摘要
迁移学习技术是一种将源域中的知识迁移到目标域任务上的新的机器学习方法,较好地解决了机器学习方法在医疗领域应用时缺少足够的有标签数据的情况。本文首先按照迁移学习方法的分类介绍了不同迁移学习方法的基本思想,并回顾了基于实例、特征、模型、关系的迁移学习研究进展。其次,结合实际案例,重点介绍了迁移学习在医疗文本数据处理,基于文本、图像、语音的疾病诊断中的应用。最后,对在医疗领域有发展潜力的迁移学习方法进行了应用展望。本文对于更好地解决传统机器学习或深度学习方法在医疗领域中的局限性提供了参考,对相关领域的工作者具有一定的借鉴价值。
Transfer learning technology is a new machine learning method that transfers knowledge from source domain to target domain.It solves the problem that machine learning methods lack enough labeled data when applied in medical field.Firstly,according to the classification of transfer learning methods,the basic ideas of different transfer learning methods were introduced,and the research progress of transfer learning based on case,feature,model and relationship was reviewed in this paper.Secondly,combined with practical cases,the application of migration learning in medical text data processing and disease diagnosis based on text,image and voice were mainly introduced.Finally,the application prospect of transfer learning method which has potential in the medical field was presented.This paper provides a reference for solving the limitations of traditional machine learning or deep learning methods in the medical field,and can be referredbyresearchers in related fields.
作者
吴骋
秦婴逸
李冬冬
王志勇
WU Cheng;QIN Yingyi;LI Dongdong;WANG Zhiyong(Department of Military Health Statistics,Naval Medical University,Shanghai 200433,China;Department of Information,First Affiliated Hospital,Naval Medical University,Shanghai 200433,China)
出处
《中国医疗设备》
2020年第9期161-164,172,共5页
China Medical Devices
基金
上海市自然科学基金(19ZR1469800)
全军后勤科研重大项目子题(AWS15J005-4)
海军军医大学第一附属医院“234学科攀峰计划”(2019YBZ002)。
关键词
迁移学习
医疗领域
机器学习
深度学习
transfer learning
medical field
machine learning
deep learning
