1.Research status and prospect of tongue image diagnosis analysis based on machine learning
Jiatuo XU ; Tao JIANG ; Shi LIU
Digital Chinese Medicine 2024;7(1):3-12
Image-based intelligent diagnosis represents a trending research direction in the field of tongue diagnosis in traditional Chinese medicine(TCM).In recent years,machine learning techniques,including convolutional neural networks(CNNs)and Transformers,have been widely used in the analysis of medical images,such as computed tomography(CT)and nucle-ar magnetic resonance imaging(MRI).These techniques have significantly enhanced the effi-ciency and accuracy of decision-making in TCM practices.Advanced artificial intelligence(AI)technologies have also provided new opportunities for the research and development of medical equipment and TCM tongue diagnosis,resulting in improved standardization and in-telligence of the tongue diagnostic procedures.Although traditional image analysis methods could transform tongue images into scientific and analyzable data,recognizing and analyzing images that capture complicated tongue features such as tooth-marked tongue,tongue spots and prickles,fissured tongue,variations in coating thickness,tongue texture(curdy and greasy),and tongue presence(peeled coating)continues posing significant challenges in con-temporary tongue diagnosis research.Therefore,the employment of machine learning tech-niques in the analysis of tongue shape and texture features as well as their applications in TCM diagnosis is the focus of this study.In the study,both traditional and deep learning im-age analysis techniques were summarized and analyzed to figure out their value in predicting disease risks by observing the tongue shapes and textures,aiming to open a new chapter for the development and application of AI in TCM tongue diagnosis research.In short,the com-bination of TCM tongue diagnosis and AI technologies,will not only enhance the scientific basis of tongue diagnosis but also improve its clinical applicability,thereby advancing the modernization of TCM diagnostic and therapeutic practices.
2.Distribution of traditional Chinese medicine pattern types and prognostic risk factors in patients undergoing percutaneous coronary intervention(PCI):a systematic review and meta-analysis
Jieyun LI ; Leixin HONG ; Jiekee LIN ; Yumo XIA ; Xin'ang XIAO ; Zhaoxia XU
Digital Chinese Medicine 2024;7(1):13-28
Objective To clarify the distribution of traditional Chinese medicine(TCM)pattern and its associated risk factors after percutaneous coronary intervention(PCI),and evaluate the re-porting quality of existing studies to guide future research standardization. Methods English databases including PubMed,Cochrane Library,and Web of Science,as well as Chinese databases including China National Knowledge Infrastructure(CNKI),China Scientific Journal Database(VIP),and Wanfang Database were searched to retrieve papers about PCI.The time span for the paper retrieval was set from the foundation of the databases to October 1,2023.Statistical analyses were performed using Stata 12 and Python(V 3.9).The Strengthening the Reporting of Observational Studies in Epidemiology(STROBE)statement was used to assess the reporting quality of included studies. Results Overall,1 356 articles were selected,and 40 cross-sectional studies were included with 10 270 participants.The most common TCM patterns before,one to two weeks after,and six months to one year after PCI was Qi stagnation and blood stasis(n=261,36.45%),inter-twined phlegm and blood stasis(n=109,27.18%),and Qi deficiency and blood stasis(n=645,37.03%),respectively.Smoking[odds ratio(OR)=1.15,95%confidence interval(CI)(0.83-1.47),I2=24.7%,P=0.257],pattern of congealing cold and Qi stagnation[OR=4.62,95%CI(1.37-7.86),I2=61.6%,P=0.074],and low-density lipoprotein(LDL)[OR=1.38,95%CI(0.92-1.85),I2=12.2%,P=0.286]were risk factors for restenosis.Hypertension[OR=7.26,95%CI(3.54-14.88),I2=91.6%,P=0.001],and overweight[i.e.,body mass index(BMI)>23][OR=1.20,95%CI(1.07-1.35),I2=85.3%,P=0.009]were significant risk factors of concomi-tant anxiety. Conclusion This systematic review and meta-analysis revealed that patients with different TCM pattern types have distinct characteristics and risk factors after PCI.More high-quality studies are warranted to provide supportive evidence for future research and clinical practice.
