| 475 | 0 | 6 |
| 下载次数 | 被引频次 | 阅读次数 |
目的 旨在探索分割一切模型(Segment Anything Model, SAM)系列图像分割大模型在数字化中医望诊研究中的应用，有望减少在中医理论指导下为训练舌、面、目、手等图像的特定部位分割模型过程中所需的大量人工标注工作。方法 首先采用快速分割一切模型(Fast Segment Anything Model, FastSAM)完成无监督分割舌、面、目等图像，评估其分割完整的舌部、面部和目部的效果，表明FastSAM能够高效分离中医望诊所需的特定部位。其次采用SAM在框选提示下可弥补FastSAM对舌部的裂纹、面部的额头、眼部的眼络、手部的指甲等细节部分分割结果不理想的缺点。结果 FastSAM在舌部和面部的交并比(Intersection over Union, IOU)值均超过0.9,目部超过0.8,表明在整体或部分图像分割方面表现良好，并且SAM能针对FastSAM无法处理的细节部位展现出良好的分割效果。结论 SAM系列模型为后续中医理论指导下实现中医望诊图像特征分析提供一种新的分割方法，实现零样本分割各类中医望诊维度图像，从而有望大量减少中医望诊图像特征的人工标注工作量，同时有效提升中医望诊的准确性和可靠性。
Abstract：Objective This study explored the application of the Segment Anything Model(SAM) series in digitalizing traditional Chinese medicine(TCM) inspection diagnosis research, aiming to reduce extensive manual annotation required for specific part segmentation models of tongue, face, eye, and hand under TCM theoretical guidance. Methods First, the Fast Segment Anything Model(FastSAM) was used to perform unsupervised segmentation on images of the tongue, face and eyes. The evaluation of its ability to fully segment these areas indicated that FastSAM is highly effective in isolating the specific regions needed for TCM diagnostics. Secondly, SAM used with bounding box prompts, can compensate for FastSAM's suboptimal performance in segmenting finer details such as cracks in the tongue, the forehead on the face, blood vessels in the eyes, and fingernails on the hands. Results FastSAM achieved Intersection over Union(IOU) values exceeding 0.9 for both the tongue and facial areas, and over 0.8 for the ocular area, indicating its good performance in overall or partial image segmentation. Additionally, SAM demonstrated the excellent segmentation results for detailed areas that FastSAM cannot handle. Conclusion The SAM series of models offers a novel segmentation approach for the subsequent analysis of image features in TCM inspection guided by TCM theory. It enables zero-shot segmentation of various TCM inspection images across different dimensions, potentially significantly reducing the manual annotation workload for TCM inspection image features. At the same time, it effectively enhances the accuracy and reliability of TCM inspection.
[1] 张孟之，高洁，李文，等.人工智能时代下的中医四诊客观化研究初探[J].贵阳中医学院学报，2019,41(1):100-102.
[2] 田飞，常俊，赵静，等.中医四诊客观化研究面临的主要问题与挑战[J].天津中医药，2015,32(7):445-448.
[3] KONG L,HUANG M,ZHANG L,et al.Enhancing diagnostic images to improve the performance of the Segment Anything Model in medical image segmentation [J].Bioengineering,2024,11:270.
[4] 吴曈，胡浩基，冯洋，等.分割一切模型(SAM)在医学图像分割中的应用[J].中国激光，2024,51(21):27-42.
[5] 封晓燕，田琪，徐云峰，等.基于双编码特征提取路径的舌体分割方法[J].生物医学工程研究，2024,43(2):123-128.
[6] 余兆钗，张祖昌，李佐勇，等.融合多颜色分量的舌图像阈值分割算法研究[J].计算机应用与软件，2019,36 (5):199-203,248.
[7] 高清河，刚晶，王和禹，等.基于GVFSnake模型的舌像分割研究[J].科技视界，2018(6):131-132.
[8] ZHOU J,ZHANG Q,ZHANG B,et al.TongueNet:A precise and fast tongue segmentation system using U-Net with a morphological processing layer [J].Applied Sciences,2019,9(15):3128.
[9] 卢运西，李晓光，张辉，等.中医舌象分割技术研究进展：方法、性能与展望[J].自动化学报，2021,47(5):1005-1016.
[10] 王一丁，孙常浩，崔家礼，等.基于深度学习的舌裂分割算法研究[J].世界科学技术-中医药现代化，2021,23(9):3065-3073.
[11] LI X,WANG D,CUI Q.WLDF:Effective statistical shape feature for cracked tongue recognition [J].Journal of Electrical Engineering and Technology,2017,12(1):420-427.
[12] 彭开来.基于深度学习的齿痕舌图像分割方法[D].杭州：杭州电子科技大学，2022:9.
