Explore the vast range of titles available.

Image-based machine learning and deep learning in particular has recently shown expert-level accuracy in medical image classification. In this study, we combine convolutional and recurrent architectures to train a deep network to predict colorectal cancer outcome based on images of tumour tissue samples. The novelty of our approach is that we directly predict patient outcome, without any intermediate tissue classification. We evaluate a set of digitized haematoxylin-eosin-stained tumour tissue microarray (TMA) samples from 420 colorectal cancer patients with clinicopathological and outcome data available. The results show that deep learning-based outcome prediction with only small tissue areas as input outperforms (hazard ratio 2.3; CI 95% 1.79–3.03; AUC 0.69) visual histological assessment performed by human experts on both TMA spot (HR 1.67; CI 95% 1.28–2.19; AUC 0.58) and whole-slide level (HR 1.65; CI 95% 1.30–2.15; AUC 0.57) in the stratification into low- and high-risk patients. Our results suggest that state-of-the-art deep learning techniques can extract more prognostic information from the tissue morphology of colorectal cancer than an experienced human observer.

Research using whole slide images (WSIs) of histopathology slides has increased exponentially over recent years. Glass slides from retrospective cohorts, some with patient follow-up data are digitised for the development and validation of artificial intelligence (AI) tools. Such resources, therefore, become very important, with the need to ensure that their quality is of the standard necessary for downstream AI development. However, manual quality control of large cohorts of WSIs by visual assessment is unfeasible, and whilst quality control AI algorithms exist, these focus on bespoke aspects of image quality, e.g. focus, or use traditional machine-learning methods, which are unable to classify the range of potential image artefacts that should be considered. In this study, we have trained and validated a multi-task deep neural network to automate the process of quality control of a large retrospective cohort of prostate cases from which glass slides have been scanned several years after production, to determine both the usability of the images at the diagnostic level (considered in this study to be the minimal standard for research) and the common image artefacts present. Using a two-layer approach, quality overlays of WSIs were generated from a quality assessment (QA) undertaken at patch-level at 5 × magnification. From these quality overlays the slide-level quality scores were predicted and then compared to those generated by three specialist urological pathologists, with a Pearson correlation of 0.89 for overall ‘usability’ (at a diagnostic level), and 0.87 and 0.82 for focus and H&E staining quality scores respectively. To demonstrate its wider potential utility, we subsequently applied our QA pipeline to the TCGA prostate cancer cohort and to a colorectal cancer cohort, for comparison. Our model, designated as PathProfiler, indicates comparable predicted usability of images from the cohorts assessed (86–90% of WSIs predicted to be usable), and perhaps more significantly is able to predict WSIs that could benefit from an intervention such as re-scanning or re-staining for quality improvement. We have shown in this study that AI can be used to automate the process of quality control of large retrospective WSI cohorts to maximise their utility for research.

Background Pathological grading of non-invasive urothelial carcinoma has a direct impact upon management. This study evaluates the reproducibility of grading these tumours on glass slides and digital pathology. Methods Forty eight non-invasive urothelial bladder carcinomas were graded by three uropathologists on glass and on a digital platform using the 1973 WHO and 2004 ISUP/WHO systems. Results Consensus grades for glass and digital grading gave Cohen’s kappa scores of 0.78 (2004) and 0.82 (1973). Of 142 decisions made on the key therapeutic borderline of low grade versus high grade urothelial carcinoma (2004) by the three pathologists, 85% were in agreement. For the 1973 grading system, agreement overall was 90%. Conclusions Agreement on grading on glass slide and digital screen assessment is similar or in some cases improved, suggesting at least non-inferiority of DP for grading of non-invasive urothelial carcinoma.

There are recognised potential pitfalls in digital diagnosis in urological pathology, including the grading of dysplasia. The World Health Organisation/International Society of Urological Pathology (WHO/ISUP) grading system for renal cell carcinoma (RCC) is prognostically important in clear cell RCC (CCRCC) and papillary RCC (PRCC), and is included in risk stratification scores for CCRCC, thus impacting on patient management. To date there are no systematic studies examining the concordance of WHO/ISUP grading between digital pathology (DP) and glass slide (GS) images. We present a validation study examining intraobserver agreement in WHO/ISUP grade of CCRCC and PRCC. Fifty CCRCCs and 10 PRCCs were graded (WHO/ISUP system) by three specialist uropathologists on three separate occasions (DP once then two GS assessments; GS1 and GS2) separated by wash-out periods of at least two-weeks. The grade was recorded for each assessment, and compared using Cohen's and Fleiss's kappa. There was 65 to 78% concordance of WHO/ISUP grading on DP and GS1. Furthermore, for the individual pathologists, the comparative kappa scores for DP versus GS1, and GS1 versus GS2, were 0.70 and 0.70, 0.57 and 0.73, and 0.71 and 0.74, and with no apparent tendency to upgrade or downgrade on DP versus GS. The interobserver kappa agreement was less, at 0.58 on DP and 0.45 on GS. Our results demonstrate that the assessment of WHO/ISUP grade on DP is noninferior to that on GS. There is an apparent slight improvement in agreement between pathologists on RCC grade when assessed on DP, which may warrant further study.

