Observer agreement in the scoring of the activity and chronicity indexes of lupus nephritis

Automated Prediction of Kidney Failure in IgA Nephropathy with Deep Learning from Biopsy Images

BACKGROUND AND OBJECTIVES

Digital pathology and artificial intelligence offer new opportunities for automatic histologic scoring. We applied a deep learning approach to IgA nephropathy biopsy images to develop an automatic histologic prognostic score, assessed against ground truth (kidney failure) among patients with IgA nephropathy who were treated over 39 years. We assessed noninferiority in comparison with the histologic component of currently validated predictive tools. We correlated additional histologic features with our deep learning predictive score to identify potential additional predictive features.

DESIGN, SETTING, PARTICIPANTS, & MEASUREMENTS

Training for deep learning was performed with randomly selected, digitalized, cortical Periodic acid-Schiff-stained sections images (363 kidney biopsy specimens) to develop our deep learning predictive score. We estimated noninferiority using the area under the receiver operating characteristic curve (AUC) in a randomly selected group (95 biopsy specimens) against the gold standard Oxford classification (MEST-C) scores used by the International IgA Nephropathy Prediction Tool and the clinical decision supporting system for estimating the risk of kidney failure in IgA nephropathy. We assessed additional potential predictive histologic features against a subset (20 kidney biopsy specimens) with the strongest and weakest deep learning predictive scores.

RESULTS

We enrolled 442 patients; the 10-year kidney survival was 78%, and the study median follow-up was 6.7 years. Manual MEST-C showed no prognostic relationship for the endocapillary parameter only. The deep learning predictive score was not inferior to MEST-C applied using the International IgA Nephropathy Prediction Tool and the clinical decision supporting system (AUC of 0.84 versus 0.77 and 0.74, respectively) and confirmed a good correlation with the tubolointerstitial score (r=0.41, P<0.01). We observed no correlations between the deep learning prognostic score and the mesangial, endocapillary, segmental sclerosis, and crescent parameters. Additional potential predictive histopathologic features incorporated by the deep learning predictive score included (1) inflammation within areas of interstitial fibrosis and tubular atrophy and (2) hyaline casts.

CONCLUSIONS

The deep learning approach was noninferior to manual histopathologic reporting and considered prognostic features not currently included in MEST-C assessment.

PODCAST

This article contains a podcast at https://www.asn-online.org/media/podcast/CJASN/2022_07_26_CJN01760222.mp3.

Multi-Task Learning-Based Immunofluorescence Classification of Kidney Disease

Chronic kidney disease is one of the most important causes of mortality worldwide, but a shortage of nephrology pathologists has led to delays or errors in its diagnosis and treatment. Immunofluorescence (IF) images of patients with IgA nephropathy (IgAN), membranous nephropathy (MN), diabetic nephropathy (DN), and lupus nephritis (LN) were obtained from the General Hospital of Chinese PLA. The data were divided into training and test data. To simulate the inaccurate focus of the fluorescence microscope, the Gaussian method was employed to blur the IF images. We proposed a novel multi-task learning (MTL) method for image quality assessment, de-blurring, and disease classification tasks. A total of 1608 patients’ IF images were included—1289 in the training set and 319 in the test set. For non-blurred IF images, the classification accuracy of the test set was 0.97, with an AUC of 1.000. For blurred IF images, the proposed MTL method had a higher accuracy (0.94 vs. 0.93, p < 0.01) and higher AUC (0.993 vs. 0.986) than the common MTL method. The novel MTL method not only diagnosed four types of kidney diseases through blurred IF images but also showed good performance in two auxiliary tasks: image quality assessment and de-blurring.

Kidney Level Lupus Nephritis Classification Using Uncertainty Guided Bayesian Convolutional Neural Networks

The kidney biopsy based diagnosis of Lupus Nephritis (LN) is characterized by low inter-observer agreement, with misdiagnosis being associated with increased patient morbidity and mortality. Although various Computer Aided Diagnosis (CAD) systems have been developed for other nephrohistopathological applications, little has been done to accurately classify kidneys based on their kidney level Lupus Glomerulonephritis (LGN) scores. The successful implementation of CAD systems has also been hindered by the diagnosing physician's perceived classifier strengths and weaknesses, which has been shown to have a negative effect on patient outcomes. We propose an Uncertainty-Guided Bayesian Classification (UGBC) scheme that is designed to accurately classify control, class I/II, and class III/IV LGN (3 class) at both the glomerular-level classification task (26,634 segmented glomerulus images) and the kidney-level classification task (87 MRL/lpr mouse kidney sections). Data annotation was performed using a high throughput, bulk labeling scheme that is designed to take advantage of Deep Neural Network's (or DNNs) resistance to label noise. Our augmented UGBC scheme achieved a 94.5% weighted glomerular-level accuracy while achieving a weighted kidney-level accuracy of 96.6%, improving upon the standard Convolutional Neural Network (CNN) architecture by 11.8% and 3.5% respectively.