[13] 马圆港，冯跃，林卓胜，等.基于文献计量学和文本分析法的智能中医面诊分区方法系统性综述[J].世界科学技术-中医药现代化，2024,26(5):1132-1141.
[14] 詹文栋，龚庆悦，朱金阳，等.基于改进U-Net的面部红外热成像的分割[J].计算机时代，2023,(10):89-93,99.
[15] LIN S,LI Z G,FU B W,et al.Feasibility of using deep learning to detect coronary artery disease based on facial photo [J].Eur Heart J,2020,41(46):4400-4411.
[16] CHEDDAD A,MOHAMAD D,MANAF A A.Exploiting Voronoi diagram properties in face segmentation and feature extraction [J].Pattern Recognition,2008,41(12):3842-3859.
[17] 姜益宏，周伟民，李莲，等.目诊在心血管疾病中应用的理论基础及研究进展[J].湖南中医药大学学报，2024,44(6):1140-1144.
[18] 李翔，邓颖，蒋鹏飞，等.基于眼底图像视网膜血管分割、分类方法的中医目诊研究[J].湖南中医药大学学报，2021,41(3):406-411.
[19] WANG YANXIA,WANG JINGYI,GUO PING.Eye-UNet:a UNet-based network with attention mechanism for low-quality human eye image segmentation[J].Signal,Image and Video Processing,2022(4):1097-1103.
[20] 王瑞云.中医甲诊系统的研究[D].天津：天津大学，2018.
[21] 赵紫娟，强彦，赵涓涓，等.图像智能处理方法在中医中的应用与挑战[J].太原理工大学学报，2022,53(3):405-419.
[22] 刘慧琳，韩吉，李福凤.基于复杂网络对舌、面诊图像特征提取及分割分类文献的可视化研究[J].世界科学技术-中医药现代化，2024,26(5):1336-1343.
[23] HE S,BAO R,LI J,et al.Computer vision benchmark segment-anything model (SAM) in medical images:Accuracy in 12 datasets [J].arXiv preprint,arXiv,2023(10):2304.
[24] WANG J L,JIA R D,ZHENG J,et al.Segment anything model-based method for precise froth size determination in flotation process[J].Chemical Engineering Science,2025(12):121657.
[25] QU C,ZHANG T,QIAO H,et al.Abdomenatlas-8k:Annotating 8,000 CT volumes for multi-organ segmentation in three weeks [C].In:Thirty-seventh Conference on Neural Information Processing Systems Datasets and Benchmarks Track,2023(2):32.
[26] ZHU Y,XIONG C,ZHAO H,et al.SAM-Att:A prompt-free SAM-related model with an attention module for automatic segmentation of the left ventricle in echocardiography [J].IEEE Access,2024(12):50335-50346.
[27] ALZATE-GRISALES YJA,MORA-RUBIO A,GARCíA-GARCíA F,et al.SAM-UNETR:Clinically significant prostate cancer segmentation using transfer learning from large model [J].IEEE Access,2023,11:118217-118228.
[28] 王卫东，刘延，邱实，等.用于铁路场景语义分割的改进动态图卷积神经网络[J].计算机辅助设计与图形学学报，2025,37(1):139-147.
[29] H ALI,M F BULBUL,Z SHAH."Prompt engineering in medical image segmentation:An overview of the paradigm shift," in 2023 IEEE International Conference on Artificial Intelligence,Blockchain,and Internet of Things (AIBThings)[J].IEEE,2023(1):1-4.
[30] 谭鲲鹏，唐甲锋，赵志斌，等.基于视觉大模型的激光粉末床熔融铺粉缺陷检测[J].中国激光，2024,51(10):275-284.
[31] ZHANG L,DENG X,LU Y.Segment Anything Model (SAM) for medical image segmentation:A preliminary review [C]// 2023 IEEE International Conference on Medical Imaging,2023:7.
[32] 倪佳宁，王启明，朱瑞虎，等.基于零样本深度学习的码头表观剥落病害区域分割与定量计算[J].水运工程，2024,(2):162-168.
[33] 黄帝内经集注素问[M].天津：天津科学技术出版社，2004:21-32.
[34] 刘冬华，蒋鹏飞，刘培，等.中医目诊的诊断原理及常见诊病部位[J].中华中医药杂志，2022,37(9):5294-5298.
[35] 张婷，王琳.中医舌诊客观化研究进展[J].实用中医内科杂志，2024,38(6):82-85.
[36] 杨紫阳，卢丙辰.中医眼科的全息观[J].中医临床研究，2017,9(6):5-7.