3.Differences in pulse manifestations at Cunkou based on simplified modeling of tactile sensing
Yubing SHI ; Hongyi YANG ; Hock Joon YEO
Digital Chinese Medicine 2024;7(1):29-39
Objective In the theories of pulse disgnosis in traditional Chinese medicine(TCM),it is em-phasized that pulse manifestations at the radial artery within the wrist(called Cunkou)signi-fy the physiological and pathological conditions of different internal organs in the human body.However,different opinions exist among researchers about the objectiveness of the pulse diagnosis technique.Some researchers mentioned that no significant differences were observed in pulse manifestations at various Cunkou areas,hence there might be some diffi-culty in evaluating the status of different organs through checking pulse manifestations at Cunkou.This research aims to analyze the pulse response at Cunkou from the aspect of the characteristics of tactile sensing,thus to give a preliminary explanation to the above question. Methods This research utilized the Weber-Fechner law to model the tactile sensing as a dy-namic low-pass signal filter with varying bandwidths under different compression levels dur-ing pulse diagnosis.The model was applied to analyzing the clinical data collected previously by our group.The arterial pressures measured invasively with equipment from 14 patients with aorta coarctation were processed to simulate different pulse manifestations at Cun,Guan,and Chi positions of Cunkou when different compression levels were applied. Results Due to the characteristics of tactile sensing,significant variations were observed in pulse manifestations at different pulse-depths under the application of changing compres-sion levels;while no such changes in pulse manifestations were observed from the employ-ment of transducer only,without tactile sensing involved.The results explained why different understandings on pulse manifestations were formed between direct pulse-taking technique in TCM and modern sphygmography using transducers.The features of pulse manifestations at Cunkou,using direct pulse-taking technique but at different depths,superficial,middle,and deep positions,respectively,predicted by the developed tactile sensing model were in line with those described in TCM pulse theories. Conclusion Based on the developed tactile sensing model,this study preliminarily explains the phenomenon that pulse manifestation at Cunkou changes in response to the compression force applied during TCM pulse-taking.Integrating the tactile sensing model with the study of TCM pulse diagnosis opens a new chapter for visualizing and quantitatively interpreting pulse manifestations.This not only expands the scope of pulse diagnosis study effectively,but also provide a scientific basis for further technical upgrading and optimization of existing pulse di-agnosis equipment.
4.Intelligent question answering system for traditional Chinese medicine based on BSG deep learning model:taking prescription and Chinese materia medica as examples
Ran LI ; Gao REN ; Junfeng YAN ; Beiji ZOU ; Qingping LIU
Digital Chinese Medicine 2024;7(1):47-55
Objective To construct a traditional Chinese medicine(TCM)knowledge base using knowl-edge graph based on deep learning methods,and to explore the application of joint models in intelligent question answering systems for TCM. Methods Textbooks Prescriptions of Chinese Materia Medica and Chinese Materia Medicawere applied to construct a comprehensive knowledge graph serving as the founda-tion for the intelligent question answering system.In the study,a BERT+Slot-Gated(BSG)deep learning model was applied for the identification of TCM entities and question inten-tions presented by users in their questions.Answers retrieved from the knowledge graph based on the identified entities and intentions were then returned to the user.The Flask framework and BSG model were utilized to develop the intelligent question answering sys-tem of TCM. Results A TCM knowledge map encompassing 3 149 entities and 6 891 relational triples based on the prescriptions and Chinese materia medica was drawn.In the question answer-ing test assisted by a question corpus,the F1 value for recognizing entities when answering 20 types of TCM questions was 0.996 9,and the accuracy rate for identifying intentions was 99.75%.This indicates that the system is both feasible and practical.Users can interact with the system through the WeChat Official Account platform. Conclusion The BSG model proposed in this paper achieved good results in experiments by increasing the vector dimension,indicating the effectiveness of the joint model method and providing new research ideas for the implementation of intelligent question answering sys-tems in TCM.
