Deep learning-based recognition of stained tongue coating images

Objective To build a dataset encompassing a large number of stained tongue coating images and process it using deep learning to automatically recognize stained tongue coating images.Methods A total of 1001 images of stained tongue coating from healthy students at Hunan University of Chinese Medicine and 1007 images of pathological(non-stained)tongue coat-ing from hospitalized patients at The First Hospital of Hunan University of Chinese Medicine withlungcancer;diabetes;andhypertensionwerecollected.Thetongueimageswererandomi-zed into the training;validation;and testing datasets in a 7:2:1 ratio.A deep learning model was constructed using the ResNet50 for recognizing stained tongue coating in the training and validation datasets.The training period was 90 epochs.The model’s performance was evaluated by its accuracy;loss curve;recall;F1 score;confusion matrix;receiver operating characteristic(ROC)curve;and precision-recall(PR)curve in the tasks of predicting stained tongue coating images in the testing dataset.The accuracy of the deep learning model was compared with that of attending physicians of traditional Chinese medicine(TCM).Results The training results showed that after 90 epochs;the model presented an excellent classification performance.The loss curve and accuracy were stable;showing no signs of overfitting.The model achieved an accuracy;recall;and F1 score of 92%;91%;and 92%;re-spectively.The confusion matrix revealed an accuracy of 92%for the model and 69%for TCM practitioners.The areas under the ROC and PR curves were 0.97 and 0.95;respectively.Conclusion The deep learning model constructed using ResNet50 can effectively recognize stained coating images with greater accuracy than visual inspection of TCM practitioners.This model has the potential to assist doctors in identifying false tongue coating and prevent-ing misdiagnosis.

Altered oral microbiota in chronic hepatitis B patients with different tongue coatings

AIM To elucidate tongue coating microbiota and metabolic differences in chronic hepatitis B(CHB) patients with yellow or white tongue coatings.METHODS Tongue coating samples were collected from 53 CHBpatients(28 CHB yellow tongue coating patients and 25 CHB white tongue coating patients) and 22 healthy controls.Microbial DNA was extracted from the tongue samples,and the bacterial 16 S ribosomal RNA gene V3 region was amplified from all samples and sequenced with the Ion Torrent PGM^(TM)sequencing platform according to the standard protocols.The metabolites in the tongue coatings were evaluated using a liquid chromatographymass spectrometry(LC-MS) platform.Statistical analyses were then performed.RESULTS The relative compositions of the tongue coating microbiotas and metabolites in the CHB patients were significantly different from those of the healthy controls,but the tongue coating microbiota abundances and diversity levels were not significantly different.Compared with the CHB white tongue coating patients,the CHB yellow tongue coating patients had higher hepatitis B viral DNA(HBV-DNA) titers(median 21210 vs 500,respectively,P = 0.03) and a significantly lower level of Bacteroidetes(20.14% vs 27.93%,respectively,P = 0.013) and higher level of Proteobacteria(25.99% vs 18.17%,respectively,P = 0.045) in the microbial compositions at the phylum level.The inferred metagenomic pathways enriched in the CHB yellow tongue coating patients were mainly those involved in amino acid metabolism,which was consistent with the metabolic disorder.The abundances of bacteria from Bacteroidales at the order level were higher in the CHB white tongue coating patients(19.2% vs 27.22%,respectively,P = 0.011),whereas Neisseriales were enriched in the yellow tongue coating patients(21.85% vs 13.83%,respectively,P = 0.029).At the family level,the abundance of Neisseriaceae in the yellow tongue patients was positively correlated with the HBV-DNA level but negatively correlated with the S-adenosyl-L-methionine level.CONCLUSION This research illustrates specific clinical features and bacterial structures in CHB patients with different tongue coatings,which facilitates understanding of the traditional tongue diagnosis.

