深度学习构建舌象分类系统在银屑病中医证型识别中的应用

doi:10.3969/j.issn.1006-5725.2026.03.006

Abstract

Abstract:

Objective To develop a deep learning-based tongue image classification system for psoriasis to improve the objective differentiation between the patterns of Spleen Deficiency with Dampness Retention （SDDR） and Blood Stasis （BS）. Methods A total of 981 tongue images from psoriasis patients diagnosed with SDDR or BS were collected. An improved U-Net model， featuring a ResNet-34 encoder， bilinear interpolation upsampling， and optimized skip connections， was employed for automatic tongue region segmentation. Macenko color normalization and the Albumentations library were applied for data augmentation to mitigate variances from imaging devices and lighting conditions. A two-stage framework was constructed： the first stage precisely extracted the tongue body region， while the second stage utilized a Hybrid Model integrating EfficientNet-B3 and Swin-Tiny architectures for pattern classification. A cross-modal multi-head attention mechanism was introduced to fuse local textural and global structural features. Results The improved U-Net achieved superior performance in tongue segmentation， with a Dice coefficient of 0.98 and an IoU of 0.89， significantly outperforming the original U-Net （Dice 0.85）. For pattern classification， the Hybrid Model demonstrated the best overall performance， achieving a 5-fold cross-validation mean accuracy of 0.9816 and a mean AUC of 0.9993. The F1-score was significantly higher than those of individual models. Macenko normalization contributed to an 8.3% increase in F1-score. The inference time per image was 38 ms on an A10 GPU， meeting the requirement for clinical real-time application. Conclusion The constructed two-stage tongue image classification model effectively and accurately distinguishes between SDDR and BS patterns in psoriasis， significantly enhancing the objectivity of tongue diagnosis. It provides a reliable tool for pattern differentiation in Traditional Chinese Medicine and shows promising potential for clinical application.

Key words: psoriasis, spleen deficiency with dampness obstruction syndrome, blood stasis syndrome, traditional chinese medicine tongue diagnosis, deep learning

CLC Number:

R241.25

Sokhan CHOI,Dongning ZHU,Haowei XU,Jiaqi HUANG,Xiaoyu LI,Zhuohao LI,Xiufeng LIU,Yuhong YAN. Application of a deep learning-based tongue image classification system in TCM syndrome identification of psoriasis[J]. The Journal of Practical Medicine, 2026, 42(3): 395-405.

