Artificial Intelligence You Can Trust: What Matters Beyond Performance When Applying Artificial Intelligence to Renal Histopathology?

Although still in its infancy, artificial intelligence (AI) analysis of kidney biopsy images is anticipated to become an integral aspect of renal histopathology. As these systems are developed, the focus will understandably be on developing ever more accurate models, but successful translation to the clinic will also depend upon other characteristics of the system.In the extreme, deployment of highly performant but "black box" AI is fraught with risk, and high-profile errors could damage future trust in the technology. Furthermore, a major factor determining whether new systems are adopted in clinical settings is whether they are "trusted" by clinicians. Key to unlocking trust will be designing platforms optimized for intuitive human-AI interactions and ensuring that, where judgment is required to resolve ambiguous areas of assessment, the workings of the AI image classifier are understandable to the human observer. Therefore, determining the optimal design for AI systems depends on factors beyond performance, with considerations of goals, interpretability, and safety constraining many design and engineering choices.In this article, we explore challenges that arise in the application of AI to renal histopathology, and consider areas where choices around model architecture, training strategy, and workflow design may be influenced by factors beyond the final performance metrics of the system.

Signaling through tumor necrosis receptor 2 induces stem cell marker in CD133+ regenerating tubular epithelial cells in acute cell-mediated rejection of human renal allografts.

Tumor necrosis factor receptor 2 (TNFR2) is strongly upregulated on renal tubular epithelial cells by acute cell-mediated rejection (ACR. In human kidney organ culture, TNFR2 signaling both upregulates TNFR2 expression and promotes cell cycle entry of tubular epithelial cells. We find significantly more cells express CD133 mRNA and protein, a putative stem cell marker, in allograft biopsy samples with ACR compared to acute tubular injury without rejection or pretransplant "normal kidney" biopsy samples. Of CD133+ cells, ~85% are within injured tubules and ~15% are interstitial. Both populations express stem cell marker TRA-1-60 and TNFR2, but only tubular CD133+ cells express proximal tubular markers megalin and aquaporin-1. TNFR2+ CD133+ cells in tubules express proliferation marker phospho-histone H3S10 (pH3S10 ). Tubular epithelial cells in normal kidney organ cultures respond to TNFR2 signaling by expressing CD133 mRNA and protein, stem cell marker TRA-1-60, and pH3S10 within 3 hours of treatment. This rapid response time suggests that CD133+ cells in regenerating tubules of kidneys undergoing ACR represent proliferating tubular epithelial cells with TNFR2-induced stem cell markers rather than expansion of resident stem cells. Infiltrating host mononuclear cells are a likely source of TNF as these changes are absent in acute tubular injury .

Clinical-pathological correlations in post-transplant thrombotic microangiopathy.

AIMS: Post-transplant thrombotic microangiopathy (TMA) is a rare and clinically challenging finding in renal transplant biopsies. In addition to recurrent atypical haemolytic uraemic syndrome, TMA in renal transplants is associated with various conditions, such as calcineurin inhibitor (CNI) treatment, antibody-mediated rejection (ABMR), viral infections, sepsis, pregnancy, malignancies, and surgery. The therapeutic implications of this diagnosis are considerable. In order to better understand post-transplant TMA and to identify histological or clinical differences between associated causes, we conducted a multicentre retrospective study. METHODS AND RESULTS: Clinical parameters and transplant renal biopsy findings from 81 patients with TMA were analysed. Biopsies from 38 patients were also analysed with electron microscopy. On the basis of clinical-pathological correlation, TMA was attributed to a main aetiology, whenever possible. TMA occurred at a median of 30 days post-transplantation. Systemic features of TMA were present in only 18% of cases. Twenty-two per cent of cases were attributed to CNI and 11% to ABMR. Although other potentially contributing factors were found in 56% of patients, in most cases (63%) no clearly attributable cause of TMA was identified. Histological differences between groups were minimal. The detection of ultrastructural features that are usually associated with ABMR may help to establish ABMR as the cause of TMA. CONCLUSIONS: Although CNI and ABMR appear to be the main contributors to post-transplant TMA, the aetiology of most cases is probably multifactorial, and TMA cannot be unequivocally attributed to a single underlying aetiology. Morphological features of TMA are not discriminating, but electron microscopy may help to identify ABMR-associated TMA.

Local renal autoantibody production in lupus nephritis.