Isolation and Characterization of Pathogenic Listonella anguillarum of Diseased Half-Smooth Tongue Sole(Cynoglossus semilaevis Günther)

Studies were conducted to determine the cause of the acute mortality of half-smooth tongue sole Cynoglossus semilaevis Günther juveniles in a fish farm in Jimo, Shandong Province, China, in June 2006. Gross signs of the diseased tongue sole included several petechiae and ecchymoses on the body and fin necrosis and hemorrhagic lesion at the base of the fin. Bacteria were isolated from kidney, liver and hemorrhagic lesions of the diseased tongue sole. Among 14 strains, SJ060621 was proved to be highly virulent to juvenile tongue sole with LD50 value of 〈1.0×10^5 colony forming units （CFU）mL^-1, while the remaining 13 were avirulent. Among the 16 antibiotics tested, SJ060621 was sensitive to gentamicin and nitrofurantoin. It was identified as Listonella anguillantm with conventional plate and tube tests in combination with API 20E analysis. 16S rRNA gene and partial HSP60 gene sequenceing analysis revealed that the strain was highly homologous with L. anguillarum. Examination of the infected musculature by electron microscopy indicated numerous bacteria and lots of macrophages containing phagocytosed bacteria. Histopathological investigations revealed severe necrotic degenerative changes in the infected organs. Indirect immunofluorescence assay （IFA） was employed to detect the location of occurrence of bacteria, and bacteria were found in aggregations in the inflammatory areas in musculature.

Application of Thermal Science to Tongue Inspection in Traditional Chinese Medicine

The study on Traditional Chinese Medicine (TCM)by the technology and method of thermal science is a new cut in point in interdisciplinary science, and its purpose is to study the tongue inspection in TCM from the view of the bio heat transfer theory. In this paper human tongues were studied with the infrared thermal imaging technology. And the characteristic relationship between temperature distribution and age, tongue color as well as infrared thermal tongue image concerning the characteristic of viscera was analyzed by the experimental data,which confirms that the change in temperature distribution in different positions of the tongue is related to the diseases of different internal organs of the body. From a series of invasive experiments in animals, the blood perfusion rate was measured with the equipment of CBI 8000 Doppler Blood Flow Meter and MP 100 Physiologic Research System for the first time and the characteristic relationship curve between the temperature on the surface of the tongue and the blood perfusion rate was obtained. From the experiments the temperature on the surface of the tongue reflects objectively the tongue color which is influenced by the blood perfusion rate. This will help to develop some new cognition in the theory of TCM.

Correlation Between Helicobacter pylori Infection and Tongue Manifestations:A Meta-analysis

Objective To analyze the relationship between Helicobacter pylori(H.pylori)infection and the manifestations of the tongue.Methods Without language restrictions,we searched databases including CAJD,CBMdisc,CDFD,CMFD,PubMed,Cochrane,and EMBASE and made meta-analysis for all literature we have collected.Results Six studies with 975 patients and 50 healthy people in total were included in the analytic pool.There were significant differences in the numbers of H.pylori infections between patients with thin and thick tongue coatings(OR=2.02,95%CI 1.32–3.07,P=0.001).In yellow tongue subgroup,there was a significant difference between patients with thick and thin coatings(OR=2.09,95%CI 1.16–3.77,P=0.01).There was a significant difference in the prevalence of H.pylori infection between patients with yellow and white coatings(OR=2.86,95%CI 2.10–3.90,P<0.001).The frequency of H.pylori infection was significantly increased in patients with red and purple tongues compared with those with pale red tongues(OR=3.42,95%CI 2.40–4.88,P<0.001;OR=7.51,95%CI 3.57–15.79,P<0.001).Conclusion Red or purple tongues and yellow tongues with thick coatings are indicators of a risk of H.pylori infection in patients with GI symptoms.Our study shows that tongue manifestations could serve as a feasible predictor of H.pylori infection in patients with GI symptoms,and determination of the exact association between tongue manifestations and H.pylori infection could enable an understanding of the objectivity of TCM.Long-term and rigorous controlled trials are needed in the future to evaluate the correlating factors.

Study on Tongue Coating Microbiota in Patients with Atrophic Gastritis

Aims: To explore the tongue coating microbiota composition in patients with atrophic gastritis by using the 16S rRNA gene sequencing technology. Methods: The study included 29 atrophic gastritis patients and 29 age and gender-matched non-atrophic gastritis controls. By sequencing the V3-V4 region of the 16S rRNA, we investigated the microbial community structure and diversity on the tongue coating. Results: There was no significant difference in the microbial diversity on the tongue coating between the two groups. However, compared with the control, the atrophic gastritis group had a smaller number of operational taxonomic units (OTUs). At the class level, patients with atrophic gastritis had lower relative abundances of Betaproteobacteria and Spirochaetia than the control group. At the gene level, the abundance of Neisseria and Aggregatibacter in atrophic gastritis group had significantly decreased compared with control ones. Furthermore, functional prediction revealed that 24 metabolic pathways significantly differed between the two groups. Conclusions: Our findings provide novel evidence that tongue coating microbiota may be a biomarker for characterizing patient with atrophic gastritis, but its mechanism needs to be further elaborated.

