| Background:Due to the problem of human-computer interaction between the existing clinical auxiliary diagnosis and treatment technology of traditional Chinese medicine and physicians,the development and promotion of clinical auxiliary diagnosis and treatment technology of traditional Chinese medicine are limited.Exploring the information correlation logic of TCM dialectical theory and treatment with information value and integrating it into the construction of auxiliary diagnosis and treatment technology is expected to become a feasible way to further guide auxiliary diagnosis and treatment technology.Objectives:Based on the universal logic of TCM dialectical treatment information correlation,an auxiliary diagnosis and treatment platform is constructed and its effect is verified.Methods:1.Theoretical discussion:Clarify the connotation,characteristics,and research dimensions of the informational association of TCM identification and treatment,and sort out the implementation ideas of this study.2.Expert investigation:First of all,starting from the three aspects of diagnosis or treatment information collection,diagnosis or treatment information processing,and presentation and diagnosis/treatment information quality control,10 experts were investigated,and the correlation logic of Traditional Chinese medicine identification and treatment information with Fang Zheng as the core was obtained.Based on those,a questionnaire was compiled to investigate 160 clinical experts in various internal medicine departments of traditional Chinese medicine to obtain the clinical universality of the above identification and treatment information correlation logic in actual diagnosis and treatment.3.By analyzing and transforming the above correlation logic content,we set a new design principle,and demonstrate the construction of an auxiliary diagnosis and treatment system that can reflect the correlation logic of various types of identification and treatment information.Among them,two diseases of chest palsy and vertigo are taken as examples,retrospective and prospective clinical data are collected,and relevant model tools for auxiliary diagnosis and recommended treatment are established.In terms of diagnosis,two different ideas of prediction models are set up.On the one hand,from November 2021 to June 2022,prospective data of patients with vertigo and chest palsy in outpatient departments of cardiology were collected,and the original data set of chest palsy and the correction dataset of dialectical results of "square measurement" were trained by Bayesian model respectively to obtain the best model for predicting chest palsy syndrome(clinical interpretability is vague);On the other hand,based on the results of expert surveys on the dialectical points of vertigo disease,the ROC curve is drawn by accumulating the expert voting scores of each dialectical key point under the symptom elements,and the Yoden index is calculated to select the optimal integral threshold for determining the symptom elements,and then the expert knowledge syndrome diagnosis model(more interpretable)is constructed.All tongue images are collected in the form of electronic pictures,and through the convolutional neural network model,training to identify the tongue color of the tongue image(light tongue,reddish tongue,red tongue,purple tongue),tooth marks,the color of tongue moss(tongue moss yellow or not),thickness of tongue moss(moss thick,tongue moss thick normal,moss less or less moss)and other characteristics.In terms of treatment,the inpatient electronic medical record records from July 2010 to January 2021 in the hospital case database were retrieved,and the various types of symptoms involved in chest palsy and vertigo were extracted and counted,and judged,segmented according to the symptomatic elements concerned in the prospective data collection,to decompose all patient type diagnosis in the retrospective case database into binarized characteristics of the symptom elements,and construct an expanded database of the stock of patients with chest palsy and vertigo.According to the intelligent correlation dialectical results of the patient’s gender,age,and individualized clinical characteristics at the time of patient visit,case matching is carried out in the database,and the association rules of the matched patients are mined based on the Apriorism association rule algorithm in the Python3 MLX tend database,and the recommendation of high-frequency single-flavor drugs and highfrequency drug combinations is formed.4.The auxiliary diagnosis and treatment platform built by the application carried out auxiliary diagnosis and treatment for 20 patients with chest palsy and 20 cases of vertigo,recorded the results of auxiliary diagnosis and treatment with the actual decision-making results of physicians,and compared their consistency,and conducted semi-structured interviews with physicians to qualitatively analyze the effect of auxiliary diagnosis and treatment and the improvement of human-computer interaction