医学数字活检在肝病精准诊疗与学科建设的实践与思考

Medical digital biopsy in precision liver disease care and discipline development: practice and reflections

摘要:

医学数字活检（MC-Biopsy）是在人工智能技术与医疗大数据深度融合背景下提出的一种面向真实临床场景的数据整合与应用技术框架，旨在解决多源异构临床数据碎片化、难以支持动态分析与循证决策的现实问题。该框架以标准化数据治理为核心，通过系统整合检验、影像、病理、病历文本及随访等多模态信息，构建覆盖疾病全周期的纵向专病数据库，并将多种人工智能方法嵌入既有临床数据结构与业务流程之中，实现高价值临床数据向可复用证据的转化。本文以笔者中心初步构建并运行的MC-Biopsy技术体系为例，系统介绍其总体构想、核心技术架构及在真实临床环境中的应用实践，重点阐述其在肝癌高危人群风险分层、肿瘤诊断分期、疗效评估与预后分析等关键场景中的应用路径及面临的挑战。实践表明，MC-Biopsy通过前置嵌入aMAP（age-Male-ALBi-Platelets score）等风险模型、整合时序实验室指标与影像数据，可支持个体化、分层化的风险管理；同时，依托自然语言处理技术对病历文本进行结构化解析，有助于提升临床信息质量并促进规范化人才培养；通过影像、病理与临床数据的多模态融合，逐步形成可溯源的数字表型库，推动肝病研究由单一模态、静态指标分析向基于纵向病程与真实世界结局的多维研究范式转变。进一步结合学科建设视角，本文讨论了MC-Biopsy在问题导向转化研究及复合型医学人才培养中的潜在价值。总体而言，MC-Biopsy为支撑临床精准诊疗、真实世界研究及学习型医疗体系建设提供了一种具有可复制性和推广潜力的技术参照。

Abstract:

Medical digital biopsy (MC-Biopsy) is a data integration and application framework developed in the context of the growing convergence of artificial intelligence and large-scale clinical data, aiming to address the fragmentation and limited reusability of heterogeneous clinical information in real-world practice. Centered on standardized data governance, MC-Biopsy systematically integrates laboratory results, imaging, pathology, clinical narratives, and longitudinal follow-up data to construct a disease-specific, multimodal database covering the entire disease course, while embedding artificial intelligence-based methods into existing clinical data structures and workflows to facilitate the transformation of high-value clinical data into reusable evidence. Taking the MC-Biopsy framework preliminarily established and implemented at our institution as an example, this article describes its overall concept, core technical architecture, and real-world deployment, with a particular focus on its application in key scenarios of liver disease management, including risk stratification of high-risk populations, tumor diagnosis and staging, treatment response assessment, and prognosis evaluation. In practice, MC-Biopsy supports individualized and stratified risk management by integrating established models such as the aMAP (age-Male-ALBi-Platelets score) into outpatient and follow-up workflows and by linking longitudinal laboratory and imaging data. In addition, natural language processing-based structuring of clinical narratives contributes to improving data quality and supporting standardized clinical training. Through multimodal integration of imaging, pathology, and clinical data, a traceable digital phenotype repository is progressively established, enabling liver disease research to shift from single-modality, static indicator-based analyses toward multidimensional investigations grounded in longitudinal disease trajectories and real-world outcomes. From the perspective of discipline development, this study further discusses the potential role of MC-Biopsy in problem-oriented translational research and interdisciplinary medical education. Overall, MC-Biopsy represents a reproducible and scalable technical reference with the potential to support precision clinical care, real-world research, and the development of learning healthcare systems.