Figures/Tables 15

Fig.1

Fig.2

Fig.3

Fig.4

Fig.5

Tab.1

Fig.6

Fig.7

Fig.8

Fig.9

Fig.10

Fig.11

Fig.12

Tab.2

Tab.3

References 28

[1]	世界中医药学会联合会.寻常型银屑病中西医结合诊疗指南［J］.世界中医药， 2024， 19（17）： 2535-2544. doi： 10.3969/j.issn.1673-7202.2024.17.001 . doi: 10.3969/j.issn.1673-7202.2024.17.001
[2]	PANNU S， ROSMARIN D. Psoriasis in Patients with Metabolic Syndrome or Type 2 Diabetes Mellitus： Treatment Challenges ［J］. Am J Clin Dermatol， 2021， 22（3）： 293-300.doi： 10.1007/s40257-021-00590-y . Epub 2021 Feb 15. doi: 10.1007/s40257-021-00590-y
[3]	李思彤，王傲，白彦萍，等.银屑病中滤泡性辅助性T细胞及IL-21对HaCaT细胞增殖及细胞周期的影响［J］.实用医学杂志，2023，39（18）：2342-2348. doi ：10.3969/j.issn.1006-5725. 2023. 18.010 . doi: 10.3969/j.issn.1006-5725. 2023. 18.010
[4]	中华医学会皮肤性病学分会银屑病专业委员会. 中国银屑病诊疗指南（2023版）［J］. 中华皮肤科杂志，2023， 56（7）：573-625. doi：10.35541/cjd.20220839 . doi: 10.35541/cjd.20220839
[5]	赵宁，底婷婷，齐聪，等.银屑病病证结合模型的研究进展［J］.中国比较医学杂志，2022，32（10）：142-147. doi：10.3969/j.issn. 1671-7856.2022.10.019 . doi: 10.3969/j.issn. 1671-7856.2022.10.019
[6]	闫玉红，卢传坚，禤国维.寻常型银屑病核心病机探讨［J］.辽宁中医杂志，2012，39（6）：1013-1015.doi：10.13192/j.ljtcm. 2012.06.60.yanyh.095 . doi: 10.13192/j.ljtcm. 2012.06.60.yanyh.095
[7]	余沐原，罗亚，闫玉红，等.卢传坚从脾虚湿瘀论治岭南地区银屑病思路探讨［J］.广州中医药大学学报，2023，40（9）：2342-2347. doi：10.13359/j.cnki.gzxbtcm.2023.09.031 . doi: 10.13359/j.cnki.gzxbtcm.2023.09.031
[8]	刘柳，孙晓颖，莫美，等.中医药临床优势病种探讨——银屑病［J］.中国实验方剂学杂志，2025，31（5）：260-268. doi：10.13422/ j.cnki.syfjx.20250897 . doi: 10.13422/ j.cnki.syfjx.20250897
[9]	张玫，钟瑞，蒋寅，等. 数智中医广义望诊系统构建与思考［J］. 北京中医药大学学报，2025，48（8）：1037-1043. doi：10.3969/ j.issn.1006-2157.2025.08.001 . doi: 10.3969/ j.issn.1006-2157.2025.08.001
[10]	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. doi：10.13729/j.issn.1671-7813.Z20230974 . doi: 10.13729/j.issn.1671-7813.Z20230974
[11]	张林子，周武，张洪来.自动舌诊技术的研究进展［J］.中国中医基础医学杂志，2023，29（5）：871-876. doi：10.19945/j.cnki.issn.1006-3250.2023.05.035 . doi: 10.19945/j.cnki.issn.1006-3250.2023.05.035
[12]	KRITTANAWONG C， ZHANG H， WANG Z， et al. Artificial Intelligence in Precision Cardiovascular Medicine［J］. J Am Coll Cardiol， 2017， 69（21）： 2657-2664.doi： 10.1016/j.jacc. 2017. 03.571 . doi: 10.1016/j.jacc. 2017. 03.571
[13]	刘军，韩燕鸿，潘建科，等. 人工智能在中医骨伤科领域应用的现状与前景［J］. 中华中医药杂志， 2019， 34（8）： 3608-3612.
[14]	HU Y， WEN G， LUO M， et al. Fully-channel regional attention network for disease-location recognition with tongue images ［J］. Artif Intell Med， 2021， 118： 102110.doi： 10.1016/j.artmed. 2021.102110 . doi: 10.1016/j.artmed. 2021.102110
[15]	赵墨，彭红叶，汪九重，等. 慢性萎缩性胃炎不同危险分期患者中医舌象客观化研究［J］. 陕西中医， 2025， 46（6）： 756-760. doi： 10.3969/j.issn.1000-7369.2025.06.007 . doi: 10.3969/j.issn.1000-7369.2025.06.007
[16]	LI J， YUAN P， HU X， et al. A tongue features fusion approach to predicting prediabetes and diabetes with machine learning ［J］. J Biomed Inform， 2021， 115： 103693.doi： 10.1016/j.jbi. 2021. 103693 . doi: 10.1016/j.jbi. 2021. 103693
[17]	YUAN L， YANG L， ZHANG S， et al. Development of a tongue image-based machine learning tool for the diagnosis of gastric cancer： a prospective multicentre clinical cohort study ［J］. EClinicalMedicine， 2023， 57： 101834.doi： 10.1016/j.eclinm. 2023. 101834 . doi: 10.1016/j.eclinm. 2023. 101834
[18]	THIRUNAVUKKARASU U， UMAPATHY S， RAVI V， et al. Tongue image fusion and analysis of thermal and visible images in diabetes mellitus using machine learning techniques ［J］. Sci Rep， 2024， 14（1）： 14571.doi： 10.1038/s41598-024-64150-0 . doi: 10.1038/s41598-024-64150-0
[19]	MA C， ZHANG P， DU S， et al. Construction of Tongue Image-Based Machine Learning Model for Screening Patients with Gastric Precancerous Lesions［J］. J Pers Med， 2023，13（2）：271.doi： 10.3390/jpm13020271 . doi: 10.3390/jpm13020271
[20]	LI P， YI N， DING C S， et al. Research on classification diagnosis model of psoriasis based on deep residual network［J］. Dig Chin Med， 2021， 4（2）： 92-101. doi： 10.1016/j.dcmed.2021.06.003 . doi: 10.1016/j.dcmed.2021.06.003
[21]	秦志远，朱峻泷，张琛，等. 基于ResNet34-UNet的静脉超声图像分割方法研究［J］. 临床超声医学杂志，2022，24（1）：74-78. doi：10.3969/j.issn.1008-6978.2022.01.022 . doi: 10.3969/j.issn.1008-6978.2022.01.022
[22]	刘伟，陈锦明，刘波，等. 基于深度学习的舌体图像分割和舌色分类研究（英文）［J］. Digital Chinese Medicine， 2022， 5（3）： 253-63. doi：10.1016/j.dcmed.2022.10.002 . doi: 10.1016/j.dcmed.2022.10.002
[23]	TAN M， LE Q V.EfficientNet：Rethinking Model Scaling for Convolutional Neural Networks［J］. arXiv， 2019. doi：10.48550/arXiv.1905.11946 . doi: 10.48550/arXiv.1905.11946
[24]	LIU Z， LIN Y， CAO Y，et al.Swin Transformer： Hierarchical Vision Transformer using Shifted Windows［J］. arXiv， 2021. doi：10.48550/arXiv.2103.14030 . doi: 10.48550/arXiv.2103.14030
[25]	宋超，王斌，许家佗.基于深度迁移学习的舌象特征分类方法研究［J］.计算机工程与科学，2021，43（8）：1488-1496. doi：10.3969/j.issn.1007-130X.2021.08.018 . doi: 10.3969/j.issn.1007-130X.2021.08.018
[26]	吴欣，徐红，林卓胜，等. 深度学习在舌象分类中的研究综述［J］. 计算机科学与探索， 2023， 17（2）： 303-323. doi：10.3778/j.issn.1673-9418.2208052 . doi: 10.3778/j.issn.1673-9418.2208052
[27]	林胜，胡聿昕，李昀泽，等.“燥湿同病”论治慢性斑块型银屑病［J］.中医杂志，2025，66（22）：2388-2392. doi：10.13288/j.11-2166/r.2025.22.016 . doi: 10.13288/j.11-2166/r.2025.22.016
[28]	王冬阳，张洋，石志超.从热毒、血燥、血瘀、血虚论治银屑病［J］.广州中医药大学学报，2020，37（7）：1391-1394. doi：10.13359/j.cnki.gzxbtcm.2020.07.034 . doi: 10.13359/j.cnki.gzxbtcm.2020.07.034

