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 2019 and some content may be unavailable. To unlock all content for 2019, please visit the archives.

Abstract: SA-PO045

Machine Learning and Glomerular Remodeling

Session Information

Category: Bioengineering

  • 300 Bioengineering

Authors

  • Pellicano, Anthony, Angion Biomedica Corp., Uniondale, New York, United States
  • Goldberg, Itzhak D., Angion Biomedica Corp., Uniondale, New York, United States
  • Narayan, Prakash, Angion Biomedica Corp., Uniondale, New York, United States
Background

Glomerular hypertrophy is an early biomarker of ongoing renal disease and informs glomerulosclerosis and proteinuria. Using machine learning we trained a computer to first identify glomeruli and then measure glomerular dimensions.

Methods

The training set comprised ~100 images (varying magnification) of hematoxylin-eosin (H-E) or periodic acid Schiff (PAS)-stained kidney tissue sourced from published literature. An open-source git-hub implementation of the mask region-convolutional neural network was used to generate a model and Keras HDF5 data used to identify, count, and measure areas of glomeruli present in the renal tissue on a Raspberry Pi 3 server. The test set comprised images (40X and 10X, n=16) of stained kidney tissue from rats sacrificed 14 days after puromycin aminonucleoside (167 mg/kg, intraperitoneal, n=3) administration, and a sham cohort of animals (n=3). Kidney sections comprising the test set had been stained with H-E or hematoxylin alone.

Results

The Raspberry Pi 3 machine was able to correctly identify glomeruli (A) while excluding non-glomerular structures in all test cases. Having correctly identified glomeruli, the machine was able to measure (B) glomerular area using a precalibrated tool.

Conclusion

Machine learning has enabled both recognition of glomeruli within renal sections and measurement of their dimensions, reducing time and labor while eliminating operator bias. This technology can be used in images from renal biopsies to diagnose ongoing glomerular kidney disease.

A glomerulus (A) amidist in a renal section was identified by the machine, greying out the surrounding nephrons. The machine was able to isolate several glomeruli and start quantifying their area (B).

Funding

  • Private Foundation Support