首页|人工智能大模型医学应用研究

人工智能大模型医学应用研究

扫码查看
近年来,以自然语言处理和视频图像分析为主的人工智能大模型技术得到快速发展,其基本特征是聚焦相关应用领域的共性需求,通过大数据、强算力和复杂算法的高效协同与深度融合,构建通用预训练模型,广泛适配下游任务,有力提高模型的处理性能与研发效率.因此,大模型技术为医学人工智能高质量发展提供了难得契机.本文通过全面梳理国内外大模型的研究进展、关键技术与核心算法,分析总结生物医学领域一系列标准数据集和预训练模型的发展特点,结合医学人工智能的研发实践,深入剖析医学领域大模型构建的应用需求、解决思路与研发经验,助力推动医学大模型创新发展.
Research on a massively large artificial intelligence model and its application in medicine
Recent years have witnessed rapid advancements in massively large artificial intelligence(AI)models based on natural language processing and video image analysis.In order to meet the requirements of relevant application fields,the universal pretraining model is developed by the efficient collaboration and deep integration of big data,large-scale computing power,and complex algorithms.The model demonstrates adaptability to a wide range of downstream tasks.In addition,the massively large model presents considerable opportunities for advancing the quality of medical AI development.Therefore,this paper comprehensively analyzes the progress of massively large models within domestic and international contexts in recent years,with an emphasis on their key technologies and algorithmic framework.Meanwhile,the developmental characteristics of a series of standard datasets and pretraining models in the biomedical field have been presented in detail.Incorporating our team's practical experience in medical AI research and development,we undertake a comprehensive analysis of the application requirements,our solutions and experiences related to the construction of massively large models in the medical field and persistently promote innovation and development within the realm of large-scale medical models.

medicineartificial intelligencemassively large modelnatural language processingmedical image analysis

郭华源、刘盼、卢若谷、杨菲菲、徐洪丽、庄严、黄高、宋士吉、何昆仑

展开 >

中国人民解放军总医院医学创新研究部医学人工智能研究中心,北京 100853

中国人民解放军总医院医学创新研究部医学大数据研究中心,北京 100853

中国人民解放军总医院第四医学中心心血管内科,北京 100048

清华大学自动化系,北京 100084

展开 >

医学 人工智能 大模型 自然语言处理 医学图像分析

工信部科技司产业技术基础公共服务平台项目(2020)工信部科技司产业技术基础公共服务平台项目(2020)

2020-0103-3-12020-0103-3-1-8

2024

中国科学(生命科学)
中国科学院

中国科学(生命科学)

CSTPCD北大核心
影响因子:0.725
ISSN:1674-7232
年,卷(期):2024.54(3)
  • 1
  • 137