模型	优势	局限性	Dice系数（x ± s）/%	本研究中的表现
基线 U-Net	简单有效，适用于多种图像分割任务	对复杂背景和光照变化敏感	0.85 ± 0.04	边缘定位精度不足
DeepLabv3+	大感受野，适合语义分割	参数量较大，计算资源消耗高	0.87 ± 0.05	显存占用16 GB，不适合实时应用
Mask R-CNN	高精度目标检测与分割	训练数据需求大（≥ 5 000样本），小样本泛化能力差	0.83 ± 0.06	误分割率高达15%
改进的U-Net	结合局部特征与全局上下文，鲁棒性强	需要预训练权重初始化	0.98 ± 0.03	边缘定位精度高，误分割率< 5%

模型类型	平均验证损失/%	平均准确率/%	平均F1分数/%	平均AUC指数/%
EfficientNet-B3	0.267 9 ± 0.004 8	0.962 4 ± 0.004 1	0.927 0 ± 0.005 3	0.996 6 ± 0.001 3
Swin-Tiny	0.252 7 ± 0.009 0	0.965 4 ± 0.007 0	0.968 4 ± 0.007 2	0.997 6 ± 0.001 4
Hybrid	0.227 5 ± 0.008 6	0.981 6 ± 0.006 1	0.981 6 ± 0.006 7	0.999 3 ± 0.000 6

模型类型	输入尺寸	参数量	单张图像推理时间
EfficientNet-B3	300 × 300	12 M	～25 ms
Swin-Tiny	224 × 224	28 M	～30 ms
Hybrid	双分支输入	40 M	～38 ms

Application of a deep learning-based tongue image classification system in TCM syndrome identification of psoriasis

RichHTML

PDF (PC)