Evaluation of the Classification Accuracy of the Kidney Biopsy Direct Immunofluorescence through Convolutional Neural Networks

BACKGROUND AND OBJECTIVES

Immunohistopathology is an essential technique in the diagnostic workflow of a kidney biopsy. Deep learning is an effective tool in the elaboration of medical imaging. We wanted to evaluate the role of a convolutional neural network as a support tool for kidney immunofluorescence reporting.

DESIGN, SETTING, PARTICIPANTS, & MEASUREMENTS

High-magnification (×400) immunofluorescence images of kidney biopsies performed from the year 2001 to 2018 were collected. The report, adopted at the Division of Nephrology of the AOU Policlinico di Modena, describes the specimen in terms of "appearance," "distribution," "location," and "intensity" of the glomerular deposits identified with fluorescent antibodies against IgG, IgA, IgM, C1q and C3 complement fractions, fibrinogen, and κ- and λ-light chains. The report was used as ground truth for the training of the convolutional neural networks.

RESULTS

In total, 12,259 immunofluorescence images of 2542 subjects undergoing kidney biopsy were collected. The test set analysis showed accuracy values between 0.79 ("irregular capillary wall" feature) and 0.94 ("fine granular" feature). The agreement test of the results obtained by the convolutional neural networks with respect to the ground truth showed similar values to three pathologists of our center. Convolutional neural networks were 117 times faster than human evaluators in analyzing 180 test images. A web platform, where it is possible to upload digitized images of immunofluorescence specimens, is available to evaluate the potential of our approach.

CONCLUSIONS

The data showed that the accuracy of convolutional neural networks is comparable with that of pathologists experienced in the field.

A Systematic Review of Interpathologist Agreement in Histologic Classification of Lupus Nephritis

Introduction

Lupus nephritis (LN) is one of the most severe manifestations of systemic lupus erythematosus (SLE), resulting in increased morbidity and mortality. The gold standard for diagnosis of LN is a renal biopsy. Considering the importance of the biopsy in determining long-term prognostication and treatment decisions, it is crucial to assess renal histopathology with utmost accuracy and precision. This review represents a systematic search of published literature to estimate the degree of interpathologist reproducibility in current assessment of LN.

Methods

Using the PubMed and Google Scholar search engines, studies analyzing the agreement of 4 or more pathologists assessing LN slides using the ISN/Renal Pathology Society (RPS) classification, activity index, and chronicity index were selected for analysis in this systematic review.

Results

In reviewing 6 qualifying studies (those analyzing the agreement of 4 or more pathologists using the ISN/RPS classification, activity index, and chronicity index) for the assignment of ISN/RPS class was 0.325 (interquartile range [IQR] 0.2405–0.425), which is “poor.” The median interpathologist concordance values for the assigned activity index and chronicity index were “moderate”: 0.52 (IQR 0.51–0.69) and 0.49 (IQR 0.36–0.58), respectively.

Conclusion

Thus, the current scoring using the ISN/RPS classification system and activity and chronicity indices for LN exhibits poor interpathologist agreement, which limits its use in clinical practice. Given that this can have severe repercussions on a patient’s treatment and prognosis, efforts to update pathology assessment guidelines, objectively measurable biomarkers, and deep learning approaches are strongly warranted.