Autoantibodies are central to the pathogenesis of several autoimmune diseases including systemic lupus erythematosus. Plasma cells secrete these autoantibodies, but the anatomical sites of these cells are not well defined. Here, we found that although dsDNA-specific plasma cells in NZB/W mice were present in spleen and bone marrow, a large number were in the kidneys and their number correlated with the serum dsDNA-IgG titer. We observed renal plasma cells only in mice with nephritis, where they located mainly to the tubulointerstitium of the cortex and outer medulla. These cells had the phenotypic characteristics of fully differentiated plasma cells and, similar to long-lived bone marrow plasma cells, they were not in cell cycle. In patients with lupus nephritis, plasma cells were often present in the medulla in those with the most severe disease, especially combined proliferative and membranous lupus nephritis. The identification of the kidney as a major site of autoreactive plasma cells has implications for our understanding of the pathogenesis of lupus nephritis and for strategies to deplete autoreactive plasma cells, a long-standing therapeutic aim.

Extradigital glomangiomyoma of the forearm mimicking peripheral nerve sheath tumour and thrombosed varicose vein.

Extradigital glomus tumour is uncommon, little-known outside of its subungual location, and may present without its classic triad of tenderness, cold sensitivity and paroxysmal pain. Imaging is non-specific and diagnosis is often delayed, sometimes for years, leading to unnecessary morbidity. Surgical excision is the treatment of choice, although technique depends on case specifics. Histological subtypes depend on the relative prominence of glomus cells, vascular structures and smooth muscle. The vast majority of glomus tumours are benign. We highlight the importance of considering extradigital glomus tumours when generating differential diagnoses of an atypical painful lesion in a variety of clinical specialties.

PreImplantation Trial of Histopathology In renal Allografts (PITHIA): a stepped-wedge cluster randomised controlled trial protocol.

INTRODUCTION: Most potential kidney transplant donors in the UK are aged over 60 years, yet increasing donor age is associated with poorer graft survival and function. Urgent preimplantation kidney biopsy can identify chronic injury, and may aid selection of better 'quality' kidneys from this group. However, the impact of biopsy on transplant numbers remains unproven. The PreImplantation Trial of Histopathology In renal Allografts (PITHIA) study will assess whether the introduction of a national, 24 hours, digital histopathology service increases the number, and improves outcomes, of kidneys transplanted in the UK from older deceased donors. METHODS AND ANALYSIS: PITHIA is an open, multicentre, stepped-wedge cluster randomised study, involving all UK adult kidney transplant centres. At 4-monthly intervals, a group of 4-5 randomly selected clusters (transplant centres) will be given access to remote, urgent, digital histopathology (total intervention period, 24 months). The trial has two primary end points: it is powered for an 11% increase in the proportion of primary kidney offers from deceased donors aged over 60 years that are transplanted, and a 6 mL/min increase in the estimated glomerular filtration rate of recipients at 12 months post-transplant. This would equate to an additional 120 kidney transplants performed in the UK annually. Trial outcome data will be collected centrally via the UK Transplant Registry held by NHS Blood and Transplant (NHSBT) and will be analysed using mixed effects models allowing for clustering within centres and adjusting for secular trends. An accompanying economic evaluation will estimate the cost-effectiveness of the service to the National Health Service. ETHICS AND DISSEMINATION: The study has been given favourable ethical opinion by the Cambridge South Research Ethics Committee and is approved by the Health Research Authority. We will present our findings at key transplant meetings, publish results within 4 years of the trial commencing and support volunteers at renal patient groups to disseminate the trial outcome. TRIAL REGISTRATION NUMBER: ISRCTN11708741; Pre-results.

Vitamin K antagonists predispose to calciphylaxis in patients with end-stage renal disease.

BACKGROUND/AIMS: Calciphylaxis is associated with a poor prognosis in dialysis patients, and its pathogenesis remains incompletely understood. Although the use of vitamin K antagonists (VKA) has been implicated, previous reports are conflicting. We aimed to determine if vitamin K antagonists conferred an increased risk of calciphylaxis in patients on dialysis. METHODS: We performed a single-centre, retrospective cohort study of 2,234 patients receiving dialysis, and compared the characteristics of those with and without calciphylaxis. RESULTS: We identified 5 cases of calciphylaxis (all female) between January 2009 and December 2013. Overall, 142 patients (6.4%) were treated with VKA during the study period. Calciphylaxis was more common in the VKA group (4 of 142 patients, OR = 61, 95% CI 6.7-546, p = 0.0001). VKA was withdrawn in all cases and treatment instituted with sodium thiosulphate, cinacalcet and supportive measures. All patients recovered, although there was one sudden cerebrovascular death during follow-up. CONCLUSION: Treatment with VKA predisposes to the development of calciphylaxis.