RT Journal Article
SR Electronic
T1 Deep learning identifies pathological abnormalities predictive of graft loss in kidney transplant biopsies
JF bioRxiv
FD Cold Spring Harbor Laboratory
SP 2021.04.18.440166
DO 10.1101/2021.04.18.440166
A1 Zhengzi Yi
A1 Fadi Salem
A1 Madhav C Menon
A1 Karen Keung
A1 Caixia Xi
A1 Sebastian Hultin
A1 M. Rizwan Haroon Al Rasheed
A1 Li Li
A1 Fei Su
A1 Zeguo Sun
A1 Chengguo Wei
A1 Weiqing Huang
A1 Samuel Fredericks
A1 Qisheng Lin
A1 Khadija Banu
A1 Germaine Wong
A1 Natasha M. Rogers
A1 Samira Farouk
A1 Paolo Cravedi
A1 Meena Shingde
A1 R. Neal Smith
A1 Ivy A. Rosales
A1 Philip J. O’Connell
A1 Robert B. Colvin
A1 Barbara Murphy
A1 Weijia Zhang
YR 2021
UL http://biorxiv.org/content/early/2021/04/19/2021.04.18.440166.abstract
AB Background Interstitial fibrosis, tubular atrophy, and inflammation are major contributors to renal allograft failure. Here we seek an objective, quantitative pathological assessment of these lesions to improve predictive utility.Methods We constructed a deep-learning-based pipeline recognizing normal vs. abnormal kidney tissue compartments and mononuclear leukocyte (MNL) infiltrates from Periodic acid-Schiff (PAS) stained slides of transplant biopsies (training: n=60, testing: n=33) that quantified pathological lesions specific for interstitium, tubules and MNL infiltration. The pipeline was applied to 789 whole slide images (WSI) from baseline (n=478, pre-implantation) and 12-month post-transplant (n=311) protocol biopsies in two independent cohorts (GoCAR: 404 patients, AUSCAD: 212 patients) of transplant recipients to correlate composite lesion features with graft loss.Results Our model accurately recognized kidney tissue compartments and MNLs. The digital features significantly correlated with Banff scores, but were more sensitive to subtle pathological changes below the thresholds in Banff scores. The Interstitial and Tubular Abnormality Score (ITAS) in baseline samples was highly predictive of 1-year graft loss (p=2.8e-05), while a Composite Damage Score (CDS) in 12-month post-transplant protocol biopsies predicted later graft loss (p=7.3e-05). ITAS and CDS outperformed Banff scores or clinical predictors with superior graft loss prediction accuracy. High/intermediate risk groups stratified by ITAS or CDS also demonstrated significantly higher incidence of eGFR decline and subsequent graft damage.Conclusions This deep-learning approach accurately detected and quantified pathological lesions from baseline or post-transplant biopsies, and demonstrated superior ability for prediction of posttransplant graft loss with potential application as a prevention, risk stratification or monitoring tool.Competing Interest StatementDr. Murphy reports stock in RenalytixAI. Dr. Zhang reports personal fees from RenalytixAI. Drs. Murphy and Zhang report the patents (1. Patents US Provisional Patent Application F&R ref 27527-0134P01, Serial No. 61/951,651, filled March 2014. Method for identifying kidney allograft recipients at risk for chronic injury; 2. US Provisional Patent Application: Methods for Diagnosing Risk of Renal Allograft Fibrosis and Rejection (miRNA); 3. US Provisional Patent Application: Method For Diagnosing Subclinical Acute Rejection by RNA sequencing Analysis of A Predictive Gene Set; 4. US Provisional Patent Application: Pretransplant prediction of post- transplant acute rejection.); Dr. O'Connell is a consultant for CSL Behring and Vitaeris. Other investigators have no financial interest to declare.