不同中医证型Wilson病患者肝脏纤维化差异及其与弹性硬度值的关系

目的:探讨不同中医证型Wilson病(WD)患者肝脏硬度值(LSM)及纤维化指标差异及两者的关系。方法:回顾性收集2021年2月至2023年4月住院确诊的110例WD患者的临床资料,中医证型采用指标聚类分析,绘制聚类图。采用简单对应分析中医证型与ADL分级的相关性。比较不同中医证型的一般临床资料、LSM及肝纤维化相关指标,从而探讨LSM与肝纤维化及中医证型分布的关系。结果:WD患者中医证型以湿热内蕴证(59例,53.64%)居多,痰瘀互结证(43例,39.09%)次之,肝肾阴虚证(5例,4.54%)、脾肾阳虚证(3例,2.73%)较少;痰瘀互结证偏向0~Ⅰ级,湿热内蕴证偏向Ⅱ级,肝肾阴虚证偏向Ⅲ~Ⅳ级,脾肾阳虚证在二维投影图中间,并未偏向WD病变程度某一分级;痰瘀互结证、肝肾阴虚证、脾肾阳虚证患者肝脏LSM均高于湿热内蕴证(P<0.05);肝肾阴虚证、脾肾阳虚证患者血清透明质酸(HA)和Ⅳ型胶原(Ⅳ-C)水平高于湿热内蕴证、痰瘀互结证组(P<0.05),4组患者Ⅲ型前胶原氨基端肽(PⅢNP)、层黏连蛋白(LN)水平差异无统计学意义(P>0.05)。湿热内蕴证、痰瘀互结证、肝肾阴虚证中LSM与HA、Ⅳ-C均呈不同程度的正相关。结论:LSM联合血清学指标能更好地评估不同证型WD患者肝纤维化程度。

Does East Meet West?—The Association between Oriental Tongue Inspection and Western Clinical Assays of White Blood Cell Subsets

The autonomic nervous system (ANS) controls white blood cell (WBC) subsets;therefore, the status of ANS can be assessed by assaying WBCs. However, this requires invasive blood sampling, time, cost, and training. Therefore, this study focused on a traditional technique, tongue inspection, which is a simpler method. The purpose of this study was to investigate whether there is an association between the traditional method of tongue inspection and clinical assay of WBC subsets. Twenty-one female alopecia areata patients were divided into two age-matched groups: 1) alopecia areata totalis (AT);and 2) alopecia areata multiplex (AM). Images of patient tongues were captured by a digital camera and categorized before blood sampling. Finally, patients were divided into five groups (normal, Yin+, Yang–, Yin– and Yang+) based on the Eight Principles of traditional Chinese medicine (TCM). Concurrently, venous blood was obtained for WBC subsets. The absolute numbers of WBCs and granulocytes of the AT group were higher than those of the AM group. The AT group was Yin+ but not Yang+, whereas the AM group was Yang+ but not Yin+. Thus, the AT group showed more elements of “cold” (Yin > Yang) compared with the AM group with elements of “hot” (Yin < Yang). Tongue inspection suggested a possibility of consistence with those of WBCs although statistical significance was not obtained. Moreover, some Yin+ and Yang+ subjects showed some trend in similarities between tongue inspection and WBC subsets although this was not statistically significant. Therefore, traditional techniques (such as tongue inspection) acupuncture must be studied further to detect whether subtle effects are induced by acupuncture treatment. As this study is underpowered, a larger scale study including males is required in the future.

Arnold Tongues for Discrete Hill’s Equation

In this work we study two types of Discrete Hill’s equation. The first comes from the discretization process of a Continuous-time Hill’s equation, we called Discretized Hill’s equation. The Second is a naturally obtained in Discrete-Time and will be called Discrete-time Hill’s equation. The objective of discretization is preserving the continuous-time behavior and we show this property. On the contrary a completely different dynamic property was found for the Discrete-Time Hill’s equation. At the end of the paper is shown that both types share the nonoscillatory behavior of solutions in the 0-th Arnold Tongue.

