Bayesian statistics is gaining momentum in biomedical research, offering a flexible and intuitive framework for integrating prior knowledge, managing uncertainty, and informing decision-making. In nephrology, a field marked by complex pathophysiology, small patient populations, and evolving therapeutic landscapes, Bayesian approaches may provide critical advantages over traditional statistical methods. Bayesian methodologies may help reshape nephrology by improving trial efficiency, refining diagnostic and prognostic precision, and strengthening causal inference. We conducted a comprehensive narrative review of peer-reviewed literature on Bayesian applications in nephrology, highlighting studies in adaptive clinical trial designs, dynamic risk prediction models, network meta-analyses, pharmacokinetic modelling, and Bayesian Mendelian randomization approaches, including Bayesian model averaging frameworks. Bayesian adaptive designs enabled more efficient and ethically sound trials, such as the WIRE platform study in renal cell carcinoma. In diagnostics and prognostics, Bayesian models allowed individualized inference from dynamic data, as demonstrated in eGFR trajectory analyses and mortality prediction. Bayesian network meta-analyses improved comparative effectiveness research by incorporating probabilistic treatment rankings, while Bayesian frameworks strengthened causal inference through methods like MR-Bayesian Model Averaging and Bayesian Kernel Machine Regression. Bayesian approaches have emerging applications in nephrology, fostering a probabilistic mindset, enhancing the interpretability of complex data, and enabling more individualized clinical strategies. Their integration into research and practice requires investment in training, interdisciplinary collaboration, and development of user-friendly tools. Embracing Bayesian thinking will be key to advancing precision medicine and improving patient outcomes in nephrology.

The digitalization of traditional glass slide microscopy into whole slide images has opened up new opportunities for pathology, such as the application of artificial intelligence techniques. Specialized software is necessary to visualize and analyze these images. One of these applications is QuPath, a popular bioimage analysis tool. This study proposes GNCnn, the first open-source QuPath extension specifically designed for nephropathology. It integrates deep learning models to provide nephropathologists with an accessible, automatic detector and classifier of glomeruli, the basic filtering units of the kidneys. The aim is to offer nephropathologists a freely available application to measure and analyze glomeruli to identify conditions such as glomerulosclerosis and glomerulonephritis. GNCnn offers a user-friendly interface that enables nephropathologists to detect glomeruli with high accuracy (Dice coefficient of 0.807) and categorize them as either sclerotic or non-sclerotic, achieving a balanced accuracy of 98.46%. Furthermore, it facilitates the classification of non-sclerotic glomeruli into 12 commonly diagnosed types of glomerulonephritis, with a top-3 balanced accuracy of 84.41%. GNCnn provides real-time updates of results, which are available at both the glomerulus and slide levels. This allows users to complete a typical analysis task without leaving the main application, QuPath. This tool is the first to integrate the entire workflow for the assessment of glomerulonephritis directly into the nephropathologists' workspace, accelerating and supporting their diagnosis.

There are no established models to predict the varied response to intensified immunosuppression in steroid-resistant nephrotic syndrome. Multimodal machine learning, integrating clinical and genetic data with nephropathology imaging, holds great promise to deliver such classifiers with perfect reproducibility. Here, we present such a theranostic classifier based on a multi-centric dataset. The PodoNet cohort with n = 201 biopsies was collected from 14 European centres with large domain shifts in whole slide images. The clinical input data contained 19 parameters including mutation (yes/no) and eGFR at biopsy; the ground truth theranostic endpoint was treatment response as no remission (n = 114), partial (n = 42) and complete (n = 45), defined by serum albumin and proteinuria. We trained our proprietary multimodal MorphSet++ architecture in a weakly supervised fashion. MorphSet++ integrates the clinical data vector from a shallow network at various stages with the transformer-based deep network analysing the nephropathology imaging data. Results are given as mean after 5-fold internal cross-validation. Mean AUC was .80 for complete, .77 for partial and .81 and xx for no remission. Mean positive predictive value was 89.72, 89.83, 89.27, mean negative predictive value was 94.83, 93.43 and 97.59; mean sensitivity was 84.21, 67.09, 98.47; mean specificity was 96.77, 98.40, 83.94; mean F1 Score was 86.88, 76.81, 93.65; mean accuracy was 84.21, 67.09, 98.47; mean balanced accuracy was 90.49, 82.75, 91.21 for no, partial and complete remission, respectively. Performance was best with both clinical data and histology as input. Our MorphSet++ architecture shows promising results as a theranostic tool, predicting response to immunosuppression in SRNS. MorphSet++ allows for rapid up-scaling with additional, larger datasets for even better and more robust performance. This might lead to a re-appraisal of nephropathology in the clinical management of SRNS.