Inter-observer variability of the histological classification of lupus glomerulonephritis in children

The gold standard for the classification of lupus nephritis is renal histology but reporting variation exists. The aim of this study was to assess the inter-observer variability of the 2003 International Society of Nephrology/Royal Pathology Society (ISN/RPS) lupus nephritis histological classification criteria in children. Histopathologists from a reference centre and three tertiary paediatric centres independently reviewed digitalized renal histology slides from 55 children with lupus nephritis. Histological ISN/RPS Class was assigned and features scored; lupus nephritis-activity [scored 0-24], lupus nephritis-chronicity [0-12] and tubulointerstitial activity [0-21]. In the cohort (73% females), the age at the time of biopsy was 15.5 ± 0.39 (mean ± standard error) years. Based on the reference centre, 42% (23/55) had ISN/RPS Class IV with lupus nephritis-activity score 4.23 ± 0.50, lupus nephritis-chronicity 1.81 ± 0.18 and tubulointerstitial activity 4.45 ± 0.35. There were 4-54 (mean 16.7) glomeruli per biopsy. Pathologists had fair agreement for ISN/RPS assignment (kappa; 0.26 ± 0.12), lupus nephritis-chronicity (intra-class correlation 0.36 ± 0.09) and tubulointerstitial activity (0.22 ± 0.09) scores. There was good agreement for lupus nephritis-activity scores (intra-class correlation 0.69 ± 0.06). When categorized into proliferative and non-proliferative disease, poor agreement among sites remained (kappa 0.24 ± 0.11). Despite unified criteria for the interpretation of histological features of lupus nephritis, marked reporting variation remains in clinical practice. As proliferative lupus nephritis is managed more intensively, this may influence renal outcomes.

Systematic review of the literature on reproducibility of the interpretation of renal biopsy in lupus nephritis

Objective Before using a test, it should be determined whether the results are reliable. The reliability of the interpretation of renal biopsy in patients with lupus nephritis has not been clearly elucidated. Our objective was to estimate inter and intra-observer reliability of the histological classification, as well as activity and chronicity indices in renal biopsy of patients with lupus nephritis. Methods We conducted a systematic search of the literature, which included articles in any language, using PubMed, Embase, Cochrane and Lilacs databases. Search terms included were: reproducibility, reliability, agreement, systemic lupus erythematosus and lupus nephritis. Comparative studies with any design were included, regardless of the year or the language of publication. Two investigators, independently, screened the literature published in accordance with pre-established inclusion and exclusion criteria. Results We found 13 relevant studies. Inter-observer reproducibility of most measurements was moderate or low, despite the fact that, in most cases, the readings were made by expert nephropathologists. There was great diversity among designs, participants, including samples and outcomes evaluated in different studies. Although there are too many reports on the clinical use, studies evaluating the reliability of classifications on renal biopsy in lupus nephritis are rare. The quality of the methodological design and reporting was fair. Conclusion The interpretation of renal biopsy in lupus nephritis is poorly reproducible, causing serious doubts about its validity and its clinical application. As it can lead to serious diagnosis, treatment and prognosis errors, it is necessary to intensify research in this field.

Concise semiquantitative histological scoring system for immunoglobulin A nephropathy

AIM

Immunoglobulin A nephropathy (IgAN) is a common and progressive glomerulonephritis. Histological lesions of IgAN are variable and considered as a risk factor for renal outcome. Establishing a relatively concise histological semiquantitative scoring system would be valuable in clinical practice.

METHODS

Renal biopsy sections of 293 patients with primary IgAN from two centres in China were reviewed. A histological scoring system was established based on multivariate survival analysis of semiquantitative histological indices, using end-stage renal disease (ESRD) as the end-point event.

RESULTS

Four indices--extracapillary glomerular activity index (exGAI), mesangial proliferation index (MsI), glomerular chronicity index (GCI) and tubulointerstitial chronicity index (TCI)--independently correlated with ESRD (relative risk (RR) = 1.16, 2.27, 1.29 and 1.80, respectively). The four indices and the sum of their scores (Total I) constituted the scoring system. Patients with exGAI of 4 or more, GCI of 4 or more, MsI of 2 or more and TCI of 2 or more were considered as having a higher risk for progression (P < 0.05). A score of 4 or more on the exGAI index could identify the patients who should be treated with immunosuppressive drugs, which showed a lower incidence of ESRD than that without (24.3% vs 52.4%, P = 0.031).

CONCLUSION

It suggested that the present concise scoring system could serve as prognostic and therapeutic indications.

