A deep-learning-based model for assessment of autoimmune hepatitis from histology: AI(H).

Histological assessment of autoimmune hepatitis (AIH) is challenging. As one of the possible results of these challenges, nonclassical features such as bile-duct injury stays understudied in AIH. We aim to develop a deep learning tool (artificial intelligence for autoimmune hepatitis [AI(H)]) that analyzes the liver biopsies and provides reproducible, quantifiable, and interpretable results directly from routine pathology slides. A total of 123 pre-treatment liver biopsies, whole-slide images with confirmed AIH diagnosis from the archives of the Institute of Pathology at University Hospital Basel, were used to train several convolutional neural network models in the Aiforia artificial intelligence (AI) platform. The performance of AI models was evaluated on independent test set slides against pathologist's manual annotations. The AI models were 99.4%, 88.0%, 83.9%, 81.7%, and 79.2% accurate (ratios of correct predictions) for tissue detection, liver microanatomy, necroinflammation features, bile duct damage detection, and portal inflammation detection, respectively, on hematoxylin and eosin-stained slides. Additionally, the immune cells model could detect and classify different immune cells (lymphocyte, plasma cell, macrophage, eosinophil, and neutrophil) with 72.4% accuracy. On Sirius red-stained slides, the test accuracies were 99.4%, 94.0%, and 87.6% for tissue detection, liver microanatomy, and fibrosis detection, respectively. Additionally, AI(H) showed bile duct injury in 81 AIH cases (68.6%). The AI models were found to be accurate and efficient in predicting various morphological components of AIH biopsies. The computational analysis of biopsy slides provides detailed spatial and density data of immune cells in AIH landscape, which is difficult by manual counting. AI(H) can aid in improving the reproducibility of AIH biopsy assessment and bring new descriptive and quantitative aspects to AIH histology.

Deep Learning-Based Image Analysis of Liver Steatosis in Mouse Models.

The incidence of nonalcoholic fatty liver disease is a continuously growing health problem worldwide, along with obesity. Therefore, novel methods to both efficiently study the manifestation of nonalcoholic fatty liver disease and to analyze drug efficacy in preclinical models are needed. The present study developed a deep neural network-based model to quantify microvesicular and macrovesicular steatosis in the liver on hematoxylin-eosin-stained whole slide images, using the cloud-based platform, Aiforia Create. The training data included a total of 101 whole slide images from dietary interventions of wild-type mice and from two genetically modified mouse models with steatosis. The algorithm was trained for the following: to detect liver parenchyma, to exclude the blood vessels and any artefacts generated during tissue processing and image acquisition, to recognize and differentiate the areas of microvesicular and macrovesicular steatosis, and to quantify the recognized tissue area. The results of the image analysis replicated well the evaluation by expert pathologists and correlated well with the liver fat content measured by EchoMRI ex vivo, and the correlation with total liver triglycerides was notable. In conclusion, the developed deep learning-based model is a novel tool for studying liver steatosis in mouse models on paraffin sections and, thus, can facilitate reliable quantification of the amount of steatosis in large preclinical study cohorts.

Automated image analysis of keratin 7 staining can predict disease outcome in primary sclerosing cholangitis.

BACKGROUND AND AIMS: Primary sclerosing cholangitis (PSC) is a chronic cholestatic liver disease that obstructs the bile ducts and causes liver cirrhosis and cholangiocarcinoma. Efficient surrogate markers are required to measure disease progression. The cytokeratin 7 (K7) load in a liver specimen is an independent prognostic indicator that can be measured from digitalized slides using artificial intelligence (AI)-based models. METHODS: A K7-AI model 2.0 was built to measure the hepatocellular K7 load area of the parenchyma, portal tracts, and biliary epithelium. K7-stained PSC liver biopsy specimens (n = 295) were analyzed. A compound endpoint (liver transplantation, liver-related death, and cholangiocarcinoma) was applied in Kaplan-Meier survival analysis to measure AUC values and positive likelihood ratios for each histological variable detected by the model. RESULTS: The K7-AI model 2.0 was a better prognostic tool than plasma alkaline phosphatase, the fibrosis stage evaluated by Nakanuma classification, or K7 score evaluated by a pathologist based on the AUC values of measured variables. A combination of parameters, such as portal tract volume and area of K7-positive hepatocytes analyzed by the model, produced an AUC of 0.81 for predicting the compound endpoint. Portal tract volume measured by the model correlated with the histological fibrosis stage. CONCLUSIONS: The K7 staining of histological liver specimens in PSC provides significant information on disease outcomes through objective and reproducible data, including variables that cannot be measured by a human pathologist. The K7-AI model 2.0 could serve as a prognostic tool for clinical endpoints and as a surrogate marker in drug trials.