5.Influencing factors analysis of dynamic change of TCM constitution based on multiple methods
Yue LUO ; Jianfu LU ; Yunsong ZHENG ; Lei BAO ; Chuanbiao WEN
Digital Chinese Medicine 2024;7(1):56-67
Objective This study aimed to explore the influencing factors of dynamic changes in tradi-tional Chinese medicine(TCM)constitution based on general statistics,Apriori-DEMATEL algorithm,and DoWhy causal inference framework methods. Methods Dynamic collection of TCM constitution identification data was conducted from the population aged 18-60,containing collection time and constitution type,and 11 constitu-tion influencing factors including dietary habit,sleeping habit,sleeping duration,exercise habit,emotion state,stress level,living environment,work/life calamity,family atmosphere,business trip frequency,and overtime situation.General statistical analysis was used to ana-lyze the relative percentage of corresponding influencing factors of different types of constitu-tion changes,the Apriori-DEMATEL algorithm was used to analyze the correlation between 11 constitution influencing factors such as dietary habit and constitution changes,and the DoWhy causal inference framework was used to analyze the causality between dietary habit,sleeping habit,sleeping duration,exercise habit,emotion state,and stress level,explore the frequency of constitution type transformation-change factors,and determine the key influ-encing factors causing dynamic changes in constitution type. Results After preprocessing,13 536 valid data points were obtained.Based on the Apriori-DEMATEL algorithm,the factors were divided into six original factors including dietary habit,sleeping habit,sleeping duration,exercise habit,emotion state,and stress level,and five re-sult factors including living environment,work/life calamity,family atmosphere,business trip frequency,and overtime situation.Combining with general statistics,we found that among the original factors,changes in dietary habit,sleeping habit,sleeping duration,and stress lev-el had a greater impact on other factors.In the process of constitution conditioning,attention should be paid to these four factors to maintain constitution balance.Among the five result factors,the absolute values of work/life calamity and family atmosphere were relatively large,indicating that these two factors were easily influenced by other factors.The dietary habit,sleeping habit,sleeping duration,exercise habit,emotion state,and stress level have higher centrality in changes,indicating that these six factors had important in constitution changes.According to the statistical frequency of constitution changes corresponding to each factor,we found that the changes of these six factors accounted for a large proportion of the constitu-tion transformation frequency among Qi deficiency constitution,balanced constitution,and allergic constitution,indicating that the changes of these six factors played an important role in the changes of the three constitution types.Combined with the results of the Apriori-DEMATEL algorithm,and DoWhy causal inference framework analysis,it was inferred that dietary habit and sleeping duration indirectly lead to constitution changes by affecting the changes of other factors. Conclusion This study explored the influencing factors of dynamic changes in TCM constitu-tion from the perspective of dynamic data and multiple analysis methods,and the results showed that the changes of dietary habit,sleeping habit,sleeping duration,exercise habit,emotion state,and stress level had a great impact on the changes of Qi deficiency constitu-tion,balanced constitution and allergic constitution.Attention should be paid to the changes of these six factors in daily life,and corresponding improvement plans should be formulated to reduce the probability of transforming into biased constitution.Our study also provided data support and objective analysis reference for the analysis of influencing factors of dynam-ic changes in TCM constitution types.
6.Correlation between tongue and pulse indicators and the outcome of live birth in frozen-thawed embryo transfer
Jinluan WANG ; Zhiling GUO ; Qinhua ZHANG ; Hua YAN ; Liping TU ; Jiatuo XU
Digital Chinese Medicine 2024;7(1):68-78
