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

Abstract: TH-PO807

A Comparison of Artificial Intelligence (AI) vs. Human-Derived Measures of Nephron Size

Session Information

Category: Pathology and Lab Medicine

  • 1800 Pathology and Lab Medicine

Authors

  • Jagtap, Jaidip M., Mayo Clinic Minnesota, Rochester, Minnesota, United States
  • Janowczyk, Andrew, Emory University, Atlanta, Georgia, United States
  • Denic, Aleksandar, Mayo Clinic Minnesota, Rochester, Minnesota, United States
  • Chen, Yijiang, Case Western Reserve University, Cleveland, Ohio, United States
  • Asghar, Muhammad Sohaib, Mayo Clinic Minnesota, Rochester, Minnesota, United States
  • Ramanathan, Sumana, Mayo Clinic Minnesota, Rochester, Minnesota, United States
  • Mullan, Aidan F., Mayo Clinic Minnesota, Rochester, Minnesota, United States
  • Kline, Timothy L., Mayo Clinic Minnesota, Rochester, Minnesota, United States
  • Erickson, Bradley J., Mayo Clinic Minnesota, Rochester, Minnesota, United States
  • Barisoni, Laura, Duke University, Durham, North Carolina, United States
  • Rule, Andrew D., Mayo Clinic Minnesota, Rochester, Minnesota, United States
Background

Enlarged nephrons as detected by glomerular volume and cortex per glomerulus (reciprocal of glomerular density) on kidney biopsy are prognostic for progressive CKD but require stereological calculations based on tedious human annotations. We sought to compare clinical associations with nephron size as calculated by AI versus human.

Methods

PAS-stained wedge section whole slide images (WSI) from nephrectomies (N=932) were manually annotated for cortex and non-sclerosed glomeruli using ImageScope software. A previously developed AI model for glomerular segmentation (NEPTUNE study) was applied to the same WSI. A threshold of 2000 µm2 for the smallest glomerular profile excluded false positives with the AI model. Spatial intersection statistics were compared between AI and human derived segmentations. Clinical correlations and the risk of progressive CKD (kidney failure or a 40% decline in eGFR sustained for at least 3 months) was assessed using AI versus human derived glomerular volume and cortex per glomerulus.

Results

The intersection between AI and human annotations of glomeruli had a Dice of 96%, precision of 99%, and recall of 93%. Glomerular volume was larger by AI than human (.0030 vs .0027 mm3, p<0.001) as was cortex per glomerulus (.080 vs .077 mm3, p<0.001). There were 52 progressive CKD events. Correlation of clinical characteristics with AI vs human estimates of glomerular volume is shown in the Table. The risk of progressive CKD with glomerular volume (per SD) was 1.76 (95%CI 1.41-2.19) by AI and 1.89 (95%CI 1.61-2.23) by human and with cortex per glomerulus (per SD) was 1.82 (95%CI 1.49-2.21) by AI and 1.71 (95%CI 1.52-1.93) by human.

Conclusion

The AI approach provides efficient quantification of nephron size measures comparable to a human approach.

Characteristics correlated with glomerular volume
 Human r (p value)AI r (p value)
Age-.09 (.009)-.13 (.0001)
Male.23 (<.0001).23 (<.0001)
BMI.31 (<.0001).32 (<.0001)
Hypertension.15 (<.0001).13 (<.0001)
Diabetes.14 (<.0001).14 (<.0001)
Smoker.10 (.004).11 (.001)
eGFR.00 (.90).04 (.26)
24hr protein.18 (<.0001).15 (<.0001)

Funding

  • NIDDK Support