Explore the vast range of titles available.

Context.—Gastrointestinal (GI) and pancreatobiliary tracts contain a variety of neuroendocrine cells that constitute a diffuse endocrine system. Neuroendocrine tumors (NETs) from these organs are heterogeneous tumors with diverse clinical behaviors. Recent improvements in the understanding of NETs from the GI and pancreatobiliary tracts have led to more-refined definitions of the clinicopathologic characteristics of these tumors. Under the 2010 World Health Organization classification scheme, NETs are classified as grade (G) 1 NETs, G2 NETs, neuroendocrine carcinomas, and mixed adenoneuroendocrine carcinomas. Histologic grades are dependent on mitotic counts and the Ki-67 labeling index. Several new issues arose after implementation of the 2010 World Health Organization classification scheme, such as issues with well-differentiated NETs with G3 Ki-67 labeling index and the evaluation of mitotic counts and Ki-67 labeling. Hereditary syndromes, including multiple endocrine neoplasia type 1 syndrome, von Hippel-Lindau syndrome, neurofibromatosis 1, and tuberous sclerosis, are related to NETs of the GI and pancreatobiliary tracts. Several prognostic markers of GI and pancreatobiliary tract NETs have been introduced, but many of them require further validation. Objective.—To understand clinicopathologic characteristics of NETs from the GI and pancreatobiliary tracts. Data Sources.—PubMed (US National Library of Medicine) reports were reviewed. Conclusions.—In this review, we briefly summarize recent developments and issues related to NETs of the GI and pancreatobiliary tracts.

Context.—Colorectal cancer is a heterogeneous disease entity with 3 molecular carcinogenesis pathways and 2 morphologic multistep pathways. Right-sided colon cancers and left-sided colon and rectal cancers exhibit differences in their incidence rates according to geographic region, age, and sex. A linear tendency toward increasing frequencies of microsatellite instability–high or CpG island methylator phenotype–high cancers in subsites along the bowel from the rectum to the cecum or the ascending colon accounts for the differences in tumor phenotypes associated with these subsites. The molecular subtypes of colorectal cancers exhibit different responses to adjuvant therapy, which might be responsible for differences in subtype-specific survival. Objectives.—To review the clinicopathologic and molecular features of the molecular subtypes of colorectal cancer generated by combined CpG island methylator phenotype and microsatellite statuses, to integrate these features with the most recent findings in the context of the prognostic implications of molecular subtypes, and to emphasize the necessity of developing molecular markers that enable the identification of adenocarcinomas involving the serrated neoplasia pathway. Data Sources.—Based on the authors' own experimental data and a review of the pertinent literature. Conclusions.—Because colorectal cancers arise from 2 different morphologic multistep carcinogenesis pathways with varying contributions from 3 different molecular carcinogenesis pathways, colorectal cancer is a heterogeneous and complex disease. Thus, molecular subtyping of colorectal cancers is an important approach to characterizing their heterogeneity with respect to not only prognosis and therapeutic response but also biology and natural history.

Histopathology image evaluation is indispensable for cancer diagnoses and subtype classification. Standard artificial intelligence methods for histopathology image analyses have focused on optimizing specialized models for each diagnostic task 1 , 2 . Although such methods have achieved some success, they often have limited generalizability to images generated by different digitization protocols or samples collected from different populations 3 . Here, to address this challenge, we devised the Clinical Histopathology Imaging Evaluation Foundation (CHIEF) model, a general-purpose weakly supervised machine learning framework to extract pathology imaging features for systematic cancer evaluation. CHIEF leverages two complementary pretraining methods to extract diverse pathology representations: unsupervised pretraining for tile-level feature identification and weakly supervised pretraining for whole-slide pattern recognition. We developed CHIEF using 60,530 whole-slide images spanning 19 anatomical sites. Through pretraining on 44 terabytes of high-resolution pathology imaging datasets, CHIEF extracted microscopic representations useful for cancer cell detection, tumour origin identification, molecular profile characterization and prognostic prediction. We successfully validated CHIEF using 19,491 whole-slide images from 32 independent slide sets collected from 24 hospitals and cohorts internationally. Overall, CHIEF outperformed the state-of-the-art deep learning methods by up to 36.1%, showing its ability to address domain shifts observed in samples from diverse populations and processed by different slide preparation methods. CHIEF provides a generalizable foundation for efficient digital pathology evaluation for patients with cancer. A study describes the development of a generalizable foundation machine learning framework to extract pathology imaging features for cancer diagnosis and prognosis prediction.