Transplant arteriopathy involves a spectrum of Leukocyte Common Antigen-positive, hypoelastotic, foam cell intimal fibrosis. Transplant arteriopathy has been associated with both Chronic Active T Cell-Mediated Rejection and Antibody-Mediated Rejection chronicity. Aim of this study was to find clinicopathological correlates of transplant arteriopathy in a single centre retrospective cohort. We retrieved 46 biopsies showing transplant arteriopathy from 33 patients, out of a total of 784 biopsies carried out between 2005 and 2014. We retrospectively evaluated Banff Lesion Scores and Additional Diagnostic Parameters as well as the transplant arteriopathy descriptors Leukocyte Common Antigen-positive, hypoelastotic, foam cell, and correlated these findings with clinical data and death-censored transplant survival. Transplant arteriopathy was frequently associated with antibody-mediated rejection-associated Banff Lesions Scores and Additional Diagnostic Parameters. Hypoelastotic, leukocyte common antigen-positive and foam cell lesions were often combined, with hypoelastotic lesion being the most frequent finding in transplant arteriopathy. Leukocyte common antigen-positive lesion appeared earlier and was associated with Banff Lesion Score v ≥ 1. About half were positive for donor-specific antibodies, about a third had concurrent transplant glomerulopathy, and about a sixth were C4d-positive. Twelve of thirty-three transplants were lost during follow-up, concurrent transplant glomerulopathy was associated with shorter transplant survival. The frequent coincidence of transplant arteriopathy and indicators of antibody-mediated rejection suggests that this arterial remodelling could indeed be antibody-mediated rejection chronicity. The transplant community should re-examine transplant arteriopathy with an expanded definition including the previously ignored hypoelastotic lesion in order to re-confirm or reject with confidence transplant arteriopathy as Additional Diagnostic Parameter of Antibody-Mediated Rejection chronicity, and to learn about its prognostic and therapeutic implications.

Glomerular nephropathy resulting from the genetic defects in COL4A3/4/5 genes including the classical Alport syndrome is the second most common hereditary kidney disease characterized by persistent haematuria progressing to the need for kidney replacement therapy, frequently associated with sensorineural deafness, and occasionally with ocular anomalies. Diagnosis and management of COL4A3/4/5 glomerulopathy is a great challenge due to its phenotypic heterogeneity, multiple modes of inheritance, variable expressivity, and disease penetrance of individual variants as well as imperfect prognostic and progression factors and scarce and limited clinical trials, especially in children. As a joint initiative of the European Rare Kidney disease reference Network (ERKNet), European Renal Association (ERA Genes&Kidney), and European Society for Paediatric Nephrology (ESPN) Inherited renal disorders working group, a team of experts including adult and paediatric nephrologists, kidney geneticists, audiologists, ophthalmologists, and a kidney pathologist were selected to perform a systematic literature review on 21 clinically relevant PICO (Patient or Population covered, Intervention, Comparator, Outcome) questions. The experts formulated recommendations and formally graded them at a consensus meeting with input from patient representatives and a voting panel of nephrologists representing all regions of the world. Genetic diagnostics comprising joint analysis of COL4A3/4/5 genes is already the key diagnostic test during the initial evaluation of an individual presenting with persistent haematuria, proteinuria, kidney failure of unknown origin, focal segmental sclerosis of unknown origin, and possibly cystic kidney disease. Early renin-angiotensin system blockade is the standard of care therapy; sodium-glucose cotransporter-2 inhibitors may be added in adults with proteinuria and chronic kidney disease. Relatives with heterozygous COL4A3/4/5 variants should only be considered as the last possible resource for living kidney donation. This guideline provides guidance for the diagnosis and management of individuals with pathogenic variants in COL4A3/4/5 genes.

