Automated detection algorithm for C4d immunostaining showed comparable diagnostic performance to pathologists in renal allograft biopsy

[Artificial intelligence in kidney transplant pathology]

BACKGROUND

Artificial intelligence (AI) systems have showed promising results in digital pathology, including digital nephropathology and specifically also kidney transplant pathology.

AIM

Summarize the current state of research and limitations in the field of AI in kidney transplant pathology diagnostics and provide a future outlook.

MATERIALS AND METHODS

Literature search in PubMed and Web of Science using the search terms "deep learning", "transplant", and "kidney". Based on these results and studies cited in the identified literature, a selection was made of studies that have a histopathological focus and use AI to improve kidney transplant diagnostics.

RESULTS AND CONCLUSION

Many studies have already made important contributions, particularly to the automation of the quantification of some histopathological lesions in nephropathology. This likely can be extended to automatically quantify all relevant lesions for a kidney transplant, such as Banff lesions. Important limitations and challenges exist in the collection of representative data sets and the updates of Banff classification, making large-scale studies challenging. The already positive study results make future AI support in kidney transplant pathology appear likely.

Banff-based histologic chronicity index is associated with graft failure but has poor interobserver reproducibility

BACKGROUND

Antibody-mediated rejection (AMR) often leads to chronic kidney allograft damage and is a critical cause of allograft failure. The Banff classification, used to diagnose AMR, has become complex and challenging for clinicians. A Banff-based histologic chronicity index (CI) was recently proposed as a simplified prognostic indicator. Its reliability and reproducibility have not been externally validated.

METHODS

This study investigated 71 kidney allograft biopsies diagnosed with AMR. Interobserver reproducibility of the recently proposed CI and its components (cg, cv, ct, and ci) were assessed. The association between CI and allograft failure was analyzed, and CI cut-off values were evaluated by Cox proportional hazards regression and Kaplan-Meier estimator with log-rank test.

RESULTS

The study confirmed the association of CI with allograft failure, but also revealed that the assessment of CI varied between pathologists, impacting its reproducibility as a prognostic tool. Only 49 (69.0%) of the biopsies showed complete agreement on the proposed cut-off value of CI < 4 or CI ≥ 4. Furthermore, this cut-off did not reliably stratify allograft failure. Notably, the cg score, which carries significant weight in the CI calculation, had the lowest agreement between observers (kappa = .281).

CONCLUSIONS

While a simplified prognostic indicator for AMR is needed, this study highlights the limitations of CI, particularly its poor interobserver reproducibility. Our findings suggest that clinicians should interpret CI cautiously and consider establishing their own cut-off values. This study underscores the need to address interobserver reproducibility before CI can be widely adopted for AMR management.

Time for a full digital approach in nephropathology: a systematic review of current artificial intelligence applications and future directions

INTRODUCTION

Artificial intelligence (AI) integration in nephropathology has been growing rapidly in recent years, facing several challenges including the wide range of histological techniques used, the low occurrence of certain diseases, and the need for data sharing. This narrative review retraces the history of AI in nephropathology and provides insights into potential future developments.

METHODS

Electronic searches in PubMed-MEDLINE and Embase were made to extract pertinent articles from the literature. Works about automated image analysis or the application of an AI algorithm on non-neoplastic kidney histological samples were included and analyzed to extract information such as publication year, AI task, and learning type. Prepublication servers and reviews were not included.

RESULTS

Seventy-six (76) original research articles were selected. Most of the studies were conducted in the United States in the last 7 years. To date, research has been mainly conducted on relatively easy tasks, like single-stain glomerular segmentation. However, there is a trend towards developing more complex tasks such as glomerular multi-stain classification.

CONCLUSION

Deep learning has been used to identify patterns in complex histopathology data and looks promising for the comprehensive assessment of renal biopsy, through the use of multiple stains and virtual staining techniques. Hybrid and collaborative learning approaches have also been explored to utilize large amounts of unlabeled data. A diverse team of experts, including nephropathologists, computer scientists, and clinicians, is crucial for the development of AI systems for nephropathology. Collaborative efforts among multidisciplinary experts result in clinically relevant and effective AI tools.

Progress in Pathological Diagnosis after Kidney Transplantation: Current Trend and Future Perspective

Advances in immunosuppressive therapy; posttransplant management of allograft rejection; and measures against infectious diseases, cardiovascular diseases, and malignancy dramatically improved graft and patient survival after kidney transplantation (KT). Among them, kidney allograft biopsy is an important tool and the gold standard for the diagnosis of various kidney allograft injuries, including allograft rejection, virus-induced nephropathy, calcineurin inhibitor toxicity, and posttransplant glomerular diseases. The Banff Conference on Allograft Pathology has contributed to establishing the diagnostic criteria for kidney allograft rejection and polyomavirus-associated nephropathy that are used as a common standard worldwide. In addition to the for-cause biopsy, many transplant centers perform protocol biopsies in the early and late posttransplant periods to detect and treat allograft injury earlier. Preimplantation biopsy in deceased-donor KT has also been performed, especially in the marginal donor, and attempts have been made to predict the prognosis in combination with clinical information and the renal resistance of hypothermic machine perfusion. Regarding the preimplantation biopsy from a living kidney donor, it can provide useful information on aging and/or early changes in lifestyle diseases, such as glomerulosclerosis, tubulointerstitial changes, and arterial and arteriolar sclerosis, and be used as a reference for the subsequent management of living donors. In this review, morphologic features of important kidney allograft pathology, such as allograft rejection and polyomavirus-associated nephropathy, according to the latest Banff classification and additional information derived from protocol biopsy, and future perspectives with recently developed technologies are discussed.

WARS1, TYMP and GBP1 display a distinctive microcirculation pattern by immunohistochemistry during antibody-mediated rejection in kidney transplantation

Antibody-mediated rejection (ABMR) is the leading cause of allograft failure in kidney transplantation. Defined by the Banff classification, its gold standard diagnosis remains a challenge, with limited inter-observer reproducibility of the histological scores and efficient immunomarker availability. We performed an immunohistochemical analysis of 3 interferon-related proteins, WARS1, TYMP and GBP1 in a cohort of kidney allograft biopsies including 17 ABMR cases and 37 other common graft injuries. Slides were interpreted, for an ABMR diagnosis, by four blinded nephropathologists and by a deep learning framework using convolutional neural networks. Pathologists identified a distinctive microcirculation staining pattern in ABMR with all three antibodies, displaying promising diagnostic performances and a substantial reproducibility. The deep learning analysis supported the microcirculation staining pattern and achieved similar diagnostic performance from internal validation, with a mean area under the receiver operating characteristic curve of 0.89 (± 0.02) for WARS1, 0.80 (± 0.04) for TYMP and 0.89 (± 0.04) for GBP1. The glomerulitis and peritubular capillaritis scores, the hallmarks of histological ABMR, were the most highly correlated Banff scores with the deep learning output, whatever the C4d status. These novel immunomarkers combined with a CNN framework could help mitigate current challenges in ABMR diagnosis and should be assessed in larger cohorts.