Chronic cholestasis detection by a novel tool: automated analysis of cytokeratin 7-stained liver specimens.

BACKGROUND: The objective was to build a novel method for automated image analysis to locate and quantify the number of cytokeratin 7 (K7)-positive hepatocytes reflecting cholestasis by applying deep learning neural networks (AI model) in a cohort of 210 liver specimens. We aimed to study the correlation between the AI model's results and disease progression. The cohort of liver biopsies which served as a model of chronic cholestatic liver disease comprised of patients diagnosed with primary sclerosing cholangitis (PSC). METHODS: In a cohort of patients with PSC identified from the PSC registry of the University Hospital of Helsinki, their K7-stained liver biopsy specimens were scored by a pathologist (human K7 score) and then digitally analyzed for K7-positive hepatocytes (K7%area). The digital analysis was by a K7-AI model created in an Aiforia Technologies cloud platform. For validation, values were human K7 score, stage of disease (Metavir and Nakunuma fibrosis score), and plasma liver enzymes indicating clinical cholestasis, all subjected to correlation analysis. RESULTS: The K7-AI model results (K7%area) correlated with the human K7 score (0.896; p < 2.2e- 16). In addition, K7%area correlated with stage of PSC (Metavir 0.446; p < 1.849e- 10 and Nakanuma 0.424; p < 4.23e- 10) and with plasma alkaline phosphatase (P-ALP) levels (0.369, p < 5.749e- 5). CONCLUSIONS: The accuracy of the AI-based analysis was comparable to that of the human K7 score. Automated quantitative image analysis correlated with stage of PSC and with P-ALP. Based on the results of the K7-AI model, we recommend K7 staining in the assessment of cholestasis by means of automated methods that provide fast (9.75 s/specimen) quantitative analysis.

Co-circulation of highly diverse Aleutian mink disease virus strains in Finland.

Aleutian mink disease virus (AMDV) is the causative agent of Aleutian disease (AD), which affects mink of all genotypes and also infects other mustelids such as ferrets, martens and badgers. Previous studies have investigated diversity in Finnish AMDV strains, but these studies have been restricted to small parts of the virus genome, and mostly from newly infected farms and free-ranging mustelids. Here, we investigated the diversity and evolution of Finnish AMDV strains by sequencing the complete coding sequences of 31 strains from mink originating from farms differing in their virus history, as well as from free-ranging mink. The data set was supplemented with partial genomes obtained from 26 strains. The sequences demonstrate that the Finnish AMDV strains have considerable diversity, and that the virus has been introduced to Finland in multiple events. Frequent recombination events were observed, as well as variation in the evolutionary rate in different parts of the genome and between different branches of the phylogenetic tree. Mink in the wild carry viruses with high intra-host diversity and are occasionally even co-infected by two different strains, suggesting that free-ranging mink tolerate chronic infections for extended periods of time. These findings highlight the need for further sampling to understand the mechanisms playing a role in the evolution and pathogenesis of AMDV.

Infection with Possible Novel Parapoxvirus in Horse, Finland, 2013.

A horse in Finland exhibited generalized granulomatous inflammation and severe proliferative dermatitis. After euthanization, we detected poxvirus DNA from a skin lesion sample. The virus sequence grouped with parapoxviruses, closely resembling a novel poxvirus detected in humans in the United States after horse contact. Our findings indicate horses may be a reservoir for zoonotic parapoxvirus.

Cerebrospinal fluid biomarkers for Alzheimer's disease in patients with frontotemporal lobar degeneration and amyotrophic lateral sclerosis with the C9ORF72 repeat expansion.

