Kidney Week

Abstract: TH-OR020

Three-Dimensional (3D) Kidney Pathology of Whole Intact Biopsies and Nephrectomies Using Lightsheet Microscopy and Computational Image Analysis

Session Information

• Artificial Intelligence in Kidney Care: From Pathology to Prediction
  October 22, 2026 | Location: Room 711, Convention Center
  Abstract Time: 04:40 PM - 04:50 PM

Category: Artificial Intelligence, Digital Health, and Data Science

• 300 Artificial Intelligence, Digital Health, and Data Science

Authors

• Poudel, Chetan, Indiana University School of Medicine, Indianapolis, Indiana, United States

Chetan Poudel, PhD

• Phillips, Carrie L., Indiana University School of Medicine, Indianapolis, Indiana, United States

Carrie L. Phillips, MD

• Kelly, Katherine J., Indiana University School of Medicine, Indianapolis, Indiana, United States

Katherine J. Kelly, MD, MS, MPH, FASN

• El-Achkar (Ashkar), Tarek M., Indiana University School of Medicine, Indianapolis, Indiana, United States

Tarek M. El-Achkar (Ashkar), MD, FASN

• Eadon, Michael T., Indiana University School of Medicine, Indianapolis, Indiana, United States

Michael T. Eadon, MD, FASN

• Dagher, Pierre C., Indiana University School of Medicine, Indianapolis, Indiana, United States

Pierre C. Dagher, MD, FASN

Background

Conventional renal pathology relies on thin, two-dimensional (2D) biopsy sections representing only a small fraction (<10% at best) of total tissue volume, leaving most tissue unassessed. This undersampling introduces diagnostic uncertainty, particularly in focal diseases. Thin sectioning also disrupts the spatial continuity of functional tissue units (FTUs), limiting correlations of glomerular, tubular, and vascular injuries within the same nephron. Emerging 3D pathology workflows integrating tissue labeling, clearing, volumetric imaging, and artificial intelligence (AI) may overcome these limitations through comprehensive tissue interrogation.

Methods

Intact renal biopsy and nephrectomy specimens of various preparation types (fresh-fixed, frozen, FFPE) were stained using rapid fluorescent labeling strategies (Path3D and FLARE), cleared using ethyl cinnamate protocols, and imaged with open-top lightsheet microscopy at high-throughput and sub-cellular resolution. Three-dimensional fluorescence datasets were computationally transformed into H&E and PAS histology volumes. Deep-learning models quantified whole-glomerular volumes, cell counts, and lesion burden in health and disease (diabetes and FSGS). Our recently developed TubuleMAP software was used to directly map glomerular injury to tubule injury within the same nephron.

Results

The 3D renal pathology pipeline generated volumetric H&E and PAS histology from intact specimens and remained compatible with downstream histologic and molecular assays. Nephrectomies subjected to 3D imaging, re-embedded in FFPE, and processed using conventional 2D histology workflows showed minimal alteration in tissue architecture. Even in modest 500-µm-thick nephrectomy specimen, 45% of glomeruli were fully captured (n = 92/206), and approximately 15 mm of a single continuous nephron segment was visualized.

Conclusion

We developed and validated a practical 3D renal pathology workflow compatible with millimeter-thick human kidney specimens across multiple preparations. This platform enables nondestructive volumetric histology, quantitative analysis of glomeruli and nephrons, and spatial correlation of injury across FTUs. This approach may improve diagnostic accuracy, disease phenotyping, and precision nephropathology.

Funding

• NIDDK Support