Discovery of Diagnosis Pattern of Coronary Heart Disease with Qi Deficiency Syndrome by the T-Test-Based Adaboost Algorithm

• Published in: Evidence-based complementary and alternative medicine Vol. 2011; no. 1; p. 408650
• Main Authors: Zhao, Huihui, Chen, Jianxin, Hou, Na, Zhang, Peng, Wang, Yong, Han, Jing, Hou, Qin, Qi, Qige, Wang, Wei
• Format: Journal Article
• Language: English
• Published: United States Hindawi Publishing Corporation 01.01.2011; John Wiley & Sons, Inc
• Subjects: Adults; Algorithms; Angina; Angina pectoris; Artificial intelligence; Blood platelets; Blood tests; Cardiovascular disease; Coronary artery disease; Data mining; Data processing; Diagnosis; Epidemiology; Heart attacks; Heart diseases; Hospitals; Internal medicine; Low density lipoprotein; Medical diagnosis; Original; Patients; Statistical methods; Student's t-test; Tongue; Traditional Chinese medicine; China; Inner Mongolia China
• ISSN: 1741-427X; 1741-4288; 1741-4288
• DOI: 10.1155/2011/408650

Coronary heart disease (CHD) is still the leading cause of death for adults worldwide. Traditional Chinese medicine (TCM) has a history of 1000 years fighting against the disease and provides a complementary and alternative treatment to it. Syndrome is the core of TCM diagnosis and it is traditionally diagnosed based on macroscopic symptoms as well as tongue and pulse recognitions of patients. Establishment of the diagnosis method in the microcosmic level is an urgent and major problem in TCM. The aim of this study was to establish characteristic diagnosis pattern for CHD with Qi deficiency syndrome (QDS). Thirty-four biological parameters were detected in 52 patients having unstable angina (UA) with or without QDS. Then, we presented a novel data mining method, t-test-based Adaboost algorithm, to establish highest prediction accuracy with the least number of biological parameters for UA with QDS. We gained a pattern composed of five biological parameters that distinguishes UA with QDS patients from non-QDS patients. The diagnosis accuracy of the patterns could reach 84.5% based on a 3-fold cross validation technique. Moreover, we included 85 UA cases collected from hospitals located in the north and south of China to further verify the association between the pattern and QDS. The classification accuracy is 83.5%, which keeps consistent with the accuracy obtained by the cross-validation technique. The association between a symptom and the five biological parameters was established by the data mining method and it reached an accuracy of ~80%. These results showed that the t-test-based Adaboost algorithm might be a powerful technique for diagnosing syndrome in TCM in the context of CHD.