BACKGROUND: The C9ORF72 expansion is one of the most common causes of frontotemporal lobar degeneration (FTLD) and amyotrophic lateral sclerosis (ALS). The C9ORF72 expansion is associated with TDP-43 and p62 neuropathology, and amyloid plaques and neurofibrillary tangles are not common in patients with the C9ORF72 expansion. Therefore, we hypothesized that cerebrospinal fluid (CSF) biomarkers for Alzheimer's disease [AD; Aß1-42, total tau (T-tau) and phospho-tau] are normal in these patients. METHODS: The CSF Aß1-42, T-tau and phospho-tau levels were measured in 40 Finnish patients with the C9ORF72 expansion (29 FTLD, 10 ALS and 1 FTLD-ALS) using ELISA. RESULTS: A decreased Aß1-42 level was found in 25% of cases, while there were only single cases with changes in the t-Tau or phospho-tau level. The patients with abnormal biomarkers fulfilled the clinical criteria of the behavioral variant frontotemporal dementia and expressed no clinical signs of AD. CONCLUSIONS: In clinical diagnostics, a decreased CSF Aß1-42 level does not exclude the C9ORF72 expansion associated with FTLD.

Molecular epidemiology of Aleutian mink disease virus (AMDV) in Estonia, and a global phylogeny of AMDV.

Aleutian mink disease virus (AMDV) causes a severe disease called Aleutian disease (AD). AMDV infects primarily mustelids, but also other mammal species. Recent evidence suggests that AMDV may also affect humans. To examine AMDV in different wild animals and in farmed mink in Estonia, we collected 203 blood samples from eight mammal species in 2007-2010, of which 152 were from species living in the wild (American mink, European mink, pine marten, polecat, raccoon dog, badger, otter, and stone marten) and 51 were from farmed mink. AMDV was tested by PCR amplification of NS1 and VP2 gene fragments, and was only detected in 4 free-ranging (14.8%) and 11 farmed (21.6%) American mink. No other species was positive for AMDV. In addition, the VP2 gene fragment was sequenced for 14 farmed mink isolates from Finland for which NS1 sequences were already publicly available. None of the four Estonian AMDV isolates found in free-ranging mink had identical sequences with farmed mink. In fact, isolates from free-ranging and farmed mink belonged to different clades, suggesting that the analyzed virus isolates circulating in nature are not from escapees of current farms. Two global phylogenies were built: one based on NS1 (336 bp, 151 taxa from nine countries); the other based on a combined NS1-VP2 dataset (871 bp, 40 taxa from six countries). AMDV genotypes did not cluster according to their geographic origin, suggesting that transport of farm mink from multiple source farms has been intense. Nevertheless, one subclade in both phylogenies was comprised solely of isolates from farmed mink, while several subclades comprised isolates only from free-ranging mink, indicating that some isolates may circulate more in the wild and others among farm animals.

Validation of an automated ELISA system for detection of antibodies to Aleutian mink disease virus using blood samples collected in filter paper strips.

BACKGROUND: Aleutian mink disease virus (AMDV) is the cause of a chronic immune complex disease, Aleutian disease (AD), which is common in mink-producing countries. In 2005, implementation of an AMDV eradication programme in Finland created a need for an automated high-throughput assay. The aim of this study was to validate an AMDV-VP2 -recombinant antigen ELISA, which we developed earlier, in an automated assay format for the detection of anti-AMDV antibodies in mink blood and to determine the accuracy of this test compared with the reference standard (counter-current immunoelectrophoresis, CIEP). METHODS: A blood sampling method based on filter paper 12-strips (blood combs) and a device to introduce these strips to an ELISA plate for elution of the samples were developed. Blood and serum samples were collected from 761 mink from two farms with low (2%) and high (81%) seroprevalences of AMDV infection in 2008. ELISA sensitivity and specificity were estimated with a Bayesian 2-test 2-population model that allowed for conditional dependence between CIEP and ELISA. Agreement between the two tests was assessed with kappa statistic and proportion agreement. RESULTS: The sensitivity and specificity of the automated ELISA system were estimated to be 96.2% and 98.4%, respectively. Agreement between CIEP and ELISA was high, with a kappa value of 0.976 and overall proportion agreement of 98.8%. CONCLUSIONS: The automated ELISA system combined with blood comb sampling is an accurate test format for the detection of anti-AMDV antibodies in mink blood and offers several advantages, including improved blood sampling and data handling, fast sample throughput time, and reductions in costs and labour inputs.

