Development of a multi-scanner facility for data acquisition for digital pathology artificial intelligence.

Whole slide imaging (WSI) of pathology glass slides using high-resolution scanners has enabled the large-scale application of artificial intelligence (AI) in pathology, to support the detection and diagnosis of disease, potentially increasing efficiency and accuracy in tissue diagnosis. Despite the promise of AI, it has limitations. 'Brittleness' or sensitivity to variation in inputs necessitates that large amounts of data are used for training. AI is often trained on data from different scanners but not usually by replicating the same slide across scanners. The utilisation of multiple WSI instruments to produce digital replicas of the same slides will make more comprehensive datasets and may improve the robustness and generalisability of AI algorithms as well as reduce the overall data requirements of AI training. To this end, the National Pathology Imaging Cooperative (NPIC) has built the AI FORGE (Facilitating Opportunities for Robust Generalisable data Emulation), a unique multi-scanner facility embedded in a clinical site in the NHS to (1) compare scanner performance, (2) replicate digital pathology image datasets across WSI systems, and (3) support the evaluation of clinical AI algorithms. The NPIC AI FORGE currently comprises 15 scanners from nine manufacturers. It can generate approximately 4,000 WSI images per day (approximately 7 TB of image data). This paper describes the process followed to plan and build such a facility. © 2024 The Author(s). The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.

Attention-guided sampling for colorectal cancer analysis with digital pathology.

Improvements to patient care through the development of automated image analysis in pathology are restricted by the small image patch size that can be processed by convolutional neural networks (CNNs), when compared to the whole-slide image (WSI). Tile-by-tile processing across the entire WSI is slow and inefficient. While this may improve with future computing power, the technique remains vulnerable to noise from uninformative image areas. We propose a novel attention-inspired algorithm that selects image patches from informative parts of the WSI, first using a sparse randomised grid pattern, then iteratively re-sampling at higher density in regions where a CNN classifies patches as tumour. Subsequent uniform sampling across the enclosing region of interest (ROI) is used to mitigate sampling bias. Benchmarking tests informed the adoption of VGG19 as the main CNN architecture, with 79% classification accuracy. A further CNN was trained to separate false-positive normal epithelium from tumour epithelium, in a novel adaptation of a two-stage model used in brain imaging. These subsystems were combined in a processing pipeline to generate spatial distributions of classified patches from unseen WSIs. The ROI was predicted with a mean F1 (Dice) score of 86.6% over 100 evaluation WSIs. Several algorithms for evaluating tumour-stroma ratio (TSR) within the ROI were compared, giving a lowest root mean square (RMS) error of 11.3% relative to pathologists' annotations, against 13.5% for an equivalent tile-by-tile pipeline. Our pipeline processed WSIs between 3.3x and 6.3x faster than tile-by-tile processing. We propose our attention-based sampling pipeline as a useful tool for pathology researchers, with the further potential for incorporating additional diagnostic calculations.

Development of a Remote Online Collaborative Medical School Pathology Curriculum with Clinical Correlations, across Several International Sites, through the Covid-19 Pandemic.

INTRODUCTION: Due to the Covid-19 social distancing restrictions, in March 2020, Weill Cornell Medicine-Qatar decided to replace students' clinical instruction with novel online electives. Hence, we implemented an innovative online and remote pathology curriculum, anchored on virtual microscopy and Zoom videoconferencing: ideal tools to support online teaching. OBJECTIVE: To assess a new curriculum implementation at Weill Cornell Medicine-Qatar. MATERIALS AND METHODS: This for-credit, 2-week elective included 6 synchronous Zoom sessions where complex clinicopathological cases were discussed in small groups. We used open access digital microscopy slides from the University of Leeds' Virtual Pathology Library (http://www.virtualpathology.leeds.ac.uk/slides/library/). Students independently prepared for these sessions by reviewing cases, slides, readings, and questions in advance (asynchronous self-directed learning anchored on a flipped classroom model), and wrote a final review of a case. An assessment and feedback were given to each student. RESULTS: Four elective iterations were offered to a total of 29 students, with learners and faculty spread over 4 countries. During the Zoom sessions, students controlled the digital slides and offered their own diagnoses, followed by group discussions to strengthen autonomy and confidence. We surveyed learners about the elective's performance (program evaluation). Students conveyed high levels of satisfaction about the elective's overall quality, their pathology learning and online interactions, with minimal challenges related to the remote nature of the course. DISCUSSION AND CONCLUSIONS: Technological innovations mitigate sudden disruptions in medical education. A remote curriculum allows instruction at any distance, at any time, from anywhere, enhancing educational exchanges, flexibility and globalization in medical education.

