Study And Application Of Alzheimer Disease Forecast Expert System

Alzheimer Disease (AD) is a high incidence of senile diseases. As the global aging of the population is becoming increasingly serious, AD is increasingly becoming a serious social problem. Mild Cognitive Impairment (MCI) is a state between normal aging and AD. The latest reports said that MCI is a representative of the very early AD stage. Therefore the study of MCI is of great significance to AD. The traditional method to diagnose MCI is based on the scale testing. It can make direct diagnosis if the result of the scale testing is consistent. But when the result comes to conflict, it needs a physician to diagnose and judge.This paper improves the traditional MCI diagnostic method, proposes and implements a diagnostic model based on one-class support vector machine. A B/S mode expert system based on this diagnostic model using J2EE framework is constructed in order to diagnose MCI.In order to operate more convenient, this paper adopts revised diagnostic criteria of MCI, and collects subjects' data of memory and cognitive ability and pretreats. The diagnostic model uses identified state data as its training data to train the inference engine, and uses uncertain state data as its test data. Finally the diagnostic model gives the result of whether there is a MCI tendency of uncertain subjects. In the establishment of the inference engine, the paper uses one-class support vector machine classification technology. It focuses on several aspects such as choice of kernel function and parameter optimization, the pretreatment of training sample, training algorithm and so on. The paper uses grid search and cross validation to avoid over fitting problems. In the process of specific, a new method of handling sample is proposed. This method reduces the training time, but the correct rate remains basically unchanged.The experiment finds out 55 persons at the age of 45-75 from Dalian and Shenyang. It collects their data of memory and cognitive ability. The experiment compares the diagnostic results of the physician and the system. The results show that the prediction accuracy of the diagnostic model reaches 85.7%. This model avoids man-made factors. It is an effective model for diagnosing MCI.