Li Jiahao, Fei Zhennan, Xie Yingjiang, Deng Da, Ming Xingcheng, Niu Fu
Academy of Systems Engineering of Academy of Military Science of Chinese PLA, Beijing, China.
Chin Med. 2025 Jul 22;20(1):116. doi: 10.1186/s13020-025-01173-3.
The development of deep learning has brought unprecedented opportunities for automatic acupoint localization, surmounting many limitations of traditional methods and machine learning, and significantly propelling the modernization of Traditional Chinese Medicine (TCM). We comprehensively review and analyze relevant research in this field in recent years, and examine the principles, classifications, commonly used datasets, evaluation metrics and application fields of acupoint localization algorithms based on deep learning. We categorize them by body part, algorithm architecture, localization strategy, and image modality, and summarize their characteristics, pros and cons, and suitable application scenarios. Then we sieve out representative datasets of high value and wide application, and detail some key evaluation metrics for better assessment. Finally, we sum up the application status of current automatic acupoint localization technology in various fields, hoping to offer practical reference and guidance for future research and practice.
深度学习的发展为穴位自动定位带来了前所未有的机遇，克服了传统方法和机器学习的诸多局限性，有力地推动了中医现代化进程。我们全面回顾和分析了近年来该领域的相关研究，考察了基于深度学习的穴位定位算法的原理、分类、常用数据集、评估指标及应用领域。我们按身体部位、算法架构、定位策略和图像模态对其进行分类，总结其特点、优缺点及适用应用场景。然后筛选出具有高价值和广泛应用的代表性数据集，并详细介绍一些关键评估指标以进行更好的评估。最后，我们总结了当前穴位自动定位技术在各个领域的应用现状，希望能为未来的研究和实践提供实际参考和指导。