The Effect of Quality Control on Accuracy of Digital Pathology Image Analysis.

Digital slide images produced from routine diagnostic histopathological preparations suffer from variation arising at every step of the processing pipeline. Typically, pathologists compensate for such variation using expert knowledge and experience, which is difficult to replicate in automated solutions. The extent to which inconsistencies affect image analysis is explored in this work, examining in detail, the results from a previously published algorithm automating the generation of tumor:stroma ratio (TSR) in colorectal clinical trial datasets. One dataset consisting of 2,211 cases and 106,268 expert-labelled images is used to identify quality issues, by visually inspecting cases where algorithm-pathologist agreement is lowest. Twelve categories are identified and used to analyze pathologist-algorithm agreement in relation to these categories. Of the 2,211 cases, 701 were found to be free from any image quality issues. Algorithm performance was then assessed, comparing pathologist agreement with image quality classification. It was found that agreement was lowest on poorly differentiated tissue, with a mean TSR difference of 0.25 (sd = 0.24). Removing images that contained quality issues increased accuracy from 80% to 83%, at the expense of reducing the dataset to 33,736 images (32%). Training the algorithm on the optimized dataset, prior to testing on all images saw a decrease in accuracy of 4%, indicating that the optimized dataset did not contain enough variation to generate a fully representative model. The results provide an in-depth perspective on image quality, highlighting the importance of the effects on downstream image analysis.

A Point-of-Use Quality Assurance Tool for Digital Pathology Remote Working.

Pathology services are facing pressures due to the COVID-19 pandemic. Digital pathology has the capability to meet some of these unprecedented challenges by allowing remote diagnoses to be made at home, during periods of social distancing or self-isolation. However, while digital pathology allows diagnoses to be made on standard computer screens, unregulated home environments may not be conducive for optimal viewing conditions. There is also a paucity of experimental evidence available to support the minimum display requirements for digital pathology. This study presents a Point-of-Use Quality Assurance (POUQA) tool for remote assessment of viewing conditions for reporting digital pathology slides. The tool is a psychophysical test combining previous work from successfully implemented quality assurance tools in both pathology and radiology to provide a minimally intrusive display screen validation task, before viewing digital slides. The test is specific to pathology assessment in that it requires visual discrimination between colors derived from hematoxylin and eosin staining, with a perceptual difference of ±1 delta E (dE). This tool evaluates the transfer of a 1 dE signal through the digital image display chain, including the observers' contrast and color responses within the test color range. The web-based system has been rapidly developed and deployed as a response to the COVID-19 pandemic and may be used by anyone in the world to help optimize flexible working conditions at: http://www. virtualpathology.leeds.ac.uk/res earch/systems/pouqa/.

Integrated eicosanoid lipidomics and gene expression reveal decreased prostaglandin catabolism and increased 5-lipoxygenase expression in aggressive subtypes of endometrial cancer.

