ASN's Mission

ASN leads the fight to prevent, treat, and cure kidney diseases throughout the world by educating health professionals and scientists, advancing research and innovation, communicating new knowledge, and advocating for the highest quality care for patients.

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: FR-PO994

Classification of Cell Types with Neural Networks in Reference and Diseased Human and Mouse Kidney Tissue Using Nuclear Morphology

Session Information

Category: Pathology and Lab Medicine

  • 1601 Pathology and Lab Medicine: Basic

Authors

  • Woloshuk, Andre, Indiana University School of Medicine, Indianapolis, Indiana, United States
  • Dean, Dawson Frank, Indiana University Division of Nephrology, Indianapolis, Indiana, United States
  • Mcnutt, Andrew, Indiana University, Indianapolis, Indiana, United States
  • Eadon, Michael T., Indiana University Division of Nephrology, Indianapolis, Indiana, United States
  • Dagher, Pierre C., Indiana University, Indianapolis, Indiana, United States
  • Winfree, Seth, Indiana University School of Medicine, Indianapolis, Indiana, United States
  • El-Achkar, Tarek M., Indiana University, Indianapolis, Indiana, United States
Background

Despite improvements in non-invasive analysis of kidney function, kidney biopsy remains the gold standard for diagnosing renal pathology. This field relies heavily on subjective interpretation and semi-quantitative image analysis of 2-dimentional (2D) images of stained tissue thin-sections. Advances in 3D confocal fluorescence imaging and machine learning approaches such as neural networks provide the opportunity for an automated and quantitative approach, and the potential for extracting new data from the 3-dimentional (3D) space. Neural networks have wide applications in image classification of natural images and medical imaging. Recent applications of neural networks in pathology are starting to explore segmentation and classification of histologically stained samples. The same approach has not been fully exploited in kidney tissue labeled with multiple fluorescent probes and imaged in 3D.

Methods

We identify individual cells in human and mouse kidney tissue and assigning each cell a ground truth classification based on validated cell markers. Images of the nuclei as 2D projections and 3D volumes from tissue are extracted and classified based on these markers using volumetric tissue exploration and analysis cytometry. Different neural network architectures are trained and evaluated using this image database. The efficacy of different architectures is assessed by their ability to distinguish different cell types within the biopsy.

Results

In this work, we create an image database of fluorescently stained nuclei collected from human and mouse renal tissue that can be used to identify and classify different cell types solely on their nuclear features. Furthermore, we begin to demonstrate the efficacy of identifying pathologies in either a mouse model of acute kidney injury or in human diabetes.

Conclusion

This work lays the groundwork for quantifying the types of cells present in biopsies, and the automatic classification of pathological cell states in a biopsy that may have otherwise gone unnoticed. Ultimately, machine learning-augmented image analysis has the potential not only to describe novel and distinct disease features, but also define a standardized approach to quantifying pathology in a kidney biopsy.

Funding

  • NIDDK Support