The use of immunohistochemistry in the reporting of prostate biopsies is an important adjunct when the diagnosis is not definite on haematoxylin and eosin (H&E) morphology alone. The process is however inherently inefficient with delays while waiting for pathologist review to make the request and duplicated effort reviewing a case more than once. In this study, we aimed to capture the workflow implications of immunohistochemistry requests and demonstrate a novel artificial intelligence tool to identify cases in which immunohistochemistry (IHC) is required and generate an automated request. We conducted audits of the workflow for prostate biopsies in order to understand the potential implications of automated immunohistochemistry requesting and collected prospective cases to train a deep neural network algorithm to detect tissue regions that presented ambiguous morphology on whole slide images. These ambiguous foci were selected on the basis of the pathologist requesting immunohistochemistry to aid diagnosis. A gradient boosted trees classifier was then used to make a slide-level prediction based on the outputs of the neural network prediction. The algorithm was trained on annotations of 219 immunohistochemistry-requested and 80 control images, and tested by threefold cross-validation. Validation was conducted on a separate validation dataset of 222 images. Non IHC-requested cases were diagnosed in 17.9 min on average, while IHC-requested cases took 33.4 min over multiple reporting sessions. We estimated 11 min could be saved on average per case by automated IHC requesting, by removing duplication of effort. The tool attained 99% accuracy and 0.99 Area Under the Curve (AUC) on the test data. In the validation, the average agreement with pathologists was 0.81, with a mean AUC of 0.80. We demonstrate the proof-of-principle that an AI tool making automated immunohistochemistry requests could create a significantly leaner workflow and result in pathologist time savings.

Due to COVID-19 outbreaks, most school pupils have had to be home-schooled for long periods of time. Two editions of a web-based competition “Beat the Pathologists” for school age participants in the UK ran to fill up pupils’ spare time after home-schooling and evaluate their ability on contributing to AI annotation. The two editions asked the participants to annotate different types of cells on Ki67 stained breast cancer images. The Main competition was at four levels with different level of complexity. We obtained annotations of four kinds of cells entered by school pupils and ground truth from expert pathologists. In this paper, we analyse school pupils’ performance on differentiating different kinds of cells and compare their performance with two neural networks (AlexNet and VGG16). It was observed that children tend to get very good performance in tumour cell annotation with the best F1 measure 0.81 which is a metrics taking both false positives and false negatives into account. Low accuracy was achieved with F1 score 0.75 on positive non-tumour cells and 0.59 on negative non-tumour cells. Superior performance on non-tumour cell detection was achieved by neural networks. VGG16 with training from scratch achieved an F1 score over 0.70 in all cell categories and 0.92 in tumour cell detection. We conclude that non-experts like school pupils have the potential to contribute to large-scale labelling for AI algorithm development if sufficient training activities are organised. We hope that competitions like this can promote public interest in pathology and encourage participation by more non-experts for annotation.

Testicular germ cell tumours (TGCT), which comprise seminoma and non-seminoma subtypes, are the most common cancers in young men. In this study, we present a comprehensive whole genome sequencing analysis of adult TGCTs. Leveraging samples from participants recruited via the UK National Health Service and data from the Genomics England 100,000 Genomes Project, our results provide an extended description of genomic elements underlying TGCT pathogenesis. This catalogue offers a comprehensive, high-resolution map of copy number alterations, structural variation, and key global genome features, including mutational signatures and analysis of extrachromosomal DNA amplification. This study establishes correlations between genomic alterations and histological diversification, revealing divergent evolutionary trajectories among TGCT subtypes. By reconstructing the chronological order of driver events, we identify a subgroup of adult TGCTs undergoing relatively late whole genome duplication. Additionally, we present evidence that human leukocyte antigen loss is a more prevalent mechanism of immune disruption in seminomas. Collectively, our findings provide valuable insights into the developmental and immune modulatory processes implicated in TGCT pathogenesis and progression. Testicular germ cell tumours (TGCT) are the most common cancers in young men. Here, the authors analyse the genomic landscape of TGCT using data from the Genomics England 100,000 Genomes Project, revealing divergent evolutionary trajectories and the prevalence of human leukocyte antigen loss.

