The Study Of Diagnosing Parkinson's Disease By Artificial Neural Network

Background Parkinson's Disease (PD) is a hackneyed and hazardous neural functional disorder disease. With the elderly population growing, there is a trend that the incidence, morbidity and the deformity rate are growing. But So far, the pathogeny of PD has not yet been known, the diagnosis of PD mainly depends on clinical experience and finally confirmed by autopsy. The rate of misdiagnosis is very high. However, at present with computer and artificial intelligence technology rapidly developing and perfecting, as the embranchment of them, the Artifical Neural Network (ANN) has triumphantly applied in some medical fields. So, the study of clinical diagnosis by using ANN has been focused gradually, however, the report about intelligently diagnosing PD hasn't been found out.Objectives To search for a new inteligent, external, exact method of diagnosing PD, to validate the feasibility of diagnosing PD by the model that is based on combining artificial neural network with Rough Set theory. Methods This study firstly proposed a frame diagnosis target and information management program, on the basis of discussing the study status of epidemic, pathogeny, diagnosis, clinical classification of PD, combining with relation database and the optimum seeking method of document and data analysis; By analyzing and discussing deeply the key technology--ANN and RS and the characteristics of PD and it's clinic diagnosis,the study built a ANN medal(59-30-1) which middle layer function contain regulation of RS theory; By using information management program, to gather clinical information of case history and then transform the format of input and output of ANN; By gathered information ,to train the ANN model by BP study arithetic and regulate the parameter of it; the last,to evaluate the model by review case method. Results the study built and a frame diagnosis target which contained 7 sects 59 items, example for personal information, risky factors, medical records, symptom, character, results of testing, effect of medication and so on; Compile a professional information gather and management program; The network training error was 0.03; The result of network diagnosis was not significantly different from that of expert diagnosis; the accuracy of the network diagnosis was 92.9%. Conclusions The project of diagnosing PD by the model that is based on combining artificial neural network with Rough Set theory is feasible, the method of diagnosing PD has strongly application value. But this study doesn't deal with complex cases, the clinical characters picked-up limit to case history.then, if the model could be applied widely to clinic, the project should be studied more deeply.