Evaluation of activity, chronicity and tubulointerstitial indices for childhood lupus nephritis

Few data exist on use of the National Institutes of Health (NIH) activity index (AI) and chronicity index (CI) in childhood lupus nephritis (LN). A tubulointerstitial activity index (TIAI) has been derived but not validated. We evaluated clinicopathologic correlations of the AI, CI and TIAI in children with LN who had undergone initial renal biopsy (n=25, age 12.4+/-2.7 years, biopsy 1) and 1 year after treatment (n=15, biopsy 2). The TIAI correlated with the AI at biopsy 1 (r=0.76, P=0.001) and biopsy 2 (r=0.52, P=0.05), but not with CI scores. Mean AI and CI scores changed substantially from biopsy 1 to biopsy 2 (P<0.05), but TIAI scores did not. Higher AI and TIAI scores correlated with proteinuria at both biopsies (r=0.51-0.76, P<0.05); CI scores correlated with estimated creatinine clearance (r=0.46-0.58, P<0.05). Improved AI score from biopsy 1 to biopsy 2 was associated with decrease in proteinuria. These results suggest that the AI and CI are useful in childhood LN. The TIAI may be a valid measure to evaluate the tubulointerstitium, but research is needed to define its responsiveness to change with therapy.

Lupus Nephritis Outcome With and Without Renal Biopsy

BACKGROUND

: Renal biopsy is an important tool in devising an adequate treatment plan for lupus nephritis. However, it is not always possible to perform a biopsy, and in many cases, treatment must rely exclusively on clinical data.

OBJECTIVE

: The aim of this study was to compare the 5-year course of patients treated without a biopsy with another group with histologic evidence of diffuse proliferative glomerulonephritis (DPGN).

METHODS

: The no-biopsy group consisted of 30 patients with lupus with strong clinical and laboratory suspicion of proliferative glomerulonephritis in whom a renal biopsy was unavailable either because of medical contraindication or the patient's refusal. The biopsy group included 30 patients undergoing biopsy and a histologic diagnosis of DPGN. Patients were followed from the onset of nephritis and at 18, 36, and 60 months.

RESULTS

: At onset, the no-biopsy group showed lower C3 levels and higher proteinuria, although both groups showed evident deterioration of the renal function. No significant differences were found in treatment, outcome, survival, renal function tests, or in the development of kidney failure.

CONCLUSIONS

: Proliferative glomerulonephritis deserves prompt diagnosis and treatment. This study demonstrates that experience in the management of lupus nephropathy, together with clinical and laboratory data, are often enough information to adequately treat proliferative glomerulonephritis even in the absence of a renal biopsy.

A new morphologic index for the evaluation of renal biopsies in lupus nephritis

BACKGROUND

Various morphologic indices for the evaluation of renal biopsies in lupus nephritis have been developed, of which the most successful have been the NIH Activity Index (AI) and Chronicity Index (CI). We wished to develop a biopsy index from standard light and immunofluorescence (IF) material that would correlate yet more closely with clinical and outcome parameters than the current indices, and be applicable to both treated and untreated cases.

METHODS

A cohort of 71 patients with lupus nephritis who had initial renal biopsies (Bx1) with systematic second biopsies (Bx2) at six months after induction therapy was studied, with a large number of light microscopic and IF variables evaluated. These were examined statistically to choose the combinations of variables with the highest overall correlations with clinical and outcome parameters.

RESULTS

The adopted biopsy index comprised four elements: Glomerular Activity Index (GAI), a modification of the standard AI with the addition of glomerular monocytes and elimination of interstitial inflammation; Tubulointerstitial Activity Index (TIAI), evaluating several tubular epithelial and inflammatory components, including interstitial inflammation, but excluding tubular atrophy; Chronic Lesions Index, a modification of the standard CI, with the addition of glomerular scars; IF Index (IFI), a semiquantitative index of IF staining for six standard antisera for glomerular capillary, mesangial, tubulointerstitial, and vascular elements. The Biopsy Index showed a statistically higher correlation with clinical and outcome parameters than the NIH AI (P = 0.0170), the NIH CI (P = 0.0009), or their combination (P = 0.0444). At Bx1, comparisons between correlation coefficients for the appropriate AI or CI value and for the Biopsy Index, were: anti-DNA antibodies (0.30 vs. 045), serum creatinine (SCr; 0.33 vs. 0.48), proteinuria (0.22 vs. 0.36), hemoglobin (-0.21 vs. -0.45), and final renal function (0.22 vs. 0.40). Spearman rank correlations showed similar superiority for outcome parameters: doubling of SCr (0.1810 vs. 0.3018) and end-stage renal disease (0.0529 vs. 0.1925). The same improvement of correlations was seen at Bx2 for most parameters, particularly doubling of SCr (0.2716 vs. 0.4753).