Alcohol-related liver disease (ALD) remains a leading cause of liver-related morbidity and mortality. Age, fibrosis stage, MELD score and continued alcohol consumption predict outcome in everyday clinical practice. In previous studies increased hepatocyte nuclear area and hepatocyte expression of p21, both markers of senescence, were associated with increased fibrosis stage and a poor outcome in non-alcohol-related fatty liver disease, while increased hepatocyte nuclear area was related to liver dysfunction in ALD cirrhosis. This study, therefore, investigated the pattern of hepatocyte cell cycle phase distribution and hepatocyte p21 expression in relation to outcome in ALD. Liver sections from two cohorts were studied. The first comprised 42 patients across the full spectrum of ALD. The second cohort comprised 77 patients with ALD cirrhosis. Immunohistochemistry assessed hepatocyte expression of cell cycle phase markers and p21. Regenerating liver (n=12) and \"normal\" liver sections (n=5) served as positive and negative controls, respectively. In the first cohort there was little cell cycle progression beyond G1/S phase and increased hepatocyte p21 expression (p<0.0001), which correlated independently with fibrosis stage (p=0.005) and an adverse liver-related outcome (p=0.03). In the second cohort, both hepatocyte p21 expression (p<0.001) and MELD score (p=0.006) were associated independently with an adverse liver-related outcome; this association was stronger with hepatocyte p21 expression (AUROC 0.74; p=0.0002) than with MELD score (AUROC 0.59; p=0.13). Further, hepatocyte p21 expression co-localised with increased hepatic stellate cell activation. The findings are consistent with impaired cell cycle progression beyond the G1/S phase in ALD. The striking independent associations between increased hepatocyte p21 expression and both fibrosis stage and an adverse liver-related outcome in both cohorts suggests hepatocyte senescence plays an important role in ALD. Measuring hepatocyte p21 expression is simple and cheap and in this series was a useful measure of long-term prognosis in ALD.

Insulin-like growth factor-1 (IGF-1) is associated with prostate cancer (PCa) development and lethality and exhibits immunosuppressive properties in other models. We investigated IGF-1’s tumor-intrinsic immune effects in PCa to understand mechanisms underlying its poor immunotherapy response. Transcriptional profiling of human (DU145, 22Rv1) and murine (Myc-CaP) PCa cells revealed that IGF-1 suppresses cytokine signalling, antigen processing and presentation, and additional immune regulatory pathways. We further examined the expression of components involved in cancer cell recognition and immune evasion: the antigen processing machinery and PD-L1 checkpoint. IGF-1 downregulated key elements such as transporters associated with antigen processing (TAPs), endoplasmic reticulum aminopeptidase-1 (ERAP-1), and Class I β2-microglobulin, without significantly altering Class I allele expression. These changes were associated with reduced surface presentation of Class I complexes on Myc-CaP cells, suggesting disrupted peptide transport, processing, and/or presentation. In contrast, IGF-1 upregulated the immune checkpoint CD274 (PD-L1) via IGF receptor/AKT/ERK-dependent signalling. Analysis of TCGA Firehose Legacy PCa data showed higher CD274 expression in tumors with elevated IGF1 and IGFBP5. Multiplex immunofluorescence in primary PCa confirmed increased PD-L1 in patients with high serum IGF-1, supporting its role in immune evasion. Overall, these findings reveal a novel IGF-1-driven immunosuppressive mechanism that may underlie PCa’s resistance to immunotherapy.

Genetic signatures have added a molecular dimension to prognostics and therapeutic decision-making. However, tumour heterogeneity in prostate cancer and current sampling methods could confound accurate assessment. Based on previously published spatial transcriptomic data from multifocal prostate cancer, we created virtual biopsy models that mimic conventional biopsy placement and core size. We then analysed the gene expression of different prognostic signatures (OncotypeDx®, Decipher®, Prostadiag®) using a step-wise approach with increasing resolution from pseudo-bulk analysis of the whole biopsy, to differentiation by tissue subtype (benign, stroma, tumour), followed by distinct tumour grade and finally clonal resolution. The gene expression profile of virtual tumour biopsies revealed clear differences between grade groups and tumour clones, compared to a benign control, which were not reflected in bulk analyses. This suggests that bulk analyses of whole biopsies or tumour-only areas, as used in clinical practice, may provide an inaccurate assessment of gene profiles. The type of tissue, the grade of the tumour and the clonal composition all influence the gene expression in a biopsy. Clinical decision making based on biopsy genomics should be made with caution while we await more precise targeting and cost-effective spatial analyses.