人工智能在中药鉴定中的应用综述
该综述检索PubMed、Google Scholar和CNKI,总结人工智能在中药图像识别、质量控制、活性成分与毒性评估及产地鉴定中的应用与局限。AI可提升鉴定准确性与效率,结合光谱、色谱等技术使分析更客观,但数据质量、可解释性、监管审批及小样本和地域差异限制模型泛化,需加强标准化与跨学科验证。
要点
- AI在中药图像识别中可提高鉴定准确性与效率。
- AI结合光谱和感官检测可更客观地进行质量控制。
- AI模型泛化性受小样本和地域差异限制,需严格验证。
核心结论
- AI在中药图像鉴定、质量鉴定、活性成分检测、产地鉴定和毒性评估中有应用前景,但存在数据质量、可解释性和监管审批等持续挑战。
研究设计与样本
- 综述检索PubMed、Google Scholar和CNKI,覆盖图像识别、质量控制、产地鉴定、植物化学分析和毒性评估等AI在中药应用的研究。
主要发现
- AI在中药鉴定中提高准确性和效率;与光谱和感官检测结合使质量控制分析更客观;与色谱和多技术方法结合支持复杂成分和毒性评估。
- 产地鉴定中支持向量机(SVM)等模型表现良好,但泛化性受小样本和区域差异限制。
临床怎么用
- 可将AI图像识别用于中药鉴定以提升准确性和效率。
- 质量控制中可结合AI与光谱、感官检测进行更客观分析。
- 产地鉴定可试用SVM等模型,但需注意小样本和区域差异对泛化性的限制。
局限性
- 数据质量、可解释性和监管审批是持续挑战,需加强标准化、严格验证和跨学科合作。
- 产地鉴定模型泛化性受小样本和区域差异限制。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
ETHNOPHARMACOLOGICAL RELEVANCE: The integration of artificial intelligence (AI) into pharmaceutical practice represents a paradigm shift, offering innovative solutions to long-standing challenges in the authentication and quality control of Chinese herbal medicine (CHM). AIM OF THE STUDY: This study aims to review the current applications and limitations of AI in the image identification, quality control, active ingredient and toxicity assessment, and origin identification of CHM, providing reference and guidance for the future development of AI-assisted traditional Chinese medicine (TCM) research and practice. MATERIALS AND METHODS: A comprehensive literature search was conducted in PubMed, Google Scholar, and CNKI to identify studies on the application of AI in TCM, covering image recognition, quality control, origin identification, phytochemical analysis, and toxicity assessment. RESULTS: The results show that AI offers significant advantages in the identification of CHM, improving both accuracy and efficiency. In quality control, the combination of AI with spectroscopic and sensory detection enables more objective analyses. When integrated with chromatographic and multi-technique approaches, AI supports the evaluation of complex components and toxicity. In origin identification, models such as support vector machines (SVM) performed well, though their generalizability remains limited by small sample sizes and regional differences. Persistent challenges in data quality, interpretability, and regulatory approval highlight the need for enhanced standardization, rigorous validation, and closer interdisciplinary collaboration. CONCLUSION: This review explores the applications and limitations of AI in TCM, covering image identification, quality identification, active ingredient detection, origin identification, and toxicity evaluation, providing new directions for the development of AI in TCM.
引用
- Jia S, Lai H, Bai H, et al. Harnessing the application of artificial intelligence in identification of traditional Chinese medicines. J Ethnopharmacol. 2025. PMID: 41052688. doi:10.1016/j.jep.2025.120704
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。