ASN's Mission

To create a world without kidney diseases, the ASN Alliance for Kidney Health elevates care by educating and informing, driving breakthroughs and innovation, and advocating for policies that create transformative changes in kidney medicine throughout the world.

learn more

Contact ASN

1401 H St, NW, Ste 900, Washington, DC 20005

email@asn-online.org

202-640-4660

The Latest on X

Kidney Week

Please note that you are viewing an archived section from 2023 and some content may be unavailable. To unlock all content for 2023, please visit the archives.

Abstract: FR-PO045

A Machine Learning Approach to Differentiating Between Typical Diabetic Nephropathy (DN) and Atypical DN with Podocytopathy in Periodic Acid-Schiff (PAS)-Stained Whole Slide Images of the Kidneys

Session Information

Category: Augmented Intelligence, Digital Health, and Data Science

  • 300 Augmented Intelligence, Digital Health, and Data Science

Authors

  • Tan, Myles Joshua Toledo, University of Florida Department of Electrical & Computer Engineering, Gainesville, Florida, United States
  • Rosenberg, Avi Z., Johns Hopkins University Department of Pathology, Baltimore, Maryland, United States
  • Paul, Anindya S., University of Florida Department of Medicine - Quantitative Health Section, Gainesville, Florida, United States
  • Han, Seung Seok, Seoul National University Department of Internal Medicine, Seoul, Korea (the Republic of)
  • Sarder, Pinaki, University of Florida Department of Medicine - Quantitative Health Section, Gainesville, Florida, United States
Background

Diabetic nephropathy (DN) is a leading cause of end-stage kidney disease. Other podocytopathies may occur in the setting of diabetes, leading to atypical presentations and potential responsiveness to immunosuppressants. Differentiation between typical and atypical cases is infeasible with light microscopy alone. We show that atypical DN with other podocytopathy can be differentiated from typical cases by uncovering clusters in a 2D projection of quantitative morphological features of glomeruli.

Methods

99 PAS-stained whole slide images of DN kidney biopsies (91 typical; 8 atypical) were fed into a deep learning model to segment glomeruli, from which 315 morphological features were quantified. Features were categorized according to compartment: global, PAS-positive (P), luminal (L), nuclear (N), both P & L, both P & N, and both L & N. They were further categorized according to type: morphological, size, color, radial, relative distance, textural. Points representing cases in a 315-dimensional hyperplane (1 dimension per feature) were projected onto 2D while preserving high-dimensional clusters using uniform manifold approximation and projection.

Results

Points representing typical DN cases clustered separately from atypical cases (top of projection). See Fig. 1.

Conclusion

Since some podocytopathies in the setting of DN respond to immunosuppressants, our approach could inform treatment in DN cases with suspicion of co-occurring podocytopathy. Given the infrequency of atypical cases and the overwhelming background diabetes, data are scarce. Future research may identify features that contribute to the clustering and to understand how morphology can inform diagnosis.

Funding

  • NIDDK Support