Non-invasive Assessment of Urinary Exfoliated Proximal Tubule Cell Multispectral Autofluorescence May Differentiate Between Causes of Kidney Transplant Dysfunction

Henry H L Wu; Yandong Lang; Shannon Handley; Aline Knab; Adnan Agha; Yuan Tian; Akanksha Bhargava; Ewa M Goldys; Carol A Pollock; Sonia Saad

doi:10.34067/KID.0000000879

Non-invasive Assessment of Urinary Exfoliated Proximal Tubule Cell Multispectral Autofluorescence May Differentiate Between Causes of Kidney Transplant Dysfunction

Kidney360. 2025 Jun 20. doi: 10.34067/KID.0000000879. Online ahead of print.

Authors

Henry H L Wu^{1

2

3}, Yandong Lang², Shannon Handley², Aline Knab², Adnan Agha², Yuan Tian², Akanksha Bhargava², Ewa M Goldys², Carol A Pollock^{1

3}, Sonia Saad¹

Affiliations

¹ Renal Research Laboratory, Kolling Institute of Medical Research, Royal North Shore Hospital & The University of Sydney, Sydney, Australia.
² ARC Centre of Excellence for Nanoscale Biophotonics, School of Biomedical Engineering, The University of New South Wales, Sydney, Australia.
³ Department of Renal Medicine, Royal North Shore Hospital, Northern Sydney Local Health District, Sydney, Australia.

PMID: 40540353
DOI: 10.34067/KID.0000000879

Abstract

Background: Complications relating to delayed or deteriorating graft function following kidney transplantation are common. There is no validated method apart from transplant kidney biopsy which can accurately identify between the histopathological causes of graft dysfunction. Considering an unmet critical need for a non-invasive approach to reliably diagnose kidney transplant complications, this work proposes a novel methodology based on the assessment of exfoliated proximal tubule cells (PTCs) extracted from urine of kidney transplant recipients by using their multispectral autofluorescence features.

Methods: Three groups of 10 patients who have undergone clinically indicated transplant kidney biopsy and was subsequently diagnosed with either acute tubular necrosis (ATN), graft rejection or non-rejection associated interstitial fibrosis and tubular atrophy (IFTA) took part in this study. Exfoliated PTCs from urine collected prior to transplant biopsy were extracted using a validated immunomagnetic separation method based on anti-CD13 and anti-SGLT2 antibodies. Imaging was performed on a custom-made multispectral autofluorescence microscopy and camera system. Multispectral autofluorescence images of PTCs were quantitatively analysed by using optimised small sets of features to prevent overfitting. Binary classification was carried out by a random forest classifier, and the AutoGluon machine learning software. Results were validated by 5-fold cross validation.

Results: For random forest classification, features were selected using entropy-based feature selection, resulting in AUC values of 0.92 (ATN versus graft rejection), 0.86 (ATN versus IFTA) and 0.62 (graft rejection versus IFTA) respectively. The AutoGluon classifier optimisation for the same features resulted in AUC values of 0.95 (ATN versus graft rejection), 0.92 (ATN versus IFTA) and 0.91 (graft rejection vs IFTA).

Conclusions: Our results demonstrate a proof-of-concept that measurement of autofluorescent features from urinary exfoliated PTCs multispectral autofluorescence could differentiate between patient groups with ATN, graft rejection and IFTA in kidney transplant recipients to an excellent degree of accuracy using AutoGluon classifier optimisation.