level.Results:1.Through expert interviews,32 TCM identification and treatment information correlation logics were obtained after the merger.In the expert questionnaire survey,104 valid questionnaires were recovered,and the results reflected that the clinical universality of the logical content of the identification and treatment information correlation was effective.2.Relying on the above three dimensions,through expert interviews,32 TCM information related logic points were obtained after merger.In the expert questionnaire survey,104 valid questionnaires were recovered,and the results showed that the clinical universality of each logical point of information related to diagnosis and treatment was good.3.Results of the diagnosis model of Chest Palsy Bayesian Symptoms:211 patients with chest palsy formed a tongue image+consulted Bayesian model,of which 147 patients with chest palsy had completed follow-up of efficacy,of which a total of 134 cases had improved diseases or symptoms,and 13 patients had no improvement in disease or symptoms.Based on the original dataset of 211 patients and their efficacy feedback,the corrective data set of"taking the square test" dialectical results were constructed,and the accuracy rates of the former diagnosis were 88.89%of blood stasis,66.67%of qi deficiency and 85.19%of sputum turbidity,respectively,and the accuracy of the latter diagnosis of blood stasis was 92.59%,qi deficiency was 70.37%,and sputum was 85.19%.The latter was chosen as the clinical diagnostic model for chest paralysis.4.Results of the diagnosis model of vertigo expert knowledge syndrome:Through literature research and two rounds of expert investigation,the selection of common symptoms and dialectical points of vertigo was carried out,and the dialectical integral thresholds of impotence,yin deficiency,and phlegm wetness were obtained through the ROC curve,which was 31.41,20.8,and 21.53,respectively.5.Tongue image identification results:210 cases of hypertension and 231 cases of chest palsy patients were collected from the outpatient clinic,and through the convolutional neural network model,431 prospectively obtained tongue images were trained.The accuracy of tooth marks is 0.775.And the accuracy of tongue color(tongue light,tongue reddish,tongue red,tongue purple and dark),tooth marks,tongue color(tongue moss yellow no),tongue thickness(moss thickness,tongue thickness is normal,no moss or less lichen)reached 0.65,0.81,0.79,respectively.6.Auxiliary diagnosis and treatment platform application verification results:20 patients with prospective vertigo were collected for verification of the interpretable four-diagnosis and composite reference method based on expert knowledge,and the predictive type was obtained,compared with the clinician’s identification type selection,the dialectical consistency rate of impotence syndrome was 80%(16/20),the dialectical consistency rate of sputum wet syndrome was 65%(13/20),the dialectical consistency of yin deficiency evidence was 85%(17/20),and all the dialectical classification of patients was completely consistent with the physician(12/20).At the same time,it matches the prescription drugs of similar patients in the case database for the reference of doctors.The intelligent correlation dialectical algorithm of 20 patients with prospective chest palsy was collected for multimodal four-diagnosis information fusion,and the predictive type was obtained,comparing the clinician’s identification type selection,the dialectical consistency rate of blood stasis was 85%(17/20),the dialectical consistency rate of qi deficiency was 90%(18/20),the dialectic consistency of sputum turbidity was 85%(17/20),and all the dialectical classification combinations of patients were completely consistent with the physicians(14/20),and the prescription drugs of patients with similar patients in the recommended case pool were matched for the reference of physicians.At the same time,the relationship network of various Chinese medicines is drawn.7.Physicians use the auxiliary diagnosis and treatment platform to experience the interview results:a semi-structured interview was conducted with two young physicians in the cardiology clinic,and the qualitative analysis of this auxiliary diagnosis and treatment system can indeed improve the level of human-computer interaction to a certain extent and assist them in dialectical medication.Conclusions:Starting from the collection,processing,presentation,and the quality control of diagnosis and treatment information,this study constructs an auxiliary diagnosis and treatment platform that embodies the information correlation logic in dialectical theory and treatment,and verifies its role in improving the docking degree of human-computer interaction in diagnosis and treatment practice,and demonstrates the construction of auxiliary diagnosis and treatment technology that integrates the information correlation logic of traditional Chinese medicine dialectical theory and treatment. |
PDF Full Text Request