药物非临床安全性评价毒性病理学深度学习方法和应用概述  

Overview of approaches and applications of deep learning in toxicologic pathology during nonclinical safety evaluation of drugs

作　　者：张頔[1] 闫振龙 杨艳伟[1] 屈哲[1] 霍桂桃[1] 李双星 林志[1] 吕建军 ZHANG Di;YAN Zhen-long;YANG Yan-wei;QU Zhe;HUO Gui-tao;LI Shuang-xing;LIN Zhi;LYU Jian-jun(National Center for Safety Evaluation of Drugs,National Institutes for Food and Drug Control,Beijing Key Laboratory,Beijing 100176,China;Innostar Biotechnology Nantong Co.,Ltd.,Nantong 226133,China;Hubei Topgene Xinsheng Biotechnology Co.,Ltd.,Wuhan 430207,China)

机构地区：[1]中国食品药品检定研究院,国家药物安全评价监测中心,北京市重点实验室,北京100176 [2]益诺思生物技术南通有限公司,南通226133 [3]湖北天勤鑫圣生物科技有限公司,武汉430207

出　　处：《中国新药杂志》2023年第21期2147-2154,共8页Chinese Journal of New Drugs

基　　金：中国食品药品检定研究院学科带头人课题(2021X2)。

摘　　要：药物非临床安全性评价毒理学试验组织病理学数据在药物发现和开发过程中非常关键,是监管部门在实验动物中评估药物相关毒性及其对临床试验人类安全性影响所必需的。随着全切片图像、数字病理学和算法研究的进步,人工智能已广泛应用于人类临床医学实践。但机器学习,尤其是深度学习在非临床毒理学试验毒性病理学的进展比较缓慢。本文简要综述了毒性病理学在药物发现和开发中的作用、人工智能和深度学习方法概述、深度学习方法在毒性病理学中的应用以及在毒性病理学中应用深度学习方法所面临的挑战,以期为我国药物非临床安全性评价毒理学试验组织病理学检查应用深度学习方法提供一定参考。Histopathology data of toxicology studies during nonclinical safety evaluation of drugs are critical during the drug discovery and development process that is necessary for regulatory bodies to assess drug-related toxicity in laboratory animals and its impact on human safety in clinical trials.Artificial intelligence(AI)has been widely used in clinical medical practice with progress in the whole slide image(WSI),digital pathology and algorithms.However,the progress of machine learning(ML)-in particular,deep learning(DL)-has been rather slow in toxicologic pathology of nonclinical toxicology studies.This paper briefly reviews the role of toxicologic pathology in drug discovery and development,overview of AI and approaches of DL,applications of approaches of DL in toxicologic pathology,as well as the challenges in implementation of approaches of DL in toxicologic pathology,in order to provide reference for implementation of approaches of DL in toxicologic pathology for toxicology studies during nonclinical safety evaluation of drugs in China.

关 键 词：药物 非临床安全性评价 毒性病理学 人工智能 深度学习 

分 类 号：R99[医药卫生—毒理学]