Prognostic outcome models might help to minimise the discard rates of renal transplants from deceased donors. To this end, we published 2-Step Scores for both delayed graft function and transplant loss with optional histology. With conventional paraffin PAS histology taking at least 3 h excluding transport time, we tested whether fast, mobile confocal histology with portable VivaScope® 2500 systems might offer a viable and more rapid alternative. After omitting 17 biopsies with less than 12 glomeruli and 1 artery, we collected 14 0-h and 16 renal transplant indication (Tx) biopsies for a combined cohort. All biopsies were scanned in less than 10 min with a VivaScope® 2500, rendering pseudo-HE images, and then underwent our regular paraffin work-up. Banff Lesion Scores ct and cv were assessed on the granular ordinal scale and binary as (ct ≤ 1 vs. ct ≥ 2 and cv ≤ 2 vs. cv3) together with the number of glomeruli as used in the previously published 2-Step Scores in a blinded fashion by an expert nephropathologist on the paraffin sections (P) and VivaScope® 2500 scans (V). Additionally, we examined the ratio of globally sclerotic glomeruli. Correlation and mixed effects linear regressions, as well as Fleiss’ kappa statistics, comparing P and V on the combined cohort were applied, supplemented with Bland-Altman statistics providing limits of agreement. Between P and V, granular Banff ct correlated with a kappa of 0.513 (p = 8.38e−05) and a Kendall’s W of 0.706 (p = 0.0694) on the combined cohort; binary Banff ct correlated with a kappa of 0.869 (p = 1.92e−06). We had to exclude two more biopsies in which no artery was found in the scanning plane in V. Granular Banff cv correlated with a kappa of 0.109 (p = 0.345) and a W of 0.677 (p = 0.103) between P and V on the combined cohort and binary Banff cv with a kappa of 0.24 (p = 0.204). The total number of glomeruli correlated between P and V with an R of 0.75 (p = 2.3e−06), the number of globally sclerotic glomeruli with an R of 0.82 (p = 4.1e−08), and the ratio thereof with an R of 0.86 (p = 8.6e−10). Instant, decentralised VivaScope® 2500 histology might deliver Banff ct and the number of glomeruli with sufficient accuracy for the 2-Step Scores to predict the risk of delayed graft function and 1-year death-censored transplant loss in deceased heart-beating donors. In contrast, VivaScope® 2500 assessment of Banff cv might not be accurate enough for use in 2-Step Scores.

BACKGROUND:The advent of digital nephropathology offers the potential to integrate deep learning algorithms into the diagnostic workflow. We introduce PICASO, a novel permutation-invariant set operator to dynamically aggregate histopathologic features from instances. We applied PICASO to two nephropathology scenarios: detecting active crescent lesions in sets of glomerular crops with IgA nephropathy and case-level classification for antibody-mediated rejection (AMR) in kidney transplant. METHODS:PICASO is a Transformer-based set operator that aggregates features from sets of instances to make predictions. It utilizes initial Histopathologic Vectors as a static memory component and continuously updates them based on input embeddings. For active crescent detection in IgA nephropathy cases, we obtained 6206 Periodic acid-Schiff-stained (PAS) glomerular crops (5792 no Active Crescent, 414 Active Crescent) from three different health institutes. For the AMR classification, we have 1655 PAS glomerular crops (769 AMR and 886 Non-AMR images) from 89 biopsies. The performance of PICASO as a set operator was compared with other set operators such as DeepSet, Set Transformer, DeepSet++, and Set Transformer++ using metrics including area under the receiver operating characteristic curves (AUROC), area under the precision-recall curves (AUPR), recall, and accuracy. RESULTS:PICASO achieved superior performance in detecting active crescent in IgA nephropathy cases, with an AUROC of 0.99 (95% confidence interval, 0.98 to 0.99) on internal validation and 0.96 (95% confidence interval, 0.95 to 0.98) on external validation, significantly outperforming other set operators (P<0.001). It also attained the highest AUROC of 0.97 (95% confidence interval, 0.90 to 1.0, P=0.02) for case-level AMR classification. The AUPR, recall, and accuracy scores were also higher when using PICASO, and it significantly outperformed baselines (P<0.001). CONCLUSIONS:PICASO can potentially advance nephropathology by improving performance through dynamic feature aggregation.

