The Study Of Neuroimaging Modeling Method For Mild Cognitive Impairment

The aging of the population leads to the rising morbidity of geriatric disease year by year, the irreversibility of Alzheimer’s Disease (AD) brings great pressure to the life and mental both patients and their family. The study about Mild Cognitive Impairment (MCI), which is a transition phase between normal aging and AD, is important to early prevention and treatment of AD. However, some factors, such as no obvious clinical symptoms of MCI, limited methods to diagnose disease and so on, bring inconvenience to clinical diagnosis. Therefore, it has important significance to study MCI. The emergence of magnetic resonance technology makes the study of brain tissue and function moving to a new stage, the previous research has shown that the gray matter of patients with MCI has a certain degree of reduction, which can be used as the basis for disease diagnosis. In addition, for the convenience of knowledge sharing and reuse, it is necessary to build a standardized and structured knowledge system in the field of MCI. Therefore, this paper constructs an ontology modeling to the diagnosis of mild cognitive impairment, and the main work is as follows:(1) Do the surface morphology analysis. We obtain the cortical thickness values from original magnetic resonance image(MRI), and correspond to the AAL template, compare and analyze the differences of cortical thickness in patients between MCI and normal aging, and then we select brain partitions whose cortex thickness has significant differences between the two groups as a diagnostic features, be ready for the following knowledge modeling.(2) Design and implement the domain knowledge modeling. On the one hand, it has been verified that ontology is effective in the medical field, however there is no available ontology used for MCI diagnosis, so in this paper, we use ontology to model and manage knowledge of MCI and MRI; On the other hand, we use the classical machine learning algorithms to assist building rule set, because the C4.5 algorithm getting better performance in the application of this field, and it has the advantages of the fast training speed and easy to analysis, so we choose C4.5 algorithm training and formalizing the ontology inference rules used for diagnosis of MCI.(3) Implement the reasoning disease. We use Jena API to implement the knowledge inference engine based on domain knowledge model in the eclipse development environment, connect the ontology and rule sets, complete the automatic reasoning of disease and update this ontology.(4) Evaluate the model. We calculate some statistical indexes of different methods by 10-fold cross-validation, such as Precision and Recall, F-score, Kappa statistic and the time to train model, to evaluate different algorithms in this field, the results show that the C4.5 algorithm has high Recall value, F-score value and Kappa statistic value, its Precision value with the same value of BN, the model of training time is second to BN. In order to make the evaluation more convincing, this paper also use paired-sample T-test to contrast the accuracy of C4.5 algorithm and other three classical machine learning algorithms in the field, in summary, C4.5 algorithm is more suitable for the application of this paper, and the knowledge model has its feasibility, it can be applied to clinical auxiliary diagnosis of MCI.