Kidney Week

Abstract: SA-PO0947

Automatic Segmentation-Derived Measurements of Structural Changes in Native and Transplant Kidney Biopsies Identify Patients at High Risk for Progression

Session Information

• Pathology: Updates and Insights
  November 08, 2025 | Location: Exhibit Hall, Convention Center
  Abstract Time: 10:00 AM - 12:00 PM

Category: Pathology and Lab Medicine

• 1800 Pathology and Lab Medicine

Authors

• Fareeduddin, Syed Khooshal, Mayo Foundation for Medical Education and Research, Rochester, Minnesota, United States

Syed Khooshal Fareeduddin, MBBS

• Denic, Aleksandar, Mayo Foundation for Medical Education and Research, Rochester, Minnesota, United States

Aleksandar Denic, MD, PhD

• Mullan, Aidan F., Mayo Foundation for Medical Education and Research, Rochester, Minnesota, United States

Aidan F. Mullan, MA

• Stetzik, Lucas, Aiforia Inc, Cambridge, Massachusetts, United States

Lucas Stetzik, PhD

• Park, Walter, Mayo Foundation for Medical Education and Research, Rochester, Minnesota, United States

Walter Park

• Stegall, Mark D., Mayo Foundation for Medical Education and Research, Rochester, Minnesota, United States

Mark D. Stegall, MD

• Barisoni, Laura, Duke University, Durham, North Carolina, United States

Laura Barisoni, MD

• Smith, Maxwell L., Mayo Foundation for Medical Education and Research, Rochester, Minnesota, United States

Maxwell L. Smith, MD

• Rule, Andrew D., Mayo Foundation for Medical Education and Research, Rochester, Minnesota, United States

Andrew D. Rule, MD, MS, FASN

Background

Structural damage of the kidney predicts disease progression, but its quantification is tedious and with limited reproducibility. We used a multiclass AI model applied to whole slide images (WSI) to assess chronic changes and test their clinical relevance.

Methods

We studied 322 adult patients from Olmsted County who underwent a native kidney biopsy, and 906 transplant recipients who had a 5-year surveillance biopsy. We used Aiforia Create, an AI-based interactive tool, to train classes across nested layers to detect 11 different kidney structures on PAS-stained WSIs. The counts and areas of kidney structures were then converted into different measures of chronic changes (e.g. %glomerulosclerosis). We assessed the risk of end stage kidney disease (ESKD) and graft loss with chronic changes and interstitial inflammation using Cox proportional hazards models measured the prediction of disease progression of these measurements.

Results

Forty (12.4%) patients with native kidney disease developed ESKD (median follow-up 7.2 years), and 71 (7.8%) transplant recipients developed graft failure (median follow-up 8.4 years). In both cohorts, higher levels of chronic changes and interstitial inflammation were associated with increased risk of ESKD and graft loss. In multivariable analysis, larger mean tubular area, higher %glomerulosclerosis, %tubular atrophy, %arteriolar hyalinosis and inflammation foci density were independent predictors of kidney failure in native kidneys (C-statistic: 0.870). In transplant recipients, higher tubular atrophy foci density, % arteriolar hyalinosis, inflammation foci density, and % luminal stenosis, were independent predictors of graft loss (C-statistic: 0.859).

Conclusion

Computationally derived measurements of kidney chronic changes and inflammation from both native and transplanted kidneys identify patients at highest risk for progression.

Funding

• NIDDK Support

Digital Object Identifier (DOI)

• doi: 10.1681/ASN.2025a6h2nmym