ABSTRACT Background Early progression of chronic histologic lesions in kidney allografts represents the main finding in graft attrition. The objective of this retrospective cohort study was to elucidate whether HLA histocompatibility is associated with progression of chronic histologic lesions in the first year post-transplant. Established associations of de novo donor-specific antibody (dnDSA) formation with HLA mismatch and microvascular inflammation (MVI) were calculated to allow for comparability with other study cohorts. Methods We included 117 adult kidney transplant recipients, transplanted between 2016 and 2020 from predominantly deceased donors, who had surveillance biopsies at 3 and 12 months. Histologic lesion scores were assessed according to the Banff classification. HLA mismatch scores [i.e. eplet, predicted indirectly recognizable HLA-epitopes algorithm (PIRCHE-II), HLA epitope mismatch algorithm (HLA-EMMA), HLA whole antigen A/B/DR] were calculated for all transplant pairs. Formation of dnDSAs was quantified by single antigen beads. Results More than one-third of patients exhibited a progression of chronic lesion scores by at least one Banff grade in tubular atrophy (ct), interstitial fibrosis (ci), arteriolar hyalinosis (ah) and inflammation in the area of interstitial fibrosis and tubular atrophy (i-IFTA) from the 3- to the 12-month biopsy. Multivariable proportional odds logistic regression models revealed no association of HLA mismatch scores with progression of histologic lesions, except for ah and especially HLA-EMMA DRB1 [odds ratio (OR) = 1.10, 95% confidence interval (CI) 1.03–1.18]. Furthermore, the established associations of dnDSA formation with HLA mismatch and MVI (OR = 5.31, 95% CI 1.19–22.57) could be confirmed in our cohort. Conclusions These data support the association of HLA mismatch and alloimmune response, while suggesting that other factors contribute to early progression of chronic histologic lesions.

Abstract Background and Aims The decision for acceptance or discard of the increasingly rare and marginal deceased donor kidneys in Eurotransplant (ET) countries has to be made without solid evidence. Thus, we developed and validated flexible clinicopathological scores we call 2-Step Scores for the prognosis of delayed graft function (DGF) as 2-Step-DGF and one-year death-censored transplant loss (1y-tl) as 2-Step-1y, reflecting current practice of six ET countries including Croatia and Belgium. Method The training set was n = 620 for DGF and n = 711 for 1y-tl, with validation sets n = 158 and n = 162. In step 1, stepwise logistic regression models including only clinical predictors were used to estimate the risks. In step 2, risk estimates were updated for statistically relevant intermediate risk percentiles with nephropathology. Results Step 1 revealed an increased risk of DGF with increased cold ischaemia time, donor and recipient BMI, dialysis vintage, number of HLA-DR mismatches or recipient CMV IgG positivity. On the training and validation set, c-statistics were 0.672 and 0.704, respectively. At a range between 18% and 36%, accuracy of DGF-prognostication improved with nephropathology including number of glomeruli and Banff cv (updated overall c statistics of 0.696 and 0.701, respectively). Risk of 1y-tl increased in recipients with cold ischaemia time, sum of HLA-A. -B, -DR mismatches and donor age. On training and validation sets, c-statistics were 0.700 and 0.769, respectively. Accuracy of 1y-tl prediction improved (c-statistics = 0.706 and 0.765) with Banff ct. Overall, calibration was good on the training, but moderate on the validation set; discrimination was at least as good as established scores when applied to the validation set. Receiver operating characteristics of both 2-Step-DGF and 2-Step-1y compared to established scores are shown in Figs 1 and 2. Conclusion Our flexible 2-Step Scores with optional inclusion of time-consuming and often unavailable nephropathology should yield good results for clinical practice in ET, and may be superior to established scores. Our scores are adaptable to donation after cardiac death and perfusion pumps use.