Eicosanoids comprise a diverse group of bioactive lipids which orchestrate inflammation, immunity, and tissue homeostasis, and whose dysregulation has been implicated in carcinogenesis. Among the various eicosanoid metabolic pathways, studies of their role in endometrial cancer (EC) have very much been confined to the COX-2 pathway. This study aimed to determine changes in epithelial eicosanoid metabolic gene expression in endometrial carcinogenesis; to integrate these with eicosanoid profiles in matched clinical specimens; and, finally, to investigate the prognostic value of candidate eicosanoid metabolic enzymes. Eicosanoids and related mediators were profiled using liquid chromatography-tandem mass spectrometry in fresh frozen normal, hyperplastic, and cancerous (types I and II) endometrial specimens (n = 192). Sample-matched epithelia were isolated by laser capture microdissection and whole genome expression analysis was performed using microarrays. Integration of eicosanoid and gene expression data showed that the accepted paradigm of increased COX-2-mediated prostaglandin production does not apply in EC carcinogenesis. Instead, there was evidence for decreased PGE2 /PGF2α inactivation via 15-hydroxyprostaglandin dehydrogenase (HPGD) in type II ECs. Increased expression of 5-lipoxygenase (ALOX5) mRNA was also identified in type II ECs, together with proportional increases in its product, 5-hydroxyeicosatetraenoic acid (5-HETE). Decreased HPGD and elevated ALOX5 mRNA expression were associated with adverse outcome, which was confirmed by immunohistochemical tissue microarray analysis of an independent series of EC specimens (n = 419). While neither COX-1 nor COX-2 protein expression had prognostic value, low HPGD combined with high ALOX5 expression was associated with the worst overall and progression-free survival. These findings highlight HPGD and ALOX5 as potential therapeutic targets in aggressive EC subtypes. Copyright © 2018 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

Biopsy proportion of tumour predicts pathological tumour response and benefit from chemotherapy in resectable oesophageal carcinoma: results from the UK MRC OE02 trial.

BACKGROUND: Neoadjuvant chemotherapy followed by surgery is the standard of care for UK patients with locally advanced resectable oesophageal carcinoma (OeC). However, not all patients benefit from multimodal treatment and there is a clinical need for biomarkers which can identify chemotherapy responders. This study investigated whether the proportion of tumour cells per tumour area (PoT) measured in the pre-treatment biopsy predicts chemotherapy benefit for OeC patients. PATIENTS AND METHODS: PoT was quantified using digitized haematoxylin/eosin stained pre-treatment biopsy slides from 281 OeC patients from the UK MRC OE02 trial (141 treated by surgery alone (S); 140 treated by 5-fluorouracil/cisplatin followed by surgery (CS)). The relationship between PoT and clinicopathological data including tumour regression grade (TRG), overall survival and treatment interaction was investigated. RESULTS: PoT was associated with chemotherapy benefit in a non-linear fashion (test for interaction, P=0.006). Only patients with a biopsy PoT between 40% and 70% received a significant survival benefit from neoadjuvant chemotherapy (N=129; HR (95%CI):1.94 (1.39-2.71), unlike those with lower or higher PoT (PoT<40%, N=39, HR:1.25 (0.66-2.35); PoT>70% (N=28, HR:0.65 (0.36-1.18)). High pre-treatment PoT was related to lack of primary tumour regression (TRG 4 or 5), P=0.0402. CONCLUSIONS: This is the first study to identify in a representative subgroup of OeC patients from a large randomized phase III trial that the proportion of tumour in the pre-chemotherapy biopsy predicts benefit from chemotherapy and may be a clinically useful biomarker for patient treatment stratification.Proportion of tumour is a novel biomarker which can be measured in the pre-treatment diagnostic biopsy and which may enable the identification of chemotherapy responders and non-responders among patients with oesophageal carcinoma. Proportion of tumour could easily become part of the routine reporting of oesophageal cancer biopsies and may aid in managing patients with borderline resectable cancer.

The prognostic significance of tumour-stroma ratio in endometrial carcinoma.

