Hölscher D.L. Bouteldja N. Joodaki M. et al. Next-Generation Morphometry for pathomics-data mining in histopathology. Nat Commun. 2023; 14: 470 Crossref PubMed Scopus (1) Google Scholar ).