Deep Learning Algorithms for the Prediction of Posttransplant Renal Function in Deceased-Donor Kidney Recipients: A Preliminary Study Based on Pretransplant Biopsy

Abstract

BackgroundPosttransplant renal function is critically important for kidney transplant recipients. Accurate prediction of graft function would greatly help in deciding acceptance or discard of allocated kidneys.Methods: Whole-slide images (WSIs) of H&E-stained donor kidney biopsies at × 200 magnification between January 2015 and December 2019 were collected. The clinical characteristics of each donor and corresponding recipient were retrieved. Graft function was indexed with a stable estimated glomerular filtration rate (eGFR) and reduced graft function (RGF). We used convolutional neural network (CNN)-based models, such as EfficientNet-B5, Inception-V3, and VGG19 for the prediction of these two outcomes.ResultsIn total, 219 recipients with H&E-stained slides of the donor kidneys were included for analysis [biopsies from standard criteria donor (SCD)/expanded criteria donor (ECD) was 191/28]. The results showed distinct improvements in the prediction performance of the deep learning algorithm plus the clinical characteristics model. The EfficientNet-B5 plus clinical data model showed the lowest mean absolute error (MAE) and root mean square error (RMSE). Compared with the clinical data model, the area under the receiver operating characteristic (ROC) curve (AUC) of the clinical data plus image model for eGFR classification increased from 0.69 to 0.83. In addition, the predictive performance for RGF increased from 0.66 to 0.80. Gradient-weighted class activation mappings (Grad-CAMs) showed that the models localized the areas of the tubules and interstitium near the glomeruli, which were discriminative features for RGF.ConclusionOur results preliminarily show that deep learning for formalin-fixed paraffin-embedded H&E-stained WSIs improves graft function prediction accuracy for deceased-donor kidney transplant recipients.

European Society for Organ Transplantation (ESOT)-TLJ 3.0 Consensus on Histopathological Analysis of Pre-Implantation Donor Kidney Biopsy: Redefining the Role in the Process of Graft Assessment

“…Luo et al [ 35 ] used donor kidney biopsy WSIs as a source of features in addition to clinical characteristics for graft function prediction, building neural network models to predict stable eGFR and reduced graft function (RGF) in deceased-donor kidney transplant recipients who underwent pre-transplantation biopsy. They tested six prediction models on 219 WSIs.…”

Multiview Clustering to Identify Novel Kidney Donor Phenotypes for Assessing Graft Survival in Older Transplant Recipients

“… 48 Although we addressed missingness with multiple imputation, there is the possibility that our outcomes may have differed with a more complete dataset. Finally, we were pragmatic in our selection of donor variables in cluster derivation and added information on donor organs (including pretransplant biopsy results 49 ) may have better predicted death/graft failure. However, these data were not available in our derivation cohort, and our intended purpose was to use factors readily available in other registry databases.…”

Deep learning applications for kidney histology analysis

“…One reason might be the lack of large cohorts needed to train such models. The currently largest kidney biopsy classification study showed the feasibility of predicting kidney allograft diseases from WSI of periodic acid-schiff (PAS), Hematoxylin and Eosin (HE), and Silver-stained biopsies [25 As a proof-of-concept, this study only predicted low-granular classes, that is, biopsies that are normal, show rejection or other diseases. The study showed that such a model could be used for prioritization of biopsies for diagnostics, that is, prioritizing biopsies with rejection and other diseases over normal ones to be handled by a nephropathologist as first.…”

“…Regression studies based on imaging data in nephropathology are less common in comparison with the number of segmentation and classification studies, however, it is possible to extract information from an image using feature extractors like CNN-based abstract feature extractors and combine this information with the features from other modalities such as features extracted from encoding the clinical data of each donor and the corresponding recipient and passing the combined features to a layer with one neuron to regress the estimated glomerular filtration rate (eGFR) values for post-transplant recipient [26 ▪ ].…”