TCM Treatment of Asthma in Children with Yin Hua Wu Mei Tang——Professor Liu Bichen's Experience in Treating Asthma

Asthma is characterized by paroxysmaldyspnea,rales in the throat,difficulty in breathing,and inability to lie horizontally.It is commonly seenin children with a weak body constitution,and tendsto be intractable,thus being harmful to the

基于双阶段元学习的小样本中医舌色域自适应分类方法

舌色是中医(TCM)望诊最关注的诊察特征之一。在实际应用中,通过一台设备采集到的舌象数据训练得到的舌色分类模型应用于另一台设备时,由于舌象数据分布特性不一致,分类性能往往急剧下降。为此,该文提出一种基于双阶段元学习的小样本中医舌色域自适应分类方法。首先,设计了一种双阶段元学习训练策略,从源域有标注样本中提取域不变特征,并利用目标域的少量有标注数据对网络模型进行微调,使得模型可以快速适应目标域的新样本特性,提高舌色分类模型的泛化能力并克服过拟合。接下来,提出了一种渐进高质量伪标签生成方法,利用训练好的模型对目标域的未标注样本进行预测,从中挑选出置信度高的预测结果作为伪标签,逐步生成高质量的伪标签。最后,利用这些高质量的伪标签,结合目标域的有标注数据对模型进行训练,得到舌色分类模型。考虑到伪标签中含有噪声问题,采用了对比正则化函数,可以有效抑制噪声样本在训练过程中产生的负面影响,提升目标域舌色分类准确率。在两个自建中医舌色分类数据集上的实验结果表明,在目标域仅提供20张有标注样本的情况下,舌色分类准确率达到了91.3%,与目标域有监督的分类性能仅差2.05%。

Contradiction between Traditional Chinese Medicine and Modern Medicine in Understanding and Treating Parkinson’s Disease and Its Solutions

Parkinson’s disease is not a lifelong disease. Although modern medicine cannot cure Parkinson’s disease, I found that traditional Chinese medicine can cure Parkinson’s disease. However, because of the misunderstanding of the treatment ideas of most of Chinese doctors, they failed to achieve the goal of curing Parkinson’s disease. There are differences not only in understanding and treating Parkinson’s disease between traditional Chinese medicine and modern medicine, but also in understanding within the Chinese medicine community. Because they have not been able to cure Parkinson’s disease, they draw the conclusions from modern medicine and also believe that Parkinson’s disease is an incurable lifelong disease. This paper deeply explores the problems existing in modern medicine and traditional Chinese medicine, puts forward new ideas for understanding and treating Parkinson’s disease, and explores the correct way to understand and cure Parkinson’s disease for the medical community to overcome Parkinson’s disease.

常见中医证型Wilson病患者情绪智力水平与病变脑区相关性

目的 探讨常见中医证型的Wilson病(Wilson’s disease, WD)患者情绪智力水平、维度特征及其与病变脑区的相关性。方法 采用情绪智力水平量表(emotional intelligence scale, EIS)对103例肝胆湿热证、痰湿阻络证及肝肾阴虚证WD患者以及47例健康者进行测评,并对WD患者进行颅脑MRI检查。结果 与正常组比较,痰湿阻络证组EIS总分及表达能力、自我管理及管理他人3项因子分均减少(P<0.05),肝肾阴虚证组EIS总分及全部4项因子分均减少(P<0.05),肝胆湿热证组EIS总分和各维度因子分差异无统计学意义(P>0.05);与肝胆湿热证组比较,痰湿阻络证组EIS总分及表达能力因子分减少(P<0.05),肝肾阴虚证组EIS总分、情绪感觉能力及自我管理因子分减少(P<0.05);与痰湿阻络证组比较,肝肾阴虚证组EIS总分和各维度因子分差异无统计学意义(P>0.05)。肝胆湿热证WD患者以基底节、脑干、丘脑损伤为主,痰湿阻络证WD患者以基底节、脑干、丘脑损伤为主,肝肾阴虚证WD患者以基底节损伤、脑干损伤、脑萎缩为主。额叶皮质损伤与WD患者EIS总分及表达能力因子分呈正相关(P<0.05);苍白球损伤与WD患者EIS总分及管理他人因子分呈负相关(P<0.05)。结论 痰湿阻络证及肝肾阴虚证WD患者情绪智力水平较低,而肝胆湿热证WD患者情绪智力正常;3种常见中医证型WD患者中,以基底节及脑干损伤较为常见;额叶皮质及苍白球损伤是影响WD患者情绪智力水平的因素。

