《生命科学》 2025, 37(12): 1615-1623

循证视角下多模态人工智能在慢性病中的应用

黄惠洁^1,2 , 于淑洁^1,2 , 管洲榆^1,2 , 吴　倩^1,2 , 盛　斌³ , 贾伟平^1,2 , 李华婷^1,2,*

¹上海交通大学医学院附属第六人民医院内分泌代谢科，上海市糖尿病研究所，上海市糖尿病 重点实验室，上海市糖尿病临床医学中心，上海 200233 ²上海交通大学医学院， 上海 200025 ³上海交通大学电子信息与电气工程学院计算机系，上海 200240

摘　要：

慢性病( 如心血管疾病、糖尿病、癌症及慢性呼吸系统疾病) 已成为全球公共卫生的重大挑战，严重威胁人类健康并造成沉重经济负担。传统筛查和管理方式存在识别效率低、依从性不足及资源分布不均等局限。随着医学人工智能的发展，多模态人工智能模型通过整合临床资料、影像学检查、组学数据、生理监测及生活方式信息，为慢性病的早期筛查、风险预测、辅助诊断及个体化干预提供了新路径。尽管如此，当前证据循证强度有限，未来需通过大规模、多中心、前瞻性研究和随机对照试验提升证据等级，解决模型推广性、可解释性及临床转化等挑战。总体而言，多模态人工智能在循证医学框架下有望成为慢性病防控的重要工具，推动精准医学与公共健康的协同发展。

通讯作者：李华婷 , Email:huarting99@sjtu.edu.cn

Application of multimodal artificial intelligence models in chronic disease management: an evidence-based perspective 

HUANG Hui-Jie^1,2 , YU Shu-Jie^1,2 , GUAN Zhou-Yu^1,2 , WU Qian^1,2 , SHENG Bin³ , JIA Wei-Ping^1,2 , LI Hua-Ting^1,2,*

¹Shanghai Clinical Center for Diabetes, Shanghai Key Laboratory of Diabetes Mellitus, Shanghai Diabetes Institute, Department of Endocrinology and Metabolism, Shanghai Sixth People′s Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai 200233, China ²Shanghai Jiao Tong University School of Medicine, Shanghai 200025, China ³Department of Computer Science and Engineering, Shanghai Jiao Tong University, Shanghai 200240, China

Abstract:

Chronic diseases, including cardiovascular disease, diabetes, cancer, and chronic respiratory disorders, have become major global public health challenges, causing substantial morbidity, mortality, and economic burden. Traditional approaches to screening and management are limited by inefficiency, poor adherence, and unequal distribution of healthcare resources. With the rapid advancement of artificial intelligence, multimodal models integrating clinical data, medical imaging, omics profiles, physiological monitoring, and lifestyle information have provided new opportunities for chronic disease prevention and control. Nevertheless, the current evidence remains limited in strength, and future efforts should focus on large-scale, multi-center, prospective studies and randomized controlled trials to enhance the evidence quality and address challenges such as model generalizability, interpretability, and clinical applicability. Overall, under the framework of evidence-based medicine, multimodal AI holds great promise as a transformative tool for chronic disease prevention and management, advancing both precision medicine and public health.

Communication Author：LI Hua-Ting , Email:huarting99@sjtu.edu.cn