# Enhancing Organ Allocation Efficiency: A Pilot Study Evaluating Artificial Intelligence-Assisted Assessment of Donor Kidney Pathology

## TL;DR

This study shows that AI can help pathologists evaluate donor kidneys faster without losing accuracy, potentially speeding up organ transplants.

## Contribution

The study introduces an AI-assisted system that improves efficiency and accuracy in assessing donor kidney viability.

## Key findings

- AI-assisted review reduced evaluation time by 54.83% compared to manual review.
- AI-assisted review showed 98.33% agreement with expert assessments on kidney viability thresholds.
- AI-assisted review had higher correlation coefficients with ground truth than manual review.

## Abstract

Purpose: The purpose of this study is to evaluate the effectiveness of an artificial intelligence (AI)-assisted review (AAR) system in improving diagnostic accuracy, efficiency, and concordance with expert assessments during the evaluation of donor kidney viability.

Methods: Sixty H&E-stained frozen-section kidney biopsy slides from explant kidneys obtained for organ donation were evaluated. A board-certified renal pathologist established ground truth (GT) through manual digital evaluation on the Techcyte Fusion Platform. The slides were independently reviewed by an AI algorithm, a board-certified pathologist (Reviewer 2 (R2)), and a board-certified transplant surgeon (Reviewer 1 (R1)). After a washout period, AI-assisted reads were performed. The performance of AAR and manual digital review (MDR) was compared to the GT for total and sclerotic glomeruli (SG) counts, as well as concordance with kidney viability thresholds (using a 20% SG cutoff rate). Secondary outcomes included comparisons of review times and concordance rates for AAR, MDR, and AI analysis alone with the GT.

Results: AAR demonstrated concordance with GT across parameters. For R1, coefficient of determination (COD) values for SG counts improved with AAR (0.833) compared to MDR (0.81). Agreement at the 20% SG threshold for kidney viability was 98.33% for both AAR and MDR. AAR reduced mean review times (minutes) by 54.83% compared to MDR, with per-slide review times decreasing from 17:09 (MDR) to 8:35 (AAR). Pearson correlation coefficients (PCC) and concordance correlation coefficients (CCC) for AAR were generally higher than MDR, particularly for the percentage of SG, indicating improved alignment with GT. Analyses revealed no systematic bias, with AAR aligning more closely with GT compared to MDR for both reviewers.

Conclusion: The Techcyte algorithm reduces review time while maintaining accuracy and concordance with experts, promoting AI adoption to improve workflow efficiency and expedite transplantation decisions.