Safety and biodistribution of a double-deleted oncolytic vaccinia virus encoding CD40 ligand in laboratory Beagles.

We evaluated adverse events, biodistribution and shedding of oncolytic vaccinia virus encoding CD40 ligand in two Beagles, in preparation for a phase 1 trial in canine cancer patients. Dog 1 received one dose of vaccinia virus and was euthanized 24 hours afterwards, while dog 2 received virus four times once weekly and was euthanized 7 days after that. Dogs were monitored for adverse events and underwent a detailed postmortem examination. Blood, saliva, urine, feces, and organs were collected for virus detection. Dog 1 had mild fever and lethargy while dog 2 experienced a possible seizure 5.5 hours after first virus administration. Viral DNA declined quickly in the blood after virus administration in both dogs but was still detectable 1 week later by quantitative polymerase chain reaction. Only samples taken directly after virus infusion contained infectious virus. Small amounts of viral DNA, but no infectious virus, were detected in a few saliva and urine samples. Necropsies did not reveal any relevant pathological changes and virus DNA was detected mainly in the spleen. The dogs in the study did not have cancer, and thus adverse events could be more common and viral load higher in dogs with tumors which allow viral amplification.

Development and evaluation of an enzyme-linked immunosorbent assay based on recombinant VP2 capsids for the detection of antibodies to Aleutian mink disease virus.

Aleutian disease (AD), a common infectious disease in farmed minks worldwide, is caused by Aleutian mink disease virus (AMDV). Serodiagnosis of AD in minks has been based on detection of AMDV antibodies by counterimmunoelectrophoresis (CIE) since the 1980s. The aim of this study was to develop and evaluate an enzyme-linked immunosorbent assay (ELISA) based on recombinant virus-like particles (VLPs) for identifying AMDV antibodies from mink sera. AMDV capsid protein (VP2) of a Finnish wild-type strain was expressed by the baculovirus system in Spodoptera frugiperda 9 insect cells and was shown to self-assemble to VLPs (with an ultrastructure similar to that of the actual virion). A direct immunoglobulin G ELISA was established using purified recombinant AMDV VP2 VLPs as an antigen. Sera from farmed minks were collected to evaluate the AMDV VP2 ELISA (n = 316) and CIE (n = 209) based on AMDV VP2 recombinant antigen in parallel with CIE performed using a commercially available traditional antigen. CIE performed with the recombinant antigen had a sensitivity and specificity of 100% and ELISA a sensitivity of 99% and a specificity of 97%, with reference to CIE performed with the commercial antigen. The results show that the recombinant AMDV VP2 VLPs are antigenic and that AMDV VP2 ELISA is sensitive and specific and encourage further development of the method for high-throughput diagnostics, involving hundreds of thousands of samples in Finland annually.

Molecular epidemiology of Aleutian mink disease virus in Finland.

Aleutian mink disease virus (AMDV) is a parvovirus that causes an immune complex-mediated disease in minks. To gain a more detailed view of the molecular epidemiology of mink AMDV in Finland, we phylogenetically analysed 14 new Finnish strains from 5 farms and all 40 strains with corresponding sequences available in GenBank. A part of the major non-structural (NS1) protein gene was amplified and analysed phylogenetically. A rooted nucleotide tree was constructed using the maximum parsimony method. The strains described in this study showed 86-100% nucleotide identity and were nearly identical on each farm. The ratio of synonymous to non-synonymous substitutions was approximately 2.7, indicating a mild purifying selection. Phylogenetic analysis confirmed that AMDV strains form three groups (I-III), all of which contained Finnish strains. The tree inferred that the three lineages of AMDV have been introduced to Finland independently. The analysis suggested that AMDV strains do not cluster into genotypes based on geographical origin, year of isolation or pathogenicity. Based on these data, the molecular clock is not applicable to AMDV, and within this gene area no recombination was detected.