BackgroundInterstitial fibrosis, tubular atrophy (IFTA), and glomerulosclerosis are indicators of irrecoverable kidney injury. Modern machine learning (ML) tools have enabled robust, automated identification of image structures that can be comparable with analysis by human experts. ML algorithms were developed and tested for the ability to replicate the detection and quantification of IFTA and glomerulosclerosis that renal pathologists perform.MethodsA renal pathologist annotated renal biopsy specimens from 116 whole-slide images (WSIs) for IFTA and glomerulosclerosis. A total of 79 WSIs were used for training different configurations of a convolutional neural network (CNN), and 17 and 20 WSIs were used as internal and external testing cases, respectively. The best model was compared against the input of four renal pathologists on 20 new testing slides. Further, for 87 testing biopsy specimens, IFTA and glomerulosclerosis measurements made by pathologists and the CNN were correlated to patient outcome using classic statistical tools.ResultsThe best average performance across all image classes came from a DeepLab version 2 network trained at 40× magnification. IFTA and glomerulosclerosis percentages derived from this CNN achieved high levels of agreement with four renal pathologists. The pathologist- and CNN-based analyses of IFTA and glomerulosclerosis showed statistically significant and equivalent correlation with all patient-outcome variables.ConclusionsML algorithms can be trained to replicate the IFTA and glomerulosclerosis assessment performed by renal pathologists. This suggests computational methods may be able to provide a standardized approach to evaluate the extent of chronic kidney injury in situations in which renal-pathologist time is restricted or unavailable.

Background and objectivesIntradialytic hypotension has high clinical significance. However, predicting it using conventional statistical models may be difficult because several factors have interactive and complex effects on the risk. Herein, we applied a deep learning model (recurrent neural network) to predict the risk of intradialytic hypotension using a timestamp-bearing dataset.Design, setting, participants, & measurementsWe obtained 261,647 hemodialysis sessions with 1,600,531 independent timestamps (i.e., time-varying vital signs) and randomly divided them into training (70%), validation (5%), calibration (5%), and testing (20%) sets. Intradialytic hypotension was defined when nadir systolic BP was <90 mm Hg (termed intradialytic hypotension 1) or when a decrease in systolic BP ≥20 mm Hg and/or a decrease in mean arterial pressure ≥10 mm Hg on the basis of the initial BPs (termed intradialytic hypotension 2) or prediction time BPs (termed intradialytic hypotension 3) occurred within 1 hour. The area under the receiver operating characteristic curves, the area under the precision-recall curves, and F1 scores obtained using the recurrent neural network model were compared with those obtained using multilayer perceptron, Light Gradient Boosting Machine, and logistic regression models.ResultsThe recurrent neural network model for predicting intradialytic hypotension 1 achieved an area under the receiver operating characteristic curve of 0.94 (95% confidence intervals, 0.94 to 0.94), which was higher than those obtained using the other models (P<0.001). The recurrent neural network model for predicting intradialytic hypotension 2 and intradialytic hypotension 3 achieved area under the receiver operating characteristic curves of 0.87 (interquartile range, 0.87–0.87) and 0.79 (interquartile range, 0.79–0.79), respectively, which were also higher than those obtained using the other models (P≤0.001). The area under the precision-recall curve and F1 score were higher using the recurrent neural network model than they were using the other models. The recurrent neural network models for intradialytic hypotension were highly calibrated.ConclusionsOur deep learning model can be used to predict the real-time risk of intradialytic hypotension.

BackgroundAcute tubulointerstitial nephritis (ATIN) is an important cause of acute kidney injury and often a potentially reversible disease. However, the role of steroids in ATIN remains controversial and the underlying mechanisms remain unresolved.MethodsA total of 113 adult patients with biopsy-proven ATIN were recruited from three tertiary referral centers. Of 102 patients with idiopathic or drug-induced ATIN, outcomes such as renal recovery, end-stage renal disease, and all-cause mortality were compared between the steroid-treated and non-treated groups. Plasma and urine inflammatory cytokine levels at the time of biopsy were analyzed in patients (n = 33) using a bead-based multiplex assay and compared with those of healthy individuals (n = 40).ResultsSteroids were used in 92 (81.4%) of the total patients and in 82 (80.3%) patients with idiopathic or drug-induced ATIN. The rate of renal recovery and the risks of end-stage renal disease and mortality were not different between the steroid-treated and non-treated groups. Despite using a propensity score matching method (n = 20 in each group), none of the outcomes were different between the two groups. Several cytokines, such as monocyte chemotactic protein-1, interferon-α, and interleukin-6 and interleukin-8 levels, were markedly elevated in plasma and urine of patients compared with those in healthy individuals. However, cytokines related to Th2 response, such as IL-10, IL-33, were not different between the two groups.ConclusionsSteroid use does not affect the overall outcome of ATIN. Based on the fact that targeting therapy should be investigated to improve outcomes, the present cytokine results will be helpful for developing a novel therapy for ATIN.Electronic supplementary materialThe online version of this article (10.1186/s12882-019-1277-2) contains supplementary material, which is available to authorized users.

