@article{1a7e7af5bf8341939c700bd4691042f4,
title = "Deep learning identified pathological abnormalities predictive of graft loss in kidney transplant biopsies",
abstract = "Interstitial fibrosis, tubular atrophy, and inflammation are major contributors to kidney allograft failure. Here we sought an objective, quantitative pathological assessment of these lesions to improve predictive utility and constructed a deep-learning–based pipeline recognizing normal vs. abnormal kidney tissue compartments and mononuclear leukocyte infiltrates. Periodic acid– Schiff stained slides of transplant biopsies (60 training and 33 testing) were used to quantify pathological lesions specific for interstitium, tubules and mononuclear leukocyte infiltration. The pipeline was applied to the whole slide images from 789 transplant biopsies (478 baseline [pre-implantation] and 311 post-transplant 12-month protocol biopsies) in two independent cohorts (GoCAR: 404 patients, AUSCAD: 212 patients) of transplant recipients to correlate composite lesion features with graft loss. Our model accurately recognized kidney tissue compartments and mononuclear leukocytes. The digital features significantly correlated with revised Banff 2007 scores but were more sensitive to subtle pathological changes below the thresholds in the Banff scores. The Interstitial and Tubular Abnormality Score (ITAS) in baseline samples was highly predictive of one-year graft loss, while a Composite Damage Score in 12-month post-transplant protocol biopsies predicted later graft loss. ITASs and Composite Damage Scores outperformed Banff scores or clinical predictors with superior graft loss prediction accuracy. High/intermediate risk groups stratified by ITASs or Composite Damage Scores also demonstrated significantly higher incidence of estimated glomerular filtration rate decline and subsequent graft damage. Thus, our deep-learning approach accurately detected and quantified pathological lesions from baseline or post-transplant biopsies and demonstrated superior ability for prediction of post-transplant graft loss with potential application as a prevention, risk stratification or monitoring tool.",
keywords = "deep learning, graft survival, kidney transplantation, renal pathology",
author = "Zhengzi Yi and Fadi Salem and Menon, {Madhav C.} and Karen Keung and Caixia Xi and Sebastian Hultin and {Haroon Al Rasheed}, {M. Rizwan} and Li Li and Fei Su and Zeguo Sun and Chengguo Wei and Weiqing Huang and Samuel Fredericks and Qisheng Lin and Khadija Banu and Germaine Wong and Rogers, {Natasha M.} and Samira Farouk and Paolo Cravedi and Meena Shingde and Smith, {R. Neal} and Rosales, {Ivy A.} and O'Connell, {Philip J.} and Colvin, {Robert B.} and Barbara Murphy and Weijia Zhang",
note = "Funding Information: This work is a substudy of the Genomics of Chronic Renal Allograft Rejection (GoCAR) study sponsored by National Institutes of Health grant no. 5U01AI070107-03 . The cost of clinical, histologic, and genomic experiments, as well as the authors{\textquoteright} effort involved in patient enrollment, data analysis, and manuscript preparation were paid by this grant. All the authors have reviewed the manuscript and agreed to submission. Funding Information: We thank Ms. Meyke Hermsen in Dr. Jeroen A.W.M. van der Laak's lab in Department of Pathology of Radboud University Medical Center in Nijmegen for suggestions in kidney compartment annotation using ASAP (Automated Slide Analysis Platform) program. We thank the Scientific Computing Division at the Icahn School of Medicine at Mount Sinai for providing computational resources. This work is a substudy of the Genomics of Chronic Renal Allograft Rejection (GoCAR) study sponsored by National Institutes of Health grant no. 5U01AI070107-03. The cost of clinical, histologic, and genomic experiments, as well as the authors? effort involved in patient enrollment, data analysis, and manuscript preparation were paid by this grant. All the authors have reviewed the manuscript and agreed to submission. ZY designed and performed computational analyses and drafted the paper. FSa supervised pathology annotation and interpretation and edited the paper. MCM contributed to study design and edited the paper. KK was involved clinical data collection of the Australian Chronic Allograft Dysfunction (AUSCAD) cohort and edited the paper. CX, MRHAR, LL, FSu, and ZS annotated slides. SH was involved clinical data and image collection of the AUSCAD cohort. CW, WH, SFr, QL, KB, and SFa helped with clinical data mining and interpretation for the Genomics of Chronic Allograft Rejection (GoCAR) cohort and critical reading of the manuscript. GW and NMR were involved in patient and sample management for the AUSCAD cohort. PC was involved in data interpretation and edited the paper. MS was involved in the AUSCAD cohort pathologic assessment. RNS and IAR were involved in pathologic assessment of the GoCAR cohort. PJO'C supervised the ASUCAD cohort study and edited the paper. RBC supervised the pathology in the GoCAR cohort and edited the paper. BM supervised the GoCAR cohort study and was involved in study conception and paper editing. WZ conceptualized and designed this study and edited the paper. Publisher Copyright: {\textcopyright} 2021 International Society of Nephrology",
year = "2022",
month = feb,
doi = "10.1016/j.kint.2021.09.028",
language = "English",
volume = "101",
pages = "288--298",
journal = "Kidney International",
issn = "0085-2538",
publisher = "Elsevier Inc.",
number = "2",
}