Kidney Week

Abstract: SA-PO044

Whole-Slide Podocyte Quantification in Renal Tissue via p57 Immunohistochemistry

Session Information

• Engineering-Based Approaches to Problems in Nephrology
  November 09, 2019 | Location: Exhibit Hall, Walter E. Washington Convention Center
  Abstract Time: 10:00 AM - 12:00 PM

Category: Bioengineering

• 300 Bioengineering

Authors

• Santo, Briana A., University at Buffalo, Buffalo, New York, United States

Briana A. Santo,

• Rosenberg, Avi Z., Johns Hopkins University, Baltimore, Maryland, United States

Avi Z. Rosenberg,

• Daneshpajouhnejad, Parnaz, Johns Hopkins University, Baltimore, Maryland, United States

Parnaz Daneshpajouhnejad,

• Sarder, Pinaki, SUNY Buffalo, Buffalo, New York, United States

Pinaki Sarder,

Background

Podocyte injury and podocytopathy drive the progression of Diabetic Nephropathy (DN). Histologic examination of podocytes in disease sections is challenging due to the lack of podocyte specific markers, as well as the limited resolution of podocytes, from all glomerular cells, in traditional stains. Based on our unique podocyte staining protocol, we have developed a whole slide podocyte computational pipeline.

Methods

Murine kidney sections (n = 66) were IHC stained for p57^kip2, and PAS post-stained. Whole slide images (WSIs) were acquired, and a neural network segmented glomeruli. WSIs entered our MATLAB pipeline, which applies color deconvolution and morphological processing to segment and quantify podocytes (Fig A). We derived 28 features from segmented podocytes and glomeruli. Pathologist estimated podocyte count per glomerulus was compared with computational results for pipeline evaluation.

Results

For a subset of WSIs, pathologist and pipeline counts were comparable (Fig B). In practice, pathologist estimates are based on small, random samples of glomeruli from WSIs. Our computational pipeline analyzes all glomeruli, thus producing more accurate results. Pathologist assessments were found to be highly variable with estimates ranging ±23% (Fig C). Our pipeline averages 0.04 seconds per glomerulus, while the pathologist averages 36. Therefore, our computational approach achieves higher precision at a conservative estimate of 900x faster rate than a pathologist.

Conclusion

Our unique staining protocol and pipeline enable rapid quantification of podocyte populations in renal tissue specimens, which will expedite disease-specific evaluation of podocytes, and inform on the contribution of podocyte structural and functional integrity to disease progression.

Funding

• NIDDK Support