Research On The Correlation Between Hypertension Syndromes And Tongue Features Based On Image Recognition Technology

ObjectiveIn this paper,study focus on multiple hypertension related information's standardized collecting and analyzing.Subjects were recruited from oil field worker in Jidong region of Hebei province who participate the physical examination.The hypertension syndromes related facial morphology,complexion and tongue character were obtained.By summarizing the living environment,living habits,past medical history,tongue and pulse diagnosis information and additional information in hypertension patients,the relevant influencing factors of hypertension were analyzed.The syndrome element distribution were also discussed.By integration the digitalized Traditional Chinese Medicine(TCM)face and tongue inspection technology,the face and tongue feature information mining of hypertension patients were realized,the relevant diagnostic value could be automatically extracted,and the optimized application scheme could be provided accordingly.MethodsSubjects were recruited in the oilfield workers in Jidong,Hebei.The collected information including: 1.the inquiry information of family and personal background;TCM clinical diagnosis information;Life status and daily schedule;Medical history and long-term health status.2.TCM physicians' diagnosis and testing information,together with the standardly collected digitalized facial and tongue images.Data researches were made by the support of automatic face recognition technology and data analysis method.In final,659 subjects were qualified from all 751 enrolled subjects,where,160 people in non-hypertension group,216 people in perhypertension group,283 people in normal blood pressure group.Results1.The comprehension correlation factors among group comparisonThe correlation analysis in overall data suggested the following results: systolic and diastolic blood pressure were suggested significantly and positively correlated with gender,age,BMI,waist-to-hip ratio,smoking and drinking habits(P < 0.01),indicate negative correlation in the final degree,years of education,family income,and life style of physical motivation.The parameter of BMI and waist-to-hip ratio could be expressed as the central obesity,and results suggested male subjects who had higher central obesity rate suffer higher risk in hypertension.According to the calculated results of the ordered multi-classification logistic regression method,the significance of the regression conclusion can be obtained from the data including waist circumference,skin fold thickness,age,waist-hip ratio and gender,and the prediction results could explain 56.90% intergroup discrepancy.In the intergroup comparison result,we could found out that in daily living state,hypertension patients always shows longer habitation duration time,keeps standing statue more frequently,earlier wake up time,more likely to suffer cough and snore,higher sleep quality,prefer to take exercise and less meat diary.In self-health evaluation results,hypertension patients were more likely in hyperactivity state and had shorter sleep waiting duration time,they also inclines to suffer edema,standing dizziness,weak voice,amnesia,emotion sensitivity,coupled with chest pain,cold limbs,afraid of cold,dry scrotum and numb feelings.In medical history,hypertension patients were more likely have medical history in hyperlipemia,brain vessel diseases and self-immune diseases.TCM facial and tongue observing diagnosis results suggested that: hypertension patients were more likely have yellow,curdy and greasy tongue fur,tortuous sublingual collaterals,red spots in tip of tongue,irregular heart rate,wide while weaker pulse and string pulse.2.Syndrome elements analysis results in hypertension subjectsThe baseline data analysis in hypertension group suggest indexes difference between sex,mainly expressed in the age,final degree,height,weight,waistline,smoking and drinking habit.Male hypertension subjects shows more in number and younger age than in females,while the blood pressure value were in no difference between sex.In factors analysis of hypertension subjects syndromes,8 dimension data were found as common factors,which could elevate the symptoms description rate level up to 60%.Classification by viscera in TCM,the results showed that the dysfunction in spleen and liver were the main syndromes of hypertension,while kidney secondly.Phlegm and dampness,deficiency of essence and deficiency of qi were suggested as the main disease characteristics of this sample.3.Facial and tongue information analysis and syndrome classificationBy apply the automatic facial identification and recognition system based on AAM model,the digitalized tongue and facial information conclusions would be calculated in more accuracy level.Results suggest that male hypertension subjects had shown wider left eyebrow,tarnish eyes,blue upper lip,higher convergence level in yellow part distribution,tarnish and dark facial color,blue like tongue color.While in female subjects,the shape information had shown hypertension subjects were more likely to have higher value in round like ? asymmetric and wider of eyebrow shape,the distance between eyebrows were larger,the angle were more flat in left eyebrow;Subjects were also suggested slimmer,smaller and the less outer lift angle in double eyes,less sharp curve in left eyes' lower edge,less area in double lips and wider nose.In color information,tarnish and dark eyes and facial color,white lips with yellow ecchymosis were more likely happened in hypertension females,coupled with darker facial color in red and blue section.The sexual difference shouldn't be neglected in hypertension intergroup identification.12 indexes in male suggest significant difference between hypertension group and non-hypertension group,including 1 in facial shape,3 in facial color and 1 in tongue color.While in female subjects,32 indexes suggest significant difference between hypertension group and non-hypertension group,including 24 in facial shape,7 in facial color and 1 in tongue color.Since the shared indexes only including 3 facial color indexes and 1 tongue color index between male and female,so sex difference should be considered when try separate the hypertension subjects by TCM inspection diagnosis methods.Based on the regression analysis results between facial features and hypertension,the number of qualified results,which means the difference between subjects' predicted BP value by facial and tongue features and measured BP values were within 10 mmHg,reached to nearly half of the total tested samples.According to the statistic results and TCM dialectic foundation,the hypertension subjects were divided into 4 subgroups,facial and tongue information were participate in the subgroup analysis.Data were also subgroup by sex.The results shows: 13 indexes in male suggest significant difference among 4 dialectic groups,including 10 in facial shape and 3 in facial color;while in females,overall 15 indexes in female suggest significant difference among 4 dialectic groups,including 3 in facial shape,11 in facial color and 1 in tongue color.In male,Yin and Yang deficiency(YYD)group suggested higher value in the right eyes' inside angle,double eyes' open distance and round like parameters than in Yin deficiency and Yang hyperactivity(YDYH)group,and YYD group shows higher left eyes' outside angle than Excessive Phlegm-dampness(EPD)group.YYD group also suggested lower value in nose width than in YDYH group.Additionally,the lower lip shows larger red component distribution in YYD group than YDYH group.Right eyes shows more yellow component in EPD group while more blue like in YYD group.The facial color blue parts were lighter in YYD group while darker in YDYH group.In female section,EPD group shows higher left eyes' outer angle,left eye round like parameter and right eyes' inner angle,while data in YDYH group suggested the opposite.EPD group shows lighter lower lip color,higher yellow component in right eyebrow and larger distribution in yellow and red part of right eyebrow,while the data in YYD group shows opposite.Additionally Liver fire hyperactivity(LFH)group shows lower light in lower lip.EPD group were more likely to have larger red and yellow component distribution in facial color,LFH group and YYD group shows higher blue component in facial color.Tongue color shows higher yellow component in EPD group and higher blue component in LFH group.Additionally,data in females could cover all 4 subgroups' classification,while in male,no signification data were shown in the liver fire symptoms discrepancy.ConclusionsThis study sorted out the hypertension related information of family background,life style,syndrome,past medical history,TCM physicians' diagnosis.Provided data support for full screening of hypertension risk factors in population,the conclusion could support to provide preclinical risk factor assessment of high-risk hypertension population and sub-healthy population during medical examination,and provide database to the syndrome evaluation and prognosis rehabilitation grading of hypertension patients.Our research applied AI image technique automatically collected the TCM physician required inspection information in facial and tongue image,digital expressed the shape and color character according to TCM principle.It elevated the expression accuracy of TCM inspective information,provide masses of data to assist the TCM physicians to elevate the dialectic results and efficiency,and finally provide more reliable data for the development of automatic TCM diagnosis system.