[37] ASIEDU KOFI,KRISHNAN ARUN V,KWAI NATALIE,et al.Conjunctival microcirculation in ocular and systemic microvascular disease.[J].Clinical & experimental optometry,2023(7):1-9.
[38] 王今觉.望目辨证诊断学[M].北京：中国中医药出版社，2013:76-224.
[39] 宋添力，马婧，李海霞，等.基于中医目诊理论和白睛成像AI及光学技术分析高血压病患者目络特征及发病机制的关联性研究[J].中华中医药学刊，2024,42(12):15-19,285-287.
[40] 刘璇，张琼艺，全虹翰，等.基于数字化目诊技术解析高血压的目络特征及与证候要素的相关性[J].中国中医基础医学杂志，2024,30(3):413-417.
[41] 朱贵冬，沈理，王今觉.基于外推跟踪的眼部白睛血管自动提取方法[J].计算机工程，2005(17):6-8.
[42] 靳士英，黄子天.舌诊的发展史略[J].现代医院，2022,22(3):469-475.
[43] 陈溶瑾，杜雅琦.《伤寒杂病论》舌诊浅析[J].河南中医，2019,39(1):9-13.
[44] 陈家旭.中医诊断学[M].北京：中国中医药出版社，2015:234-260.
[45] 江涛，屠立平，许家佗.中医舌象智能诊断技术研究述评及展望[J].中国中医药信息杂志，2024,31(7):182-187.
[46] 刘璐，周艳霞，亓鲁光.面诊在内分泌疾病诊疗中的应用[J].四川中医，2014,32(3):33-34.
[47] 胡洁娴，刘旺华.中医面部色诊现代研究进展[J].中医学报，2020,35(5):1001-1005.
[48] 夏淑洁，李灿东.基于整体观念的五辨论治思维探析[J].天津中医药，2020,37(2):158-161.
[49] 关茜，徐莹，杨帅，等.中医面诊特征与疾病关系探究[J].中华中医药杂志，2022,37(2):902-905.
[50] 朱方石.《内经》面部五色诊浅析[J].中华中医药学刊，1990(3):17-18.
[51] 位庚，周睿，李福凤.中医面诊脏腑分属理论的研究概况[J].求医问药(下半月),2013,11(2):340-341.
[52] 宋·钱乙.小儿药证直诀[M].北京：人民卫生出版社，2006:24.
[53] 刘舒悦，焦媛，张若诗，等.中医甲诊理论源流与临床应用探微[J].中国中医药图书情报杂志，2022,46(5):46-49.
[54] 王文华，李捷珈.指甲诊病[M].上海：上海中医学院出版社，1990:23.
[55] 孙朝润.中医学对指甲的认识[J].中医研究，2020,33(12):10-12.
[56] MA J,HE Y,LI F,et al.Segment anything in medical images [J].Nat Commun,2024,15:654.
[57] YUYANG SHENG,SOPHIA BANO,MATTHEW J CLARKSON,et al.Surgical-DeSAM:decoupling SAM for instrument segmentation in robotic surgery.[J].International journal of computer assisted radiology and surgery,2024(7):1267-1271.
[58] WANG C Y,CHEN H B,ZHOU X,et al.SAM-IE:SAM-based image enhancement for facilitating medical image diagnosis with segmentation foundation model[J].Expert Systems With Applications,2024(PC):123795.
[59] 高诗岳，杨珺涵，张世祺，等.“手机端中医舌诊疾病分析系统”及舌诊仪研究进展[J].实用中医内科杂志，2024,38(11):75-77.
[60] GEERT LITJENS,THIJS KOOI,BABAK EHTESHAMI BEJNORDI,et al.A survey on deep learning in medical image analysis[J].Medical Image Analysis,2017(1):60-88.
[61] ZHOU S KEVIN,GREENSPAN HAYIT,DAVATZIKOS CHRISTOS,et al.A review of deep learning in medical imaging:Imaging traits,technology trends,case studies with progress highlights,and future promises.[J].Proceedings of the IEEE.Institute of Electrical and Electronics Engineers,2021(5):820-838.
[62] ZHANG Y C,SHEN Z R,JIAO R S.Segment anything model for medical image segmentation:Current applications and future directions[J].Computers in Biology and Medicine,2024(12):108238.
基本信息:
DOI：10.13193/j.issn.1673-7717.2025.09.005
中图分类号:R241.2;TP391.41
引用信息:
[1]刘迈,王晶,王曼睿等.基于SAM与FastSAM交互式分割技术在中医望诊客观化中的应用探究[J].中华中医药学刊,2025,43(09):21-25+263-265.DOI:10.13193/j.issn.1673-7717.2025.09.005.
基金信息:
国家重点研发计划项目(2022YFC3502301)