Research Of Bayesian Network And Its Application In Diagnosis Of Mild Cognitive Impairment

With the population aging, the number of Alzheimer disease patients has increased year by year. And there is no effective treatment method for the disease. Mild cognitive impairment is a state of cognitive impairment between normal aging and dementia. Therefore, the research of mild cognitive impairment is of important significance to the early diagnosis and preventing, and it has become a hot research recently.As a graphical model, Bayesian network provides a method which enables knowledge graphic, and has many advantages. It can deal with the uncertainty when describing correlation conditions between different components of knowledge. Bayesian network can be applied in diagnosis of mild cognitive impairment to describe the standard indicators by the way of processing uncertain knowledge.The paper introduces the learning of Bayesian structure and parameters, and inference. Using naive Bayesian technology and the means of clinical scale, the paper builds diagnosis model in which diagnosis indicators are used as the nodes. The finding shows that the model can effectively completed the diagnosis of mild cognitive impairment.The traditional algorithms of structure learning need the sequence of nodes in the network, when learning the network structure. To rectify the shortcoming, a new algorithm is presented. The algorithm can learn automatically Bayesian network structure just by mutual information and probability of the training samples, and does not need the sequence of nodes. By analyzing the structure-structure associations and structure-function associations, the paper uses the new algorithm to build Bayesian diagnosis model based atrophy in brain structures. The structures in brain are used as the nodes of the model. Finally, the paper introduces the design and implement of expert diagnosis system for mild cognitive impairment.