Machine learning applications in the pathology clinical domain are emerging rapidly. As decision support systems continue to mature, laboratories will increasingly need guidance to evaluate their performance in clinical practice. Currently there are no formal guidelines to assist pathology laboratories in verification and/or validation of such systems. These recommendations are being proposed for the evaluation of machine learning systems in the clinical practice of pathology. To propose recommendations for performance evaluation of in vitro diagnostic tests on patient samples that incorporate machine learning as part of the preanalytical, analytical, or postanalytical phases of the laboratory workflow. Topics described include considerations for machine learning model evaluation including risk assessment, predeployment requirements, data sourcing and curation, verification and validation, change control management, human-computer interaction, practitioner training, and competency evaluation. An expert panel performed a review of the literature, Clinical and Laboratory Standards Institute guidance, and laboratory and government regulatory frameworks. Review of the literature and existing documents enabled the development of proposed recommendations. This white paper pertains to performance evaluation of machine learning systems intended to be implemented for clinical patient testing. Further studies with real-world clinical data are encouraged to support these proposed recommendations. Performance evaluation of machine learning models is critical to verification and/or validation of in vitro diagnostic tests using machine learning intended for clinical practice.

Multiple independent studies have shown that tumor-infiltrating lymphocytes (TIL) are prognostic in breast cancer with potential relevance for response to immune-checkpoint inhibitor therapy. Although many groups are currently evaluating TIL, there is no standardized system for diagnostic applications. This study reports the results of two ring studies investigating TIL conducted by the International Working Group on Immuno-oncology Biomarkers. The study aim was to determine the intraclass correlation coefficient (ICC) for evaluation of TIL by different pathologists. A total of 120 slides were evaluated by a large group of pathologists with a web-based system in ring study 1 and a more advanced software-system in ring study 2 that included an integrated feedback with standardized reference images. The predefined aim for successful ring studies 1 and 2 was an ICC above 0.7 (lower limit of 95% confidence interval (CI)). In ring study 1 the prespecified endpoint was not reached (ICC: 0.70; 95% CI: 0.62–0.78). On the basis of an analysis of sources of variation, we developed a more advanced digital image evaluation system for ring study 2, which improved the ICC to 0.89 (95% CI: 0.85–0.92). The Fleiss' kappa value for <60 vs ≥60% TIL improved from 0.45 (ring study 1) to 0.63 in RS2 and the mean concordance improved from 88 to 92%. This large international standardization project shows that reproducible evaluation of TIL is feasible in breast cancer. This opens the way for standardized reporting of tumor immunological parameters in clinical studies and diagnostic practice. The software-guided image evaluation approach used in ring study 2 may be of value as a tool for evaluation of TIL in clinical trials and diagnostic practice. The experience gained from this approach might be applicable to the standardization of other diagnostic parameters in histopathology.

Since 1966, when Donald Gleason, MD, first proposed grading prostate cancer based on its histologic architecture, there have been numerous changes in clinical and pathologic practices relating to prostate cancer. Patterns 1 and 2, comprising more than 30% of cases in the original publications by Gleason, are no longer reported on biopsy and are rarely diagnosed on radical prostatectomy. Many of these cases may even have been mimickers of prostate cancer that were described later with the use of contemporary immunohistochemistry. The original Gleason system predated many newly described variants of prostate cancer and our current concept of intraductal carcinoma. Gleason also did not describe how to report prostate cancer on biopsy with multiple cores of cancer or on radical prostatectomy with separate tumor nodules. To address these issues, the International Society of Urological Pathology first made revisions to the grading system in 2005, and subsequently in 2014. Additionally, a new grading system composed of Grade Groups 1 to 5 that was first developed in 2013 at the Johns Hopkins Hospital and subsequently validated in a large multi-institutional and multimodal study was presented at the 2014 International Society of Urological Pathology meeting and accepted both by participating pathologists as well as urologists, oncologists, and radiation therapists. In the present study, we describe updates to the grading of prostate cancer along with the new grading system.

