Research On TCM Auxiliary Intelligent Inquiry Model For T2DM Based On Symptom Association Law Mining

Inquiry is one of the important contents of the four TCM diagnosis.Taking the inquiry of type 2 diabetes as an example,this study provides new ideas for the development of the AI inquiry model of traditional Chinese medicine by designing a computer algorithm in line with the theory of traditional Chinese medicine.Objectives:(1)It is to establish a standardized database of common clinical symptoms of type 2diabetes mellitus;(2)It is to construct a TCM assisted intelligent inquiry model for type 2 diabetes which conforms to the TCM clinical syndrome differentiation thinkingMethods:(1)With "diabetes" as the key word,TCM medical cases of type 2 diabetes were collected from an ancient and modern medical case cloud platform,China National Knowledge Network,VIP,Wanfang,famous studios such as "On Sugar and Miscellaneous Diseases",online quality periodical databases and modern famous books.Data cleaning was carried out on the medical case data,and 300 TCM medical cases of type 2 diabetes meeting the inclusion criteria were finally included.(2)The "Four-diagnosis Information Collection Table" of TCM Syndrome Research Base of Fujian University of Traditional Chinese Medicine is used to standardize symptoms in medical records.If symptoms cannot be standardized by the "Fourdiagnosis Information Collection table",they should be standardized by combining the "Differential Diagnosis of TCM Symptoms" and the "Standard of Terms for Common Clinical Symptoms of TCM".In case of different synonymous words,alias and proper name,etc.,A standardized database of common clinical symptoms of type 2 diabetes was finally established based on the expert opinions of the team.(3)By referring to "TCM Internal Medicine" and "TCM Clinical Diagnosis and Treatment Terms Part 2: Syndromes" issued by the National Health Commission and the State Administration of Traditional Chinese Medicine,the syndrome types were standardized,and the common clinical witness types of type 2 diabetes were obtained.(4)A feasible TCM assisted intelligent inquiry model for diabetes was constructed using a Heterogeneous graph based diabetes mellitus search algorithm(HGDMSA).The accuracy rate and recall rate are used to evaluate the performance of the algorithm model.Results:(1)After standardized treatment,a total of 252 symptoms were included in the final database of Type 2 diabetes symptoms.The total frequency of symptoms was 3959 times,and there were 11 symptoms with frequency greater than 80 times.The symptoms with the frequency from high to low was fatigue,dry mouth,thin pulse,thirsty drinking,pulse string,tongue cloy,thirst,chronic frequent urination,heavy pulse,red tongue,and waist and knee soreness.(2)After standard treatment,a total of 16 syndrome types were obtained,and the syndromes with frequency statistics from high to low were: Syndrome of spleen and kidney deficiency,syndrome of Qi and Yin deficiency,syndrome of unregulated liver and spleen,syndrome of spleen deficiency and dampness accumulation,syndrome of liver and kidney Yin deficiency,syndrome of qi and blood stasis,syndrome of dampness-heat,syndrome of kidney Yin deficiency,syndrome of kidney deficiency and blood stasis,syndrome of stomach heat incandescence,syndrome of qi stagnation and blood stasis,syndrome of liver stagnation and fire,syndrome of Yin and Yang deficiency and blood stasis,syndrome of lung heat and fluid injury,syndrome of phlegm and dampness,syndrome of Yin deficiency and fire flourishing.(3)A diabetes graph search algorithm model based on heterogeneous graph was constructed.When the proportion of symptom input was 0.5-0.9,the accuracy rate was greater than 0.6,the lowest was 0.6308,and the highest was 0.8655.Although the accuracy rate had certain changes,it did not affect the change of recall rate on the whole,and the recall rate was all 1.It showed that this method can ensure that all the recommended symptoms can be recommended and the coverage rate of recommended symptoms can be guaranteed.Conclusions:(1)The establishment of a standardized database of common clinical symptoms of diabetes mellitus in traditional Chinese medicine is conducive to the mining of symptom association rules by computer algorithm,so as to infer accurate syndrome types.It can be seen that the higher the standard degree of traditional Chinese medicine symptoms,the more conducive to the implementation of computer algorithms,and the more accurate the syndrome type inference.(2)By constructing the diabetes graph search algorithm model based on heterogeneous graph,symptoms association and recommendation can be quickly and effectively completed with high accuracy.It can be seen that the graph search algorithm is a suitable algorithm model for the study of intelligent inquiry and diagnosis of traditional Chinese medicine.