Objective To investigate the correlation between tongue and pulse indicators and the out-come of live birth in patients undergoing frozen-thawed embryo transfer(FET),as well as the association between these indicators and patients'endocrine parameters. Methods This study was conducted at Reproductive Medicine Center,Shuguang Hospital Af-filiated to Shanghai University of Traditional Chinese Medicine,Shanghai,China,from March 8,2021 to January 5,2022.Patients undergoing FET were divided into live birth and non-live birth groups according to their live birth outcome.The differences between the endocrine pa-rameters[basic follicle stimulating hormone(b FSH),basic luteinizing hormone(b LH),basic estradiol(b E2),basic progesterone(b P),basal endometrial thickness,follicle stimulating hormone(FSH)on endometrial transition day,luteinizing hormone(LH)on endometrial transition day,estradiol(E2)on endometrial transition day,progesterone(P)on endometrial transition day,and endometrial thickness on endometrial transition day]and the tongue and pulse indicators[tongue body(TB)-L,TB-a,TB-b,tongue coating(TC)-L,TC-a,TC-b,perAll,perPart,h1,h4,h5,t1,h1/t1,and h4/h1]of patients in the two groups were analyzed,with the correlation between these variables analyzed as well using Spearman's correlation coefficient.Multivariate logistic regression was employed to identify the influential factors in the live birth prediction models across various datasets,including Model 1 consisting of endocrine indica-tors only,Model 2 solely consisting of tongue and pulse indicators,and Model 3 consisting of both tongue,pulse,and endocrine indicators,as well as to evaluate efficacy of the models de-rived from different datasets. Results This study included 78 patients in live birth group and 144 patients in non-live birth group.Compared with non-live birth group,live birth group exhibited higher levels of TB-L(P=0.01)and TB-a(P=0.04),while demonstrated lower levels of b FSH(P=0.01),perAll(P=0.04),and h4/h1(P=0.03).The Spearman's correlation coefficient analysis revealed statisti-cally significant correlation(P<0.05)between TB-L,TB-b,TC-L,TC-b,perAll,perPart,h4,h5,t1,h1/t1 and b FSH,b LH,basal endometrial thickness,LH on endometrial transition day,E2 on endometrial transition day,P on endometrial transition day,and endometrial thickness on endometrial transition day in live birth group.The multivariate logistic regression analysis showed that the prediction Model 3 for live birth outcome[area under the curve(AUC):0.917,95%confidence interval(CI):0.863-0.971,P<0.001]surpassed the Model 1(AUC:0.698,95%CI:0.593-0.803,P=0.001),or the Model 2(AUC:0.790,95%CI:0.699-0.880,P<0.001).The regression equations for the live birth outcomes,integrating tongue and pulse indicators with endocrine parameters,included the following measures:FSH on endometrial transition day[odds ratio(OR):0.523,P=0.025],LH on endometrial transition day(OR:1.277,P=0.029),TB-L(OR:2.401,P=0.001),perPart(OR:1.018,P=0.013),h1(OR:0.065,P=0.021),t1(OR:4.354,P=0.024),and h4/h1(OR:0.018,P=0.016). Conclusion In infertility patients undergoing FET,there exists a correlation between tongue and pulse indicators and endocrine parameters.The corporation of tongue and pulse indica-tors significantly improved the predictive capability of the model for live birth outcomes.Specifically,tongue and pulse indicators such as TB-L,perPart,h1,t1,and h4/h1 exhibited a discernible correlation with the ultimate live birth outcomes.
7.Mechanism of Wenyang Shengji Ointment in treating diabetic wounds based on network pharmacology and animal experiments
Yarong DING ; Chenlei XIE ; Shuihua FENG ; Zhonghang YUAN ; Wei WANG ; Mulin LIU ; Zhongzhi ZHOU ; Li CHEN
Digital Chinese Medicine 2024;7(1):79-89
Objective To explore the mechanism of Wenyang Shengji Ointment(温阳生肌膏,WYSJO)in the treatment of diabetic wounds from the perspective of network pharmacology,and to veri-fy it by animal experiments. Methods The Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform(TCMSP)and related literature were used to screen active compounds in WYSJO and their corresponding targets.GeneCards,Online Mendelian Inheritance in Man(OMIM),DrugBank,PharmGkb,and Therapeutic Target Database(TTD)databases were employed to identify the targets associated with diabetic wounds.Cytoscape 3.9.0 was used to map the ac-tive ingredients in WYSJO,which was the diabetic wound target network.Search Tool for the Retrieval of Interaction Gene/Proteins(STRING)platform was utilized to construct protein-protein interaction(PPI)network.Kyoto Encyclopedia of Genes and Genomes(KEGG)and Gene Ontology(GO)enrichment analyses were performed to identify signaling pathways be-tween WYSJO and diabetic wounds.AutoDock 1.5.6 was used for molecular docking of core components in WYSJO to their targets.Eighteen rats were randomly divided into control,model,and WYSJO groups(n=6).The model and WYSJO groups were used to prepare the model of refractory wounds in diabetes rats.The wound healing was observed on day 0,5,9,and 14 after treatment,and the wound tissue morphology was observed by hematoxylin-eosin(HE)staining.The expression levels of core genes were detected by quantitative real-time polymerase chain reaction(qPCR). Results A total of 76 active compounds in WYSJO,206 WYSJO drug targets,3 797 diabetic wound targets,and 167 diabetic wound associated WYSJO targets were screened out through network pharmacology.With the use of WYSJO-diabetic wound target network,core targets of seven active compounds encompassing quercetin,daidzein,kaempferol,rhamnetin,rham-nocitrin,strictosamide,and