Background: Henoch-Schönlein purpura nephritis (HSPN), a small-vessel vasculitis, shares renal pathological features with immunoglobulin A nephropathy. Oxford classification of immunoglobulin A nephropathy pathology has been updated to the MEST-C score, but its application in HSPN remains unresolved. Methods: Two hundred and thirteen patients with biopsy-proven HSPN were retrieved from the Seoul National University Hospital between 2000 and 2017. Renal outcome risks (i.e., end-stage renal disease or doubling of serum creatinine) were evaluated according to MEST-C scores after stratification by age: 113 children aged < 18 years (9.2 ± 3.6 years) and 100 adults aged ≥18 years (38.6 ± 18.3 years). We pooled our data with four previous cohort studies in which MEST or MEST-C scores were described in detail.Results: Twenty-one child (19%) and 16 adult (16%) patients reached the renal outcome during the median followup periods of 12 years and 13 years, respectively (maximum 19 years). In children, M1 and T1/T2 scores revealed worse renal outcomes than did M0 and T0 scores, respectively, whereas the T score was the only factor related to worse outcomes in adult patients after adjusting for multiple clinical and laboratory variables. The pooled data showed that M1, S1, and T1/T2 in children and E1 and T1/T2 in adults were correlated with poorer renal outcomes than those of their counterpart scores. Conclusions: The Oxford classification MEST-C scores can predict long-term renal outcomes in patients with HSPN.

No abstract

Background Although patients undergoing continuous renal replacement therapy (CRRT) due to acute kidney injury (AKI) frequently have instability in mean arterial pressure (MAP), no consensus exists on the target value of MAP related to high mortality after CRRT. Methods A total of 2,292 patients who underwent CRRT due to AKI in three referral hospitals were retrospectively reviewed. The MAPs were divided into tertiles, and the 3rd tertile group served as a reference in the analyses. The major outcome was all-cause mortality during the intensive care unit period. The odds ratio (OR) of mortality was calculated using logistic regression after adjustment for multiple covariates. The nonlinear relationship regression model was applied to determine the threshold value of MAP related to increasing mortality. Results The mean value of MAP was 80.7 ± 17.3 mmHg at the time of CRRT initiation. The median intensive care unit stay was 5 days (interquartile range, 2–12 days), and during this time, 1,227 (55.5%) patients died. The 1st tertile group of MAP showed an elevated risk of mortality compared with the 3rd tertile group (adjusted OR, 1.28 [1.03–1.60]; P = 0.029). In the nonlinear regression analysis, the threshold value of MAP was calculated as 82.7 mmHg. Patients with MAP < 82.7 mmHg had a higher mortality rate than those with ≥ 82.7 mmHg (adjusted OR, 1.21 [1.01–1.45]; P = 0.037). Conclusions Low MAP at CRRT initiation is associated with a high risk of mortality, particularly when it is < 82.7 mmHg. This value may be used for risk classification and as a potential therapeutic target.

Background Hyperlactatemia occurs frequently in critically ill patients, and this pathologic condition leads to worse outcomes in several disease subsets. Herein, we addressed whether hyperlactatemia is associated with the risk of mortality in patients undergoing continuous renal replacement therapy (CRRT) due to acute kidney injury. Methods A total of 1,661 patients who underwent CRRT for severe acute kidney injury were retrospectively reviewed between 2010 and 2020. The patients were categorized according to their serum lactate levels, such as high (≥ 7.6 mmol/l), moderate (2.1–7.5 mmol/l) and low (≤ 2 mmol/l), at the time of CRRT initiation. The hazard ratios (HRs) for the risk of in-hospital mortality were calculated with adjustment of multiple variables. The increase in the area under the receiver operating characteristic curve (AUROC) for the mortality risk was evaluated after adding serum lactate levels to the Sequential Organ Failure Assessment (SOFA) and the Acute Physiology and Chronic Health Evaluation (APACHE) II score-based models. Results A total of 802 (48.3%) and 542 (32.6%) patients had moderate and high lactate levels, respectively. The moderate and high lactate groups had a higher risk of mortality than the low lactate group, with HRs of 1.64 (1.22–2.20) and 4.18 (2.99–5.85), respectively. The lactate-enhanced models had higher AUROCs than the models without lactates (0.764 vs. 0.702 for SOFA score; 0.737 vs. 0.678 for APACHE II score). Conclusions Hyperlactatemia is associated with mortality outcomes in patients undergoing CRRT for acute kidney injury. Serum lactate levels may need to be monitored in this patient subset.

Background Precise prediction of contrast media–induced acute kidney injury (CIAKI) is an important issue because of its relationship with poor outcomes. Objective Herein, we examined whether a deep learning algorithm could predict the risk of intravenous CIAKI better than other machine learning and logistic regression models in patients undergoing computed tomography (CT). Methods A total of 14,185 patients who were administered intravenous contrast media for CT at the preventive and monitoring facility in Seoul National University Hospital were reviewed. CIAKI was defined as an increase in serum creatinine of ≥0.3 mg/dL within 2 days or ≥50% within 7 days. Using both time-varying and time-invariant features, machine learning models, such as the recurrent neural network (RNN), light gradient boosting machine (LGM), extreme gradient boosting machine (XGB), random forest (RF), decision tree (DT), support vector machine (SVM), κ-nearest neighbors, and logistic regression, were developed using a training set, and their performance was compared using the area under the receiver operating characteristic curve (AUROC) in a test set. Results CIAKI developed in 261 cases (1.8%). The RNN model had the highest AUROC of 0.755 (0.708-0.802) for predicting CIAKI, which was superior to that obtained from other machine learning models. Although CIAKI was defined as an increase in serum creatinine of ≥0.5 mg/dL or ≥25% within 3 days, the highest performance was achieved in the RNN model with an AUROC of 0.716 (95% confidence interval [CI] 0.664-0.768). In feature ranking analysis, the albumin level was the most highly contributing factor to RNN performance, followed by time-varying kidney function. Conclusions Application of a deep learning algorithm improves the predictability of intravenous CIAKI after CT, representing a basis for future clinical alarming and preventive systems.