Introduction: Kidneys of marginal quality are increasingly being used to overcome the shortage of donor organs. However, accurate prediction of outcome is needed to optimize the use of these kidneys. We aimed to test the performance of a recently proposed score consisting of delayed graft function (DGF), renal function recovery (RFR), and glomerular filtration rate (GFR) < 30 mL/min per 1.73 m2 90 days after transplantation for risk assessment of patient and graft survival. Material and Methods: A total of 221 adult brain death donors with marginal kidneys transplanted into 223 recipients within Eurotransplant were included in the analysis. Multivariable Cox proportional hazards models were constructed to assess death-censored and all-cause censored graft failure and recipient mortality at 1 and 3 years. Results: Recipients with DGF had a higher risk of death-censored graft loss (HR, 95% CIs: 3.058 [1.195-7.825]). Recipients with a GFR < 30 mL/min/1.73 m2 at 90 days after transplantation had a higher risk of death censored and all- cause graft failure (HR, 95% CIs: 2.122 [1.129-3.990] and 2.122 [1.129-3.990]). None of the three components of the proposed score was associated with a higher risk of mortality. Conclusion: DGF and estimated GFR < 30 mL/min/ 1.73 m2 but not RFR at 90 days predicted graft failure after transplantation of marginal kidneys. However, no combination of these factors was able to predict short-term patient and graft survival. (c) 2024 The Author(s). Published by S. Karger AG, Basel

Abstract Background and Aims Diagnostic applications of machine learning in nephropathology are only beginning to emerge. We hypothesized that we could develop a machine learning classifier for 12 different classes of glomerulonephritis with CNN and self-attention-based architectures, following the nephropathology paradigm that globally sclerosed glomeruli are not useful for diagnostic purposes. Method The dataset contains 11,000 PAS-stained glomerular crops from 350 biopsies (four institutions). Each crop retained the diagnosis label from the 12 classes ABMGN, ANCA, C3-GN, CryoGN, DDD, Fibrillary, infection-associated GN (IAGN), IgAGN, MPGN, Membranous, PGNMID, SLEGN-IV; globally sclerotic glomerular crops were stripped from this diagnostic label and were just labeled as the 13th class Sclerotic. This dataset was divided into 75% of samples for training, 15% for validation and 10% testing, avoiding information leak within biopsies. Moreover, a hold-out validation set of another 50 biopsies with 2,000 new crops that were taken from another three centres. A classifier was trained in a fully supervised fashion for 13 classes (12 GN classes and Sclerotic), based on an ensemble of multiple transformer-based classification networks, including Swin-Transformer and ConvNext. This allows classification of each glomerular crop by different network instances. Since each network was trained under different conditions, the whole system acquired a more global knowledge understanding rather than a relying on a single method. For the final decision, the system takes the prediction with the largest confidence threshold, which is also predicted along with the class. Results The Table and the Figure list the metrics for classification performance calculated as Precision, Sensitivity, Specificity, F1 Score and balanced Accuracy. Balanced accuracy was between 0.4797 for CryoGN and 0.5949 for Membranous, it was 0.6892 for Sclerotic. AUCs for ROCs were between 0.40 for PGNMID and 0.82 for Membranous, with 0.81 for Sclerotic. Conclusion This proof-of-concept study establishes a baseline for this challenging classification task, which usually requires immunostains, electron microscopy and even clinical data. Our classification results even on single PAS glomerular crops appear promising. Combined with our automatic glomerular segmentation models, we could rapidly expand the training cohorts sizes and even add more classes of GN.

Pre-transplant procurement biopsy interpretation is challenging, also because of the low number of renal pathology experts. Artificial intelligence (AI) can assist by aiding pathologists with kidney donor biopsy assessment. Herein we present the “Galileo” AI tool, designed specifically to assist the on-call pathologist with interpreting pre-implantation kidney biopsies. A multicenter cohort of whole slide images acquired from core-needle and wedge biopsies of the kidney was collected. A deep learning algorithm was trained to detect the main findings evaluated in the pre-implantation setting (normal glomeruli, globally sclerosed glomeruli, ischemic glomeruli, arterioles and arteries). The model obtained on the Aiforia Create platform was validated on an external dataset by three independent pathologists to evaluate the performance of the algorithm. Galileo demonstrated a precision, sensitivity, F1 score and total area error of 81.96

Complement 3 glomerulopathy (C3G) and immune complex membranoproliferative glomerulonephritis (IC-MPGN) are ultra-rare chronic kidney diseases with an overall poor prognosis, with approximately 40–50

