基于人工智能深度学习的腹针穴位辅助定位系统的研发  

Development of an abdominal acupoint localization system based on AI deep learning

作　　者：张默 李宇明 史宗明[1] ZHANG Mo;LI Yuming;SHI Zongming(Department of TCM,Integration of Traditional Chinese and Western Medicine,First Hospital of Peking University,Beijing 100034,China;Department of Electronic Engineering,City University of Hong Kong,Hong Kong 201337)

机构地区：[1]北京大学第一医院中医、中西医结合科,北京100034 [2]香港城市大学电子工程系,中国香港201337

出　　处：《中国针灸》2025年第3期391-396,共6页Chinese Acupuncture & Moxibustion

基　　金：国家中医药管理局中医传承与创新“百千万”人才工程(岐黄工程)——国家中医药领军人才支撑计划项目:国中医人教函[2022]6号;山西省中医药管理局中医药创新团队建设项目:zyytd2024030;中央高水平医院临床科研业务费资助(北京大学第一医院青年临床研究专项基金)项目:2022CR107。

摘　　要：本研究旨在开发一套基于计算机视觉和卷积神经网络(CNNs)的腹针穴位辅助定位系统。此系统针对腹部穴位定位的问题,构建并训练了一套多任务CNNs架构,分别定位出神阙穴及人体边界;利用网络定位到的神阙穴进一步推导出上脘、曲骨及双侧大横4个关键穴位特征点;利用仿射变换矩阵实现由图像空间坐标到穴位模板空间的精确映射,从而实现腹部穴位的精准定位。通过测试验证,该系统能够精确地在图像中识别并定位腹部的针灸穴位。此定位系统的研制为中医远程教育、辅助诊断治疗及智能针灸机器人等先进中医设备的研制提供了技术支撑,有助于推动中医针灸的标准化和智能化进程。This study aims to develop an abdominal acupoint localization system based on computer vision and convolutional neural networks(CNNs).To address the challenge of abdominal acupoint localization,a multi-task CNNs architecture was constructed and trained to locate the Shenque(CV8)and human body boundaries.Based on the identified Shenque(CV8),the system further deduces key characteristics of four acupoints:Shangwan(CV13),Qugu(CV2),and bilateral Daheng(SP15).An affine transformation matrix is applied to accurately map image coordinates to an acupoint template space,achieving precise localization of abdominal acupoints.Testing has verified that this system can accurately identify and locate abdominal acupoints in images.The development of this localization system provides technical support for TCM remote education,diagnostic assistance,and advanced TCM equipment,such as intelligent acupuncture robots,facilitating the standardization and intelligent advancement of acupuncture.

关 键 词：穴位定位 卷积神经网络 机器学习 针灸 

分 类 号：R245-33[医药卫生—针灸推拿学] TP18[医药卫生—中医临床基础] TP391.41[医药卫生—中医学]