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 Twitter

Kidney Week

Abstract: TH-PO567

Machine-Learning-Quantified Lupus Nephritis Histological Features Correlate With NIH Activity and Chronicity Index Subscores

Session Information

  • Pathology and Lab Medicine
    November 03, 2022 | Location: Exhibit Hall, Orange County Convention Center‚ West Building
    Abstract Time: 10:00 AM - 12:00 PM

Category: Pathology and Lab Medicine

  • 1700 Pathology and Lab Medicine

Authors

  • Moll, Solange, Universite de Geneve, Geneve, GE, Switzerland
  • Austin, Cary D., Genentech Inc, South San Francisco, California, United States
  • Toth, Balazs, Genentech Inc, South San Francisco, California, United States
  • Kos, Filip, PathAI, Boston, Massachusetts, United States
  • Hennek, Stephanie, PathAI, Boston, Massachusetts, United States
  • Jayson, Christina, PathAI, Boston, Massachusetts, United States
  • Baker, Brian Hartman, PathAI, Boston, Massachusetts, United States
  • Trotter, Benjamin, PathAI, Boston, Massachusetts, United States
  • Lincoln, Webster U., Genentech Inc, South San Francisco, California, United States
  • Brosnan-Cashman, Jacqueline A., PathAI, Boston, Massachusetts, United States
  • Schindler, Thomas, F Hoffmann-La Roche AG, Basel, Basel-Stadt, Switzerland
  • Resnick, Murray, PathAI, Boston, Massachusetts, United States
  • Chang, Anthony, University of Chicago Pritzker School of Medicine, Chicago, Illinois, United States
  • Wapinski, Ilan, PathAI, Boston, Massachusetts, United States
  • Montalto, Michael, PathAI, Boston, Massachusetts, United States
  • Garg, Jay P., Genentech Inc, South San Francisco, California, United States
  • Tew, Wei, Genentech Inc, South San Francisco, California, United States
  • Prunotto, Marco, F Hoffmann-La Roche AG, Basel, Basel-Stadt, Switzerland
Background

Histologic evaluation of renal biopsies is necessary for lupus nephritis (LN) diagnosis and treatment decisions; however, interobserver variability and poor quantitation limit the utility of histology-based metrics for precision medicine. To mitigate these challenges, we developed ML-based models to quantify histologic features in LN.

Methods

374 hematoxylin and eosin (H&E)-stained whole-slide images (WSI) of non-LN kidney and LN biopsies were obtained, mainly from a LN cohort at the University of Geneva and a clinical trial of obinutuzumab (OBZ) in proliferative LN (NCT02550652). WSI were split into training (286; 76%) and validation (88; 24%) sets. Expert pathologist annotations trained deep convolutional neural networks, yielding two distinct segmentation models covering anatomic regions and histopathological features. Model performance was tested on a held-out set of 94 WSI. For each model, 20-30 image frames were annotated by 3-5 pathologists to derive ground truth consensus. Whole slide predictions on 73 baseline cases from the OBZ trial were correlated to manual revised NIH LN activity and chronicity index (CI) subscores and kidney function metrics using Spearman method.

Results

The model performed comparably to pathologists on both WSI and frames identifying tissue regions (e.g. cortex, F1model=0.78; F1pathologist=0.75) and tissue features (e.g. interstitial inflammation, F1model=0.68; F1pathologist=0.60). ML-quantified interstitial inflammation and sclerotic glomeruli regions correlated with the NIH activity index interstitial inflammation (r=0.638; p<0.0001) and CI glomerulosclerosis subscores (r=0.702; p<0.0001), respectively, as well as with eGFR, creatinine, and UPCR (r=0.32 - 0.47; p<0.01).

Conclusion

We developed ML models that quantify histologic features on LN H&E biopsies, revealing significant correlations with NIH disease index subscores and kidney function metrics. The findings demonstrate the feasibility of ML for quantifying LN histologic features. The utility of this approach in predicting treatment response is being evaluated.

Funding

  • Commercial Support