diisobutyl phthalate(DIBP)in WYSJO were found.GO enrich-ment analysis showed that the treatment of diabetes wounds with WYSJO may involve lipopolysaccharide,bacteria-derived molecules,metal ions,foreign stimuli,chemical stress,nutrient level,hypoxia,and oxidative stress in the biological processes.KEGG enrichment analysis showed that the treatment of diabetes wounds with WYSJO may involve advanced glycation end products(AGE-RAGE),p53,interleukin(IL)-17,tumor necrosis factor(TNF),hypoxia inducible factor-1(HIF-1),apoptosis,lipid,atherosclerosis,etc.The results of animal experiments showed that WYSJO could significantly accelerate the healing process of diabetic wounds(P<0.05),alleviate inflammatory response,promote the growth of granulation tis-sues,and down-regulate the expression levels of eight core genes[histone crotonyltrans-ferase p300(EP300),protoc gene-oncogene c-Jun(JUN),myelocytomatosis(MYC),hypoxia inducible factor 1A(HIF1A),mitogen-activated protein kinase 14(MAPK14),specificity pro-tein 1(SP1),tumor protein p53(TP53),and estrogen receptor 1(ESR1)]predicted by the net-work pharmacology(P<0.05). Conclusion The mechanism of WYSJO in treating diabetes wounds may be closely related to AGE-RAGE,p53,HIF-1,and other pathways.This study can provide new ideas for the phar-macological research of WYSJO,and provide a basis for its further transformation and appli-cation.
8.Effects of Tuina targeting different body parts on the behaviors and gut microflora of autistic spectrum disorder rat models
Tao LI ; Xiang FENG ; Hui ZHI ; Wentao HUANG ; Jiangshan LI ; Wu LI
Digital Chinese Medicine 2024;7(1):90-100
Objective To investigate the effects of Tuina targeting different body parts on the behaviors and gut microflora of rat models with valproic acid(VPA)-induced autistic spectrum disorder(ASD). Methods Twenty female Sprague Dawley(SD)rats with 12.5 d of pregnancy were randomly divided into VPA model group[intraperitoneal injection of VPA(600 mg/kg),n=15]and saline group(intraperitoneal injection of equal amount of normal saline,n=5).The offspring male rats injected with saline were secleted as control group.The offspring male rats injected with VPA were randomly divided into VPA,dorsal,and abdominal groups(n=7 in each group).On the 21st day after birth,three-chamber social test,open field test,and marble-burying test were carried out to observe the social abilities,anxiety behaviors,and stereotypi-cal behaviors of rats in the four groups.Rats in dorsal and abdominal groups underwent Tuina for 14 d,twice a day.On the 35th day,behavioral tests were conducted again,and in-testinal contents were taken for species composition and structural analysis,as well as mark-er and differential species analysis. Results(i)According to behavioral observations,compared with VPA group,the social and movement time in the central open field of rats in dorsal group increased significantly(P<0.05),and the number of buried marbles decreased markedly(P<0.01),indicating improve-ment on their social abilities,anxiety behaviors,and stereotypical behaviors as consequences of dorsal Tuina;and the number of buried marbles was reduced as well in abdominal group when compared with VPA group(P<0.05),suggesting the improvement on their stereotypi-cal behaviors following abdominal Tuina.In the marble-burying test,the number of marbles buried in dorsal group was less than in abdominal group,and the stereotypical behaviors were improved more significantly(P<0.05),and there were no significant differences in the three-chamber social and open field tests between the two groups(P>0.05).(ii)In accor-dance with intestinal microflora detection results,compared with VPA group,both dorsal and abdominal groups showed increased richness(P<0.05)and elevated diversity(P<0.05 in dorsal group and P<0.01 in abdominal group)in intestinal microflora.The results of differen-tial analysis indicated that at the phylum level,compared with VPA group,the relative abun-dance of Firmicutes in rats in abdominal group showed a significant reduction trend(P<0.05);at the genus level,compared with VPA group,the relative abundance of Lactobacillus in rats in dorsal and abdominal groups decreased significantly(P<0.05).Dorsal group also showed significant increase in the genus Blautia in the analysis of marker species compared with VPA group(P<0.05). Conclusion Tuina impacted the behavior and gut microflora structure of ASD model rats.Dorsal intervention had a significant effect on social abilities,anxiety behaviors,and stereo-typical behaviors of ASD model rats,while abdominal intervention only had an obvious effect on stereotypical behaviors.Both dorsal and abdominal interventions increased the richness and diversity of gut microflora of ASD model rats,with abdominal intervention improving the intestinal microbial diversity more significantly and resulting in a more uniform species dis-tribution.