CONCLUSIONS

The Biopsy Index and/or its components show better correlations with clinical and outcome parameters than the standard AI and CI and other similar indices.

Value of gallium-67 scintigraphy in monitoring the renal activity in lupus nephritis

In this study, we evaluated the relationship between the gallium uptake in the kidneys and the results of renal biopsy. The renal uptake of gallium in 32 patients with lupus nephritis were evaluated according to the classification of the World Health Organization, the activity index (AI), and the chronicity index (CI). AI was significantly higher in patients with a positive gallium scan when compared with patients with a negative gallium scan (p = 0.0007). However, change in gallium uptake was not related to the CI or WHO classification (p = 0.68 and p = 0.79 respectively). Of the 16 patients with AI > or = 9, 13 (81.25%) had a positive gallium scan, while only 3 (18.7%) of the patients with AI < 9 had a positive scan. In addition, all patients (100%) with AI > or = 11 showed a positive gallium scan. We conclude that renal imaging with gallium has a good correlation with AI, and may be a useful alternative to renal biopsy in patients with lupus nephritis.

Factors associated with chronic renal failure in 121 patients with diffuse proliferative lupus nephritis: a case-control study

Lupus nephritis remains an important problem in patients with systemic lupus erythematosus (SLE). Some patients with diffuse proliferative lupus nephritis (DPLN) develop chronic renal failure (CRF). A case-control study was designed to determine the variables associated with CRF in patients with DPLN. We studied 121 patients with biopsy-proven DPLN seen in our institution from 1970 to 1988. There were 34 patients who developed CRF, the remaining were their controls. Clinical charts were reviewed and a pathologist re-scored blindly both activity and chronicity indices. The mean of age at SLE onset was 24.1 +/- 7.9 years; the mean disease duration was 9.2 +/- 6.1 years for controls and 6.1 +/- 5 years for patients. The main variables associated with CRF were male sex. HR (hazard ratio): 12.6 (95% CI 1.6-98.2); activity index, HR 2.59 (1.07-6.3); severe infections, HR 2.9 (1.2-7.3): number of antihypertensive drugs, HR 2.5 (1.4-4.7); cellular crescents, HR 1.6 (1.2-2): and interstitial inflammation, HR 2.7 (1.5-5.1). A protective effect was observed with longer use of < or = 20 mg of prednisone, HR 0.53 (95% CI 0.34-08): azathioprine, HR 0.6 (0.4-0.8); and length of formal education. HR 0.3 (0.09-0.94). Our results indicate that maleness, activity index, extracapillary proliferation and interstitial inflammation, as well as hypertension and severe infections associate with CRF in patients with DPLN, and treatment and higher education, perhaps through better therapeutic compliance, may be protective.

Irreproducibility of the activity and chronicity indices limits their utility in the management of lupus nephritis. Lupus Nephritis Collaborative Study Group

Histological indices of renal pathology in lupus nephritis have been proposed as prognostic signs and as a method to quantitate therapeutic response. We tested the reproducibility of the activity (AI) and chronicity indices (CI) in the renal biopsies from 83 patients with lupus nephritis, enrolled in a controlled therapeutic trial. The AI/CI were calculated separately by four renal pathologists using published criteria. Pair-wise Spearman's rank correlation coefficient was used to examine the relationship among the scores of the four raters, and their degree of reproducibility was evaluated using the coefficient of reliability. The mean CI scores ranged from 2.84 to 4.61, and the mean AI ranged from 9.64 to 12.89. The correlation among the different pathologist's scores ranged from 0.44 to 0.63 for the AI and 0.60 to 0.76 for the CI. One pathologist (M.M.S.) rated the biopsies twice, and the correlation between the two ratings was 0.58 for the AI and 0.74 for the CI. Thus, the AI and CI calculated by different pathologists and the temporally separate observations of a single observer were only moderately correlated. The reproducibility of a single rating was low, showing a reliability coefficient of 0.48 for the AI and 0.57 for the CI. The low reliability coefficient suggests that the variability among pathologists was the result of interpretative differences. We conclude that the AI/CI are too subjective to be used as therapeutic guides or as prognosticators.