Kidney biopsies are routinely used for diagnostic and prognostic purposes but their utility in the intensive care unit (ICU) setting is limited. We investigated the associations of clinical and histopathological risk factors with ICU-acute kidney injury (AKI) in donors with brain death (DBD) with kidneys of lower quality and procurement biopsies. Overall, 221 donors with brain death, 239 biopsies and 197 recipients were included. The biopsies were reread and scored according to the Banff recommendations. Clinical and histopathological data were compared between donors with and without AKI defined by serum creatinine and by urine output. Logistic regression analysis was applied to identify independent clinical and histopathological risk factors for both phenotypes. Lastly, the impact of each AKI phenotype on outcome was explored. AKI was diagnosed based on the RIFLE (Risk, Injury, Failure, Loss of function, End-stage kidney disease) AKIN (Acute Kidney Injury Network) or KDIGO (Kidney Disease Improving Global Outcomes) criteria. Acute kidney injury occurred in 65

Background The decision to accept or discard the increasingly rare and marginal brain-dead donor kidneys in Eurotransplant (ET) countries has to be made without solid evidence. Thus, we developed and validated flexible clinicopathological scores called 2-Step Scores for the prognosis of delayed graft function (DGF) and 1-year death-censored transplant loss (1y-tl) reflecting the current practice of six ET countries including Croatia and Belgium.Methods The training set was n = 620 for DGF and n = 711 for 1y-tl, with validation sets n = 158 and n = 162, respectively. In Step 1, stepwise logistic regression models including only clinical predictors were used to estimate the risks. In Step 2, risk estimates were updated for statistically relevant intermediate risk percentiles with nephropathology.Results Step 1 revealed an increased risk of DGF with increased cold ischaemia time (CIT), donor and recipient body mass index, dialysis vintage, number of HLA-DR mismatches or recipient cytomegalovirus immunoglobulin G positivity. On the training and validation set, c-statistics were 0.672 and 0.704, respectively. At a range between 18% and 36%, accuracy of DGF-prognostication improved with nephropathology including number of glomeruli and Banff cv (updated overall c-statistics of 0.696 and 0.701, respectively). Risk of 1y-tl increased in recipients with CIT, sum of HLA-A, -B, -DR mismatches, and donor age. On training and validation sets, c-statistics were 0.700 and 0.769, respectively. Accuracy of 1y-tl prediction improved (c-statistics = 0.706 and 0.765) with Banff ct. Overall, calibration was good on the training, but moderate on the validation set; discrimination was at least as good as established scores when applied to the validation set.Conclusion Our flexible 2-Step Scores with optional inclusion of time-consuming and often unavailable nephropathology should yield good results for clinical practice in ET, and may be superior to established scores. Our scores are adaptable to donation after cardiac death and perfusion pump use.

Living kidney donors are screened pre-donation to estimate the risk of end-stage kidney disease (ESKD). We evaluate Machine Learning (ML) to predict the progression of kidney function deterioration over time using the estimated GFR (eGFR) slope as the target variable. We included 238 living kidney donors who underwent donor nephrectomy. We divided the dataset based on the eGFR slope in the third follow-up year, resulting in 185 donors with an average eGFR slope and 53 donors with an accelerated declining eGFR-slope. We trained three Machine Learning-models (Random Forest [RF], Extreme Gradient Boosting [XG], Support Vector Machine [SVM]) and Logistic Regression (LR) for predictions. Predefined data subsets served for training to explore whether parameters of an ESKD risk score alone suffice or additional clinical and time-zero biopsy parameters enhance predictions. Machine learning-driven feature selection identified the best predictive parameters. None of the four models classified the eGFR slope with an AUC greater than 0.6 or an F1 score surpassing 0.41 despite training on different data subsets. Following machine learning-driven feature selection and subsequent retraining on these selected features, random forest and extreme gradient boosting outperformed other models, achieving an AUC of 0.66 and an F1 score of 0.44. After feature selection, two predictive donor attributes consistently appeared in all models: smoking-related features and glomerulitis of the Banff Lesion Score. Training machine learning-models with distinct predefined data subsets yielded unsatisfactory results. However, the efficacy of random forest and extreme gradient boosting improved when trained exclusively with machine learning-driven selected features, suggesting that the quality, rather than the quantity, of features is crucial for machine learning-model performance. This study offers insights into the application of emerging machine learning-techniques for the screening of living kidney donors.