Developing AI-powered tools for histopathology: opportunities, challenges and new friends along the way

Diagnostic Histopathology Pub Date : 2025-05-01 DOI:10.1016/j.mpdhp.2025.03.003

Sharon Ruane, Korsuk Sirinukunwattana, Anna Kotanska, Alan Aberdeen

{"title":"Developing AI-powered tools for histopathology: opportunities, challenges and new friends along the way","authors":"Sharon Ruane, Korsuk Sirinukunwattana, Anna Kotanska, Alan Aberdeen","doi":"10.1016/j.mpdhp.2025.03.003","DOIUrl":null,"url":null,"abstract":"<div><div>Advancements in slide digitization and artificial intelligence (AI) offer immense transformative potential for pathology. While much focus is placed on AI's potential to automate tasks and standardize assessments, its most significant impact may come from uncovering novel tissue-based biomarkers and deepening our understanding of disease. Properly developed and validated AI-based tools could improve the quantification of known biomarkers, identify novel tissue-based biomarkers beyond human perception, and enable inter-sample comparisons to reveal disease subtypes and heterogeneity. This article draws on our practical experience working with pathologists to develop AI-based algorithms for assessing bone marrow in patients with blood cancer. We provide an overview of approaches to AI model development from perspectives typically of most interest to our pathologist collaborators: the data requirements and the resulting model interpretability. We discuss the limitations of the current manual assessment of histopathology samples and the opportunities provided by AI-based approaches. We then address major challenges in the field and discuss how an interdisciplinary approach combining expertise across disciplines is essential to maximizing the potential of AI-powered pathology tools.</div></div>","PeriodicalId":39961,"journal":{"name":"Diagnostic Histopathology","volume":"31 5","pages":"Pages 277-283"},"PeriodicalIF":0.0000,"publicationDate":"2025-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Diagnostic Histopathology","FirstCategoryId":"1085","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S1756231725000349","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}

引用次数: 0

Abstract

Advancements in slide digitization and artificial intelligence (AI) offer immense transformative potential for pathology. While much focus is placed on AI's potential to automate tasks and standardize assessments, its most significant impact may come from uncovering novel tissue-based biomarkers and deepening our understanding of disease. Properly developed and validated AI-based tools could improve the quantification of known biomarkers, identify novel tissue-based biomarkers beyond human perception, and enable inter-sample comparisons to reveal disease subtypes and heterogeneity. This article draws on our practical experience working with pathologists to develop AI-based algorithms for assessing bone marrow in patients with blood cancer. We provide an overview of approaches to AI model development from perspectives typically of most interest to our pathologist collaborators: the data requirements and the resulting model interpretability. We discuss the limitations of the current manual assessment of histopathology samples and the opportunities provided by AI-based approaches. We then address major challenges in the field and discuss how an interdisciplinary approach combining expertise across disciplines is essential to maximizing the potential of AI-powered pathology tools.

查看原文本刊更多论文

为组织病理学开发人工智能工具：机遇、挑战和新朋友

幻灯片数字化和人工智能（AI）的进步为病理学提供了巨大的变革潜力。虽然人工智能在自动化任务和标准化评估方面的潜力备受关注，但其最重大的影响可能来自于发现新的基于组织的生物标志物，并加深我们对疾病的理解。适当开发和验证的基于人工智能的工具可以改善已知生物标志物的量化，识别超越人类感知的新的基于组织的生物标志物，并使样本间比较能够揭示疾病亚型和异质性。本文借鉴了我们与病理学家合作的实践经验，开发基于人工智能的算法来评估血癌患者的骨髓。我们从病理学家合作者通常最感兴趣的角度概述了人工智能模型开发的方法：数据要求和由此产生的模型可解释性。我们讨论了目前人工评估组织病理学样本的局限性和基于人工智能的方法提供的机会。然后，我们讨论了该领域的主要挑战，并讨论了跨学科方法结合跨学科专业知识对于最大限度地发挥人工智能病理学工具的潜力至关重要。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Diagnostic Histopathology Medicine-Pathology and Forensic Medicine

CiteScore

1.30

自引率

0.00%

发文量

期刊介绍： This monthly review journal aims to provide the practising diagnostic pathologist and trainee pathologist with up-to-date reviews on histopathology and cytology and related technical advances. Each issue contains invited articles on a variety of topics from experts in the field and includes a mini-symposium exploring one subject in greater depth. Articles consist of system-based, disease-based reviews and advances in technology. They update the readers on day-to-day diagnostic work and keep them informed of important new developments. An additional feature is the short section devoted to hypotheses; these have been refereed. There is also a correspondence section.