我国三级中医医院门诊患者中医药特色服务利用情况调查

目的:调查我国三级中医医院门诊患者中医药特色服务的利用情况,为中医医院的特色化发展提供依据。方法:对全国37家中医医院门诊患者中药饮片和中医非药物疗法的利用情况进行分析。合并中药饮片和中医非药物疗法利用为中医药特色服务利用情况,并采用卡方检验进行不同分层间的差异性分析。结果:共收集7,542例门诊患者,55.81%的患者开具了中药饮片处方,27.95%的患者采用了中医非药物疗法,31.77%的患者在诊疗过程中未采用中医药特色方法治疗。不同地理位置、就诊科室、医生职称患者中医药特色服务利用存在显著性差异(χ^(2)=26.459、393.670、54.343,P<0.01),而不同患者社会人口学特征间差异无统计学意义。结论:我国三级中医医院门诊患者中医药特色服务利用情况良好,但不同科室对中医药技术方法的利用各有侧重。中西部地区中医医院门诊患者中医药特色服务利用情况优于东部地区。职称越高的医生提供中医药特色服务的比例也越高。

63例SARS中医症候群的动态分析

目的 :观察严重急性呼吸综合征 (SARS)患者的症候及舌象动态特征及与实验室检查的关系。方法 :观察 2 0 0 3年 3月 11日— 4月 30日在我院感染科住院 6 3例SARS患者的症候、舌象及实验室指标的动态变化。结果 :(1)早期最常见的症状为发热 (5 6例 )、咳嗽 (4 1例 )、四肢乏力或困重 (4 0例 ) ;中期 (极期 )最常见的症状为咳嗽 (5 3例 )、四肢乏力或困重 (4 3例 ) ;恢复期患者临床症状多不典型 ,偶可见四肢乏力或困重(17例 )、胸闷胸痛 (13例 )、低热 (11例 )、咳嗽 (12例 )、咳痰 (11例 )。 (2 )舌象 :早期多舌淡红或舌尖红 ,舌苔薄白、白腻或黄腻 ;中期多舌质红 ,舌苔白腻或黄厚腻 ;恢复期多舌质淡红或舌尖红 ,偶有舌质暗紫 ,苔薄白或白腻 ,偶有苔黄腻。 (3)SARS患者早期白细胞、血小板及CD4 + 、CD8+ T淋巴细胞均低于正常 ,至恢复期均显著升高 ,与早期比较 ,差异均有显著性 (P <0 0 1)。结论 :SARS不同病期的中医症状有其特点和变化规律 ,与实验室检查结果有一致性。

210例HIV感染者/AIDS患者中医四诊信息分布特点

目的研究HIV感染者/AIDS患者中医症状体征的分布特点。方法以某HIV感染者/AIDS患者门诊病例资料为数据来源,进行HIV感染者/AIDS患者常见临床流行病学信息回顾性分析。结果基于"三期十二型"进行HIV感染者/AIDS患者临床证型诊断分类结果 :气虚血瘀,邪毒壅滞型占27.14%;气血两亏型占15.71%;气阴两虚,肺阴不足型占9.52%;热毒内蕴,痰热壅肺型占6.19%。HIV感染者/AIDS患者常见临床症状体征包括乏力(64.29%)、头痛(50.00%)、腰痛(49.05%)、气短(42.38%)、咳嗽(41.43%)、皮肤瘙痒(38.57%)、恶心(36.19%)、腹泻(30.48%)、发热(28.10%)、皮疹(27.62%)等;舌象以红舌(40.95%)为主,舌苔以白苔(36.19%)为多见;脉象以细脉(71.90%)、沉脉(25.71%)和弦脉(27.62%)多见。结论 HIV感染者/AIDS患者中医四诊信息分布有一定规律,症状体征主要集中在肺脾肾三脏。