9.The enlightenment of artificial intelligence large-scale model on the research of intelligent eye diagnosis in traditional Chinese medicine
Yuan GAO ; Zixuan WU ; Boyang SHENG ; Fu ZHANG ; Yong CHENG ; Junfeng YAN ; Qinghua PENG
Digital Chinese Medicine 2024;7(2):101-107
Eye diagnosis is a method for inspecting systemic diseases and syndromes by observing the eyes.With the development of intelligent diagnosis in traditional Chinese medicine(TCM),artificial intelligence(AI)can improve the accuracy and efficiency of eye diagnosis.However,the research on intelligent eye diagnosis still faces many challenges,including the lack of standardized and precisely labeled data,multi-modal information analysis,and artificial in-telligence models for syndrome differentiation.The widespread application of AI models in medicine provides new insights and opportunities for the research of eye diagnosis intelli-gence.This study elaborates on the three key technologies of AI models in the intelligent ap-plication of TCM eye diagnosis,and explores the implications for the research of eye diagno-sis intelligence.First,a database concerning eye diagnosis was established based on self-su-pervised learning so as to solve the issues related to the lack of standardized and precisely la-beled data.Next,the cross-modal understanding and generation of deep neural network models to address the problem of lacking multi-modal information analysis.Last,the build-ing of data-driven models for eye diagnosis to tackle the issue of the absence of syndrome dif-ferentiation models.In summary,research on intelligent eye diagnosis has great potential to be applied the surge of AI model applications.
10.Constitution identification model in traditional Chinese medicine based on multiple features
Anying XU ; Tianshu WANG ; Tao YANG ; Xiao HAN ; Xiaoyu ZHANG ; Ziyan WANG ; Qi ZHANG ; Xiao LI ; Hongcai SHANG ; Kongfa HU
Digital Chinese Medicine 2024;7(2):108-119
Objective To construct a precise model for identifying traditional Chinese medicine(TCM)constitutions,thereby offering optimized guidance for clinical diagnosis and treatment plan-ning,and ultimately enhancing medical efficiency and treatment outcomes. Methods First,TCM full-body inspection data acquisition equipment was employed to col-lect full-body standing images of healthy people,from which the constitutions were labelled and defined in accordance with the Constitution in Chinese Medicine Questionnaire(CCMQ),and a dataset encompassing labelled constitutions was constructed.Second,heat-suppres-sion valve(HSV)color space and improved local binary patterns(LBP)algorithm were lever-aged for the extraction of features such as facial complexion and body shape.In addition,a dual-branch deep network was employed to collect deep features from the full-body standing images.Last,the random forest(RF)algorithm was utilized to learn the extracted multifea-tures,which were subsequently employed to establish a TCM constitution identification mod-el.Accuracy,precision,and F1 score were the three measures selected to assess the perfor-mance of the model. Results It was found that the accuracy,precision,and F1 score of the proposed model based on multifeatures for identifying TCM constitutions were 0.842,0.868,and 0.790,respectively.In comparison with the identification models that encompass a single feature,either a single facial complexion feature,a body shape feature,or deep features,the accuracy of the model that incorporating all the aforementioned features was elevated by 0.105,0.105,and 0.079,the precision increased by 0.164,0.164,and 0.211,and the F1 score rose by 0.071,0.071,and 0.084,respectively. Conclusion The research findings affirmed the viability of the proposed model,which incor-porated multifeatures,including the facial complexion feature,the body shape feature,and the deep feature.In addition,by employing the proposed model,the objectification and intel-ligence of identifying constitutions in TCM practices could be optimized.