Digital pathology provides a possibility for computational analysis of histological slides and automatization of routine pathological tasks. Histological slides are very heterogeneous concerning staining, sections’ thickness, and artifacts arising during tissue processing, cutting, staining, and digitization. In this study, we digitally reproduce major types of artifacts. Using six datasets from four different institutions digitized by different scanner systems, we systematically explore artifacts’ influence on the accuracy of the pre-trained, validated, deep learning-based model for prostate cancer detection in histological slides. We provide evidence that any histological artifact dependent on severity can lead to a substantial loss in model performance. Strategies for the prevention of diagnostic model accuracy losses in the context of artifacts are warranted. Stress-testing of diagnostic models using synthetically generated artifacts might be an essential step during clinical validation of deep learning-based algorithms.

Digital pathology (DP) implementations vary in scale, based on aims of intended operation. Few laboratories have completed a full-scale DP implementation, which may be due to high overhead costs that disrupt the traditional pathology workflow. Neither standardized criteria nor benchmark data have yet been published showing practical return on investment after implementing a DP platform. To provide benchmark data and practical metrics to support operational efficiency and cost savings in a large academic center. Metrics reviewed include archived pathology asset retrieval; ancillary test request for recurrent/metastatic disease; cost analysis and turnaround time (TAT); and DP experience survey. Glass slide requests from the department slide archive and an off-site surgery center showed a 93% and 97% decrease, respectively. Ancillary immunohistochemical orders, compared in 2014 (52%)-before whole slide images (WSIs) were available in the laboratory information system-and 2017 (21%) showed $114 000/y in anticipated savings. Comprehensive comparative cost analysis showed a 5-year $1.3 million savings. Surgical resection cases with prior WSIs showed a 1-day decrease in TAT. A DP experience survey showed 80% of respondents agreed WSIs improved their clinical sign-out experience. Implementing a DP operation showed a noteworthy increase in efficiency and operational utility. Digital pathology deployments and operations may be gauged by the following metrics: number of glass slide requests as WSIs become available, decrease in confirmatory testing for patients with metastatic/recurrent disease, long-term decrease in off-site pathology asset costs, and faster TAT. Other departments may use our benchmark data and metrics to enhance patient care and demonstrate return on investment to justify adoption of DP.

Breast pathology reports include many important details to guide clinical management. Reports with missing critical data elements are commonly seen in non-subspecialized pathology practices. The use of synoptic templates has been shown to improve pathology reports. Although synoptic templates are readily available from professional societies, many are not tailored to low-resource settings. To perform an assessment of current breast pathology reporting at 3 referral hospitals in sub-Saharan Africa and design a locally adapted breast cancer synoptic template. We conducted semi-structured interviews with key stakeholders involved in breast cancer care, including pathologists, radiologists, oncologists, and surgeons, from Nigeria, Tanzania, and Mozambique. Moreover, each stakeholder reviewed a preliminary synoptic template that was compiled by using templates from the College of American Pathologists, Royal College of Pathologists, and International Collaboration on Cancer Reporting and was asked to score each data element as essential, optional, or exclude. A locally adapted synoptic template was then designed from the needs assessment. Using the adapted templates, a retrospective review of breast cancer pathology reports from 2020 to 2022 was conducted to determine the completeness of reports at the 3 institutions. A total of 17 physicians were interviewed. Review of pathology reports revealed that none of the reports across all 3 sites contained all data elements considered essential by local physicians. There is an urgent need to improve breast pathology reporting in sub-Saharan Africa. Development and implementation of synoptic templates in collaboration with key stakeholders has the potential to improve pathology reporting practices in low-resource settings.

Machine learning techniques, especially deep learning techniques such as convolutional neural networks, have been successfully applied to general image recognitions since their overwhelming performance at the 2012 ImageNet Large Scale Visual Recognition Challenge. Recently, such techniques have also been applied to various medical, including histopathological, images to assist the process of medical diagnosis. In some cases, deep learning–based algorithms have already outperformed experienced pathologists for recognition of histopathological images. However, pathological images differ from general images in some aspects, and thus, machine learning of histopathological images requires specialized learning methods. Moreover, many pathologists are skeptical about the ability of deep learning technology to accurately recognize histopathological images because what the learned neural network recognizes is often indecipherable to humans. In this review, we first introduce various applications incorporating machine learning developed to assist the process of pathologic diagnosis, and then describe machine learning problems related to histopathological image analysis, and review potential ways to solve these problems.