The Study Of Standard Model Establishment And Application Of TCM Syndrome Differentiation Of Type 2 Diabetes Based On Data Mining

Objective: To explore methods for establishing diagnostic criterion of TCM (traditional Chinese medicine) syndrome differentiation by studying the set-up of diagnostic standard model of syndrome differentiation of Type 2 diabetes . Methods: Our research was carried out in the following steps: collecting the Fisher-iris data of model check-up ,which is popular in the world; Secondly, the document data from 1984 to 2005 on type 2 diabetes syndromes was gathered by way of document datebase index on Internet(such as CBMdisc and CNKI ). By clinical epidemiological investigation, and clinical questionnaires of the patients, pooling the clinical data from the diagnosed patients of Type 2 diabetes,who were met with the selective criterion, from No.1 Affiliated Hospital of Henan College of TCM and Anyang TCM Hospital and Kaifeng TCM Hospital in 2003 - 2006, and establishing document data-base and clinical data-base on the basis of preprocessed. The study of standard model establishment and application of syndromes diagnosis criterion was made by use of Artificial Neural Network(ANN) and fuzzy system (FS). And a finally programming was completed with MATLAB 6.5.The successive procedures included setting up dynamic kohonen net model, that is, based on it, increasing formation of dynamic neuron in order to make up a dynamic adaptive kohonen net, whose output could reflect the distributive characteristics of sample input chart. The whole process still continued with following procedures: after dynamic neurons being stable, changing weight vector between input layer and dynamic layer into fuzzy inference system and corresponding regular subsidiary function centre, and further repeatedly adjusting fuzzy regulations and corresponding function parameter in accordance with recognition value and thus obtaining the most optimal fuzzy regulations. The model reliability was tested with Fisher-iris data check-up. With this model the clinical data was explored and in accordance of TCM basic theories, the diagnostic criterion of syndrome differentiation of Type 2 diabetes was achieved, whose rationality was, too, determined. Result: Based on the Fisherman-Iris data, When the dynamic layer was turn into stability, Nine neurons in the dynamic layer of the neuron model obtained by dynamic study, the correct recognition rate of the sample test was 94%. The result gained based on document data was that The composition of 22 neuron group in the dynamic layer with the dynamic study in the method and The rate of coincidence identity of test sample was almost 86%. The outcome obtained based on clinical data showed that The composition of 118 neuron group in the dynamic layer with the dynamic study in the method and The rate of coincidence identity of test sample was almost 74%. It was clear for the 6 syndromes to be differentiated according to the main symptom and the subsymptom by transformed rule and the selected coordinate of Primary symptoms and Subsidiary symptoms and differentiated syndrome criterion of TCM. The 6 syndromes were deficiency syndrome of both qi and yin(DSBQY), deficiency syndrome of kidney-yin(DSKY), blood stasis syndrome (BSS), syndrome of lung-heat exhausting the body fluid(SLHEB), syndrome of dominant of hyperheat of stomach(SDHS), syndrome of obstruction by dampness and heat(SODH).Common syndromes diagnosis criterion of type 2 diabetes include such 6 types as:1) DSBQY consisted of Primary symptoms and Subsidiary symptoms. Primary symptoms included lassitude, red tongue, palpitation, thin fur on the tongue. Subsidiary symptoms included thirst and polydipsia, thready and rapid pulse, spontaneous sweating, scandy fur, insomnia and dream-disturbed sleep, feverish sensation in the chest, palms and soles, dryness of the mouth and the tongue, dry stools, shortness of breath, night sweating. 2)DSKY consisted of Primary symptoms and Subsidiary symptoms. Primary symptoms included frequent urination, polyuria, red tongue, lassitude in lumbus and limp knees, rice-water urine and lipuria. Subsidiary symptoms included scandy fur on the tongue, thready and rapid pulse, feverish sensation in the palms and soles, thirst, tinnitus and deafness, dryness of the mouth and the tongue, insomnia and dream-disturbed sleep. 3) BSS consisted of Primary symptoms and Subsidiary symptoms. Primary symptoms included ecchymosis in the tongue, tarnished complexion, taut pulse, dark and grayish tongue. Subsidiary symptoms included dark red tongue, numbness of the four limbs, pain of the four limbs.4) SLHEB consisted of Primary symptoms and Subsidiary symptoms. Primary symptoms included thirst and polydipsia, red tongue, dryness of the throat, yellow fur on the tongue, frequent urination, polyuria. Subsidiary symptoms included thin fur on the tongue, polyphagia and frequent hunger.5) SDHS consisted of Primary symptoms and Subsidiary symptoms. Primary symptoms included eat more and hungry, yellow moss, constipation, thirsty and drink more, red tongue. Subsidiary symptoms included heat body, tantrum, heartbeats, more urinate. 6) SODH consisted of Primary symptoms and Subsidiary symptoms. Primary symptoms included feel tightness in the stomach, greasy tongue, yellow moss, feel tightness in the abdomen, fatigue. Subsidiary symptoms included red tongue, tinnitus, bitter mouth, thirsty and drink much, greasy mouth, dry mouth and dry tongue.Conclusion: The result of the fisherman-Iris data showed that the fuzzy classification rules expressed in higher accuracy the law lies in learning sample and proved the reliablity of the model. By comparing the result gained based on document data with the outcome obtained based on clinical data, it were found for DSBQY, BSS, SLHEB, DSKY to have the same Primary symptoms and Subsidiary symptoms, the result showed the model could be used for the study on syndromes diagnosis standard system of type 2 diabetes.