BACKGROUND: High tumour stromal content has been found to predict adverse clinical outcome in a range of epithelial tumours. The aim of this study was to assess the prognostic significance of tumour-stroma ratio (TSR) in endometrial adenocarcinomas and investigate its relationship with other clinicopathological parameters. METHODS: Clinicopathological and 5-year follow-up data were obtained for a retrospective series of endometrial adenocarcinoma patients (n=400). TSR was measured using a morphometric approach (point counting) on digitised histologic hysterectomy specimens. Inter-observer agreement was determined using Cohen's Kappa statistic. TSR cut-offs were optimised using log-rank functions and prognostic significance of TSR on overall survival (OS) and disease-free survival (DFS) were determined using Cox Proportional Hazards regression analysis and Kaplan-Meier curves generated. Associations of TSR with other clinicopathological parameters were determined using non-parametric tests followed by Holm-Bonferroni correction for multiple comparisons. RESULTS: TSR as a continuous variable associated with worse OS (P=0.034) in univariable Cox-regression analysis. Using the optimal cut-off TSR value of 1.3, TSR-high (i.e. low stroma) was associated with worse OS (HR=2.51; 95% CI=1.22-5.12; P=0.021) and DFS (HR=2.19; 95% CI=1.15-4.17; P=0.017) in univariable analysis. However, TSR did not have independent prognostic significance in multivariable analysis, when adjusted for known prognostic variables. A highly significant association was found between TSR and tumour grade (P<0.001) and lymphovascular space invasion (P<0.001), both of which had independent prognostic significance in this study population. CONCLUSIONS: Low tumour stromal content associates with both poor outcome and with other adverse prognostic indicators in endometrial cancer, although it is not independently prognostic. These findings contrast with studies on many--although not all--cancers and suggest that the biology of tumour-stroma interactions may differ amongst cancer types.

Prospector: A web-based tool for rapid acquisition of gold standard data for pathology research and image analysis.

BACKGROUND: Obtaining ground truth for pathological images is essential for various experiments, especially for training and testing image analysis algorithms. However, obtaining pathologist input is often difficult, time consuming and expensive. This leads to algorithms being over-fitted to small datasets, and inappropriate validation, which causes poor performance on real world data. There is a great need to gather data from pathologists in a simple and efficient manner, in order to maximise the amount of data obtained. METHODS: We present a lightweight, web-based HTML5 system for administering and participating in data collection experiments. The system is designed for rapid input with minimal effort, and can be accessed from anywhere in the world with a reliable internet connection. RESULTS: We present two case studies that use the system to assess how limitations on fields of view affect pathologist agreement, and to what extent poorly stained slides affect judgement. In both cases, the system collects pathologist scores at a rate of less than two seconds per image. CONCLUSIONS: The system has multiple potential applications in pathology and other domains.

RandomSpot: A web-based tool for systematic random sampling of virtual slides.

This paper describes work presented at the Nordic Symposium on Digital Pathology 2014, Linköping, Sweden. Systematic random sampling (SRS) is a stereological tool, which provides a framework to quickly build an accurate estimation of the distribution of objects or classes within an image, whilst minimizing the number of observations required. RandomSpot is a web-based tool for SRS in stereology, which systematically places equidistant points within a given region of interest on a virtual slide. Each point can then be visually inspected by a pathologist in order to generate an unbiased sample of the distribution of classes within the tissue. Further measurements can then be derived from the distribution, such as the ratio of tumor to stroma. RandomSpot replicates the fundamental principle of traditional light microscope grid-shaped graticules, with the added benefits associated with virtual slides, such as facilitated collaboration and automated navigation between points. Once the sample points have been added to the region(s) of interest, users can download the annotations and view them locally using their virtual slide viewing software. Since its introduction, RandomSpot has been used extensively for international collaborative projects, clinical trials and independent research projects. So far, the system has been used to generate over 21,000 sample sets, and has been used to generate data for use in multiple publications, identifying significant new prognostic markers in colorectal, upper gastro-intestinal and breast cancer. Data generated using RandomSpot also has significant value for training image analysis algorithms using sample point coordinates and pathologist classifications.