92例SARS患者舌象的分析

为了解严重急性呼吸综合征(severeacuterespiratorysyndrome,SARS)患者的中医证候特征、演变规律,舌象特点与X线胸片变化之间的关系,探索舌象在防治SARS中的作用,采用前瞻性方法对入院的92例SARS患者的临床表现、舌象照片、X线胸片等临床资料归纳总结和统计分析。结果患者的舌象出现芒剌与X线胸片的进展期和高峰期呈正相关,且与高峰期有显著性相关(P<0.05);舌体胖大、齿痕与痊愈期呈正相关,具有非常显著性相关(P<0.01);少苔或剥脱苔与吸收期、痊愈期和病程呈正相关,其中与吸收期和病程具有显著性相关(P<0.05)。提示虽然SARS患者的临床表现、舌象、X线胸片呈多样化,但放射学仍有独特改变,且舌象与X线胸片之间具有一定的相关性,舌象的动态追踪有助于及早发现病情变化,中医舌诊在SARS的诊疗中具有重要意义。

基于噪声样本渐近修正的中医舌色分类方法

基于深度学习的中医舌色分类模型具备良好的性能,但是依赖大量正确标注的样本.由于人工标注样本费时费力,不可避免地存在错误标注,导致模型在训练过程中对噪声样本过拟合,使其泛化能力变差.为此,本文提出了一种基于噪声样本渐近修正的中医舌色分类方法.首先,根据舌色分类的特点,提出了一种全局-局部特征融合方法,将其嵌入到ResNet18骨干网络中,构建了舌色分类网络,并采用集成学习范式,提高分类模型的可靠性和稳定性;其次,针对噪声样本下的舌色分类网络训练问题,提出了样本注意力机制和噪声样本标签重新标注机制,在训练过程中对干净样本和噪声样本加以区分,赋予不同的权重,并逐步对噪声样本标签进行修正;最后,采用Boostrapping损失函数降低模型对噪声样本的关注度,抑制噪声样本对分类性能的影响.将提出的方法在两个自建的舌色分类数据集上进行了实验验证,结果表明,该方法通过渐进地对噪声标签进行校正,可以获得比现有的有噪样本下图像分类方法更高的分类精度,Acc指标分别达到了94.6%和93.65%.

基于持续学习的中医舌色苔色协同分类方法

中医(traditional Chinese medicine, TCM)舌诊客观化研究中需要分析的舌象特征很多,不同的舌象特征往往采用单独的方法进行分析,导致分析系统的整体实现复杂度大幅增加。为此,基于持续学习的思想,提出一种中医舌色苔色协同分类方法,该方法将舌色分类作为旧任务,将苔色分类作为新任务,充分利用2个任务的相似性和相关性,仅通过一个网络结构就同时实现舌色和苔色的准确分类。首先,设计一种基于全局-局部混合注意力机制(global local hybrid attention, GLHA)的双分支网络结构,将网络高层语义特征与低层特征相融合,提升特征的表达能力;然后,提出基于正则化和回放相结合的持续学习策略,使得该网络在学习新任务知识的同时有效防止对旧任务知识的遗忘。在2个自建的中医舌象特征分析数据集上的实验结果表明,提出的协同分类方法可以获得与单个任务相当的分类性能,同时可以将2个分类任务的整体复杂度降低一半左右。其中,舌色分类准确率分别达到93.92%和92.97%,精确率分别达到93.69%和92.87%,召回率分别达到93.96%和93.16%;苔色分类准确率分别达到90.17%和90.26%,精确率分别达到90.05%和90.17%,召回率分别达到90.24%和90.29%。

基于双模型互学习的半监督中医舌诊图像分割方法

舌体图像的准确分割是中医舌诊客观化分析的重要前提,目前广泛采用的全监督分割方法需要对大量像素级标注的样本进行训练,而基于单一模型的半监督分割方法对学习到的错误知识缺乏自我纠正能力。针对这一问题,本文提出一种新颖的基于双模型互学习的半监督舌体图像分割方法。首先,模型A、B分别在有标签数据集上进行监督训练。之后,模型A、B进入互学习阶段,使用本文设计的互学习损失函数,根据双模型对无标签数据预测的分歧而赋予不同的权重。先由模型A对无标签数据集生成伪标签,模型B在有标签数据集和伪标签数据集上进行微调,然后模型B再对无标签数据集生成伪标签,模型A以同样的方式进行微调。双模型微调结束后,选择性能更佳的一个作为最终的舌体图像分割模型。实验结果表明,本文方法的有标签数据比例为1/100、1/50、1/25、1/8时,mIoU分别达到96.67%、97.92%、98.52%、98.85%,优于对比的其他典型半监督方法。本文方法仅需少数标签数据便可达到较高的舌体图像分割精度,可为后续的中医舌色、舌形等舌象分析应用奠定坚实基础,推动中医诊疗数字化进程。