The Mining Method Of Clinical Thinking And Regularities Of Prescription Use Of Famous Doctors Through Clustering And Association Rule

Objective:To study the applicability of the method of "Associating after Clustering" to the mining of regularities of prescription use and the inference of clinical thinking with the sample of the records of childhood Gilles de la Tourtte's syndrome treated by Professor Liu and tumours treated by Professor Shen.Methods and processes:1. EM algorithm clustering:Cluster 76 prescriptions about childhood Gilles de la Tourtte's syndrome treated by Professor Liu with EM algorithm.2. "Core prescriptions"searching:"Core prescription" is the herbs used very frequently in some type of prescription.3. Hierarchical clustering:Cluster every type of prescription, and find their heart prescription, compare which with the core prescription to verify the rationality of the results of EM algorithm clustering.4. Association rules mining:Mine the associations of symptoms and herbs to gain the the herb-use rules according to different symptoms.5. Analyze the doctor's clinical thinking and herb-ues rules by analyzing core prescriptions and association rules and verify the rationality of the references of herbs used frequently, and deduce the types of syndromes by core prescriptions.6. Verify the rationality of results above by Professor Liu's disciples.7. Analyzing 498 records about tumors by repeating the above 1 to 5 processes.8. Analyzing the records about lung cancer, breast cancer, hepatic carcinoma, ect. by repeating the abovel to 5 processes..9. Verify the rationality of results above by Professor Shen.Results:1. The prescriptions about childhood Gilles de la Tourtte's syndrome were classified into four types. The four core prescriptions all contain Gouteng, Huanglian, Baifuzi, Quanchong, Juhua, Mugua, Banxia, Shenjincao, so it is concluded that liver heat and wind agitation was the basic pathogenesis of childhood Gilles de la Tourtte's syndrome. The first core prescription, with External Wind Invading Lung, also contains Cangerzi, Xuanshen, Banlangen, Shandougen, Xinyihua. The second core prescription, with Pathogenic Factor Disturbing Vitality and Liver Yin deficiency, also contains Zhenzhumu, Cishi, Shichangpu, Chanyi, Danshen. The third core prescription, with less Pathogenic Factor Disturbing Vitality and Liver Yin deficiency, also contains Baichao, Chaihu, Zhigancao, Shichangpu, Yuanzhi, Yujin. The fourth core prescription, with Spleen-deficiency, also contains Baichao, Chaihu, Zhigancao, Zhiqiao, Zhenzhumu and Cishi.2. The prescriptions about tumors were classified into five types. The front four types all contain Biejia. Beishashen, Maidong, Fuling. so it is concluded that Qi-Yin deficiency was the common pathogenesis of tumors. The first core prescription, with stasis, also contains Weijing, Yiyiren, Dongguaren, Jiegeng and Zhebei. The second core prescription, with heat, also contains Houpu, Shihu, Huangqi, Baihuasheshecao and Banzhilian. The third core prescription, with spleen and stomach disorders, also contains Houpu, Shihu. Baizhu. Fangfeng. Huangqi. Baishao, Yanhusuo and Qiyeyizhihua. The fourth core prescription contains Shudi, Danggui, Fuling, Chenpi, Banxia, Zhigancao, Jiegeng, Zhebei, Niubangzi, Gouteng and Suzi, so the corresponding syndrome is Lung and Kidney deficiency with stasis.3. The results above were supported by association rules, by which many herb-use rules according to different symptoms were gained.Conclusion:The method of "Associating after Clustering" that successively analyzes prescriptions with EM algorithm clustering, searching core prescriptions, hierarchical clustering, and then association rules mining between symptoms and herbs, can be used to mine a doctor's clinical thinking and regularities of prescription use rules from its medical records data, no matter which is a large or small one.