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 15

References 28

Related Articles 15

Recommended Articles

Metrics

Comments

[1]	Yuchan ZHOU,Rongchang ZHENG,Huarun LI,Jinping HUANG,Si QIN,Ting LI,Zhenyu LU,Sihui LI,Xianwen LI,Mujin LI,Ju WEN. Expression of T_RM cells in the lesions of imiquimod⁃induced models of psoriasis in mice [J]. The Journal of Practical Medicine, 2025, 41(9): 1327-1331.
[2]	Yingmei HAN,Yijie LI,Heng ZHANG,Weiqing LI,Ze FENG,Feng WANG. The application value of deep learning in imaging studies for predicting the conversion of Alzheimer′s disease [J]. The Journal of Practical Medicine, 2025, 41(9): 1413-1424.
[3]	Jianbin LI,Rui. WU. Correlation study on pericyte depletion， eye⁃signs in blood stasis syndrome， and blood⁃brain barrier dysfunction in neuropsychiatric systemic lupus erythematosus [J]. The Journal of Practical Medicine, 2025, 41(6): 790-799.
[4]	Gaokai HU,Ya'nan NIU,Yukang GONG,Yang HU,Ruixuan XU,Wenshan. GAO. Research progress on the application of deep learning in lumbar spine disease [J]. The Journal of Practical Medicine, 2025, 41(6): 921-928.
[5]	Jinfen ZHENG,Cuiping SHI,Yunxia LING,Dehua ZHANG,Qianyu ZHAI,Lijia ZHU,Doukou JIANG,Xiaohong WANG,Yonghui. LAI. Effect of TINCR⁃MAF： MAFB transcription factor network on proliferation and differentiation of human kerathnocytes [J]. The Journal of Practical Medicine, 2025, 41(4): 509-514.
[6]	Dan ZHU,Wanjun LU,Ying ZHU,Jinlu CAO,Yingzi CHEN. Developing an unsupervised deep learning model for diabetic nephropathy prediction using panoramic fundus retinal images [J]. The Journal of Practical Medicine, 2025, 41(22): 3598-3608.
[7]	Qian ZHANG,Yuntai CAO,Zhijie WANG,Boqi. ZHOU. Advances in deep learning for endoscopic image⁃based diagnosis of early gastric cancer [J]. The Journal of Practical Medicine, 2025, 41(14): 2160-2166.
[8]	Xi HU,Yan LI,Yang. LIU. Recent advances on the application of deep learning in assisted reproductive technology [J]. The Journal of Practical Medicine, 2024, 40(7): 893-897.
[9]	Jia XIANG,Qiantong CHEN,Yingxin LU,Sijie ZHENG,Junjie HUANG,Yingying CHEN,Suidan HUANG,Huai. CHEN. Study on diffuse cystic lung disease based on deep learning [J]. The Journal of Practical Medicine, 2024, 40(19): 2747-2754.
[10]	Xinyu LI,Yang WU,Hongmei ZHANG,Lixue YIN,Bo PENG,Shenghua. XIE. Deep learning technology for quality control of echocardiography images [J]. The Journal of Practical Medicine, 2024, 40(1): 108-113.
[11]	Runcai GUO,Lei WANG,Zhenguo HUANG,Linfeng XI,Shuai ZHANG,Min. LIU. Deep learning convolutional neural network model trained from scratch algorithm in the evaluation of acute pulmonary thromboembolism [J]. The Journal of Practical Medicine, 2023, 39(22): 2979-2983.
[12]	Sitong LI,Ao WANG,Yanping BAI,Ying. WANG. Effects of T follicular helper cells and IL⁃21 on proliferation and cell cycle of HaCaT cellsinpsoriasis [J]. The Journal of Practical Medicine, 2023, 39(18): 2342-2348.
[13]	ZHANG Kaitai, DING Minggang, ZHOU Jingxiang, FANG Chongtao, ZHANG Tingting. . A clinical observation on modified Sijunzi decoction for heart failure with reduced ejection fraction of Qi deficiency and blood stasis syndrome [J]. The Journal of Practical Medicine, 2023, 39(15): 1982-1986.
[14]	WU Weiwei , HAN Junya, ZHANG Yan, HOU Xiaoqiang, FANG Xianfeng.. The role of reflectance confocal microscopy in the efficacy evaluation of secukinumab for psoriasis [J]. The Journal of Practical Medicine, 2022, 38(24): 3072-3076.
[15]	LUO Gang, PAN Si⁃lin, QIAO Sibo, PANG Shanchen, CHEN Taotao, SUN Lingyu, DONG Yukun.. Deep learning technology for automatic recognition of fetal echocardiography images [J]. The Journal of Practical Medicine, 2022, 38(14): 1830-1833.