| Alzheimer’s disease（AD）is a disease that causes memory loss,slow brain function,and gradual change of intelligence.The disease can lead to loss of cognitive function,especially the loss of cerebral cortex function,including memory,judgment,and abstract thinking ability.An effective way to prevent AD is to effectively diagnose and predict the probability of suffering from AD at an early stage.Predecessors used fine-grained data to conduct certain research on AD.However,obtaining fine-grained data is costly and time-consuming,which is not conducive to the early research of AD,especially the screening stage.Compared with the use of biomarkers such as blood tests,the use of coarse-grained（i.e.,low-cost,short time）heterogeneous data can improve the effectiveness of assisting clinical decision support systems.The goal of this article is to use CDR-SB（Clinical Dementia Rating Scale-Sum of Boxes）as the standard to extract early signs of AD.Data preprocessing based on the open-source dataset ADNI（Alzheimer’s Disease Neuroimaging Initiative）,including raw data cleaning,data normalization and discretization,feature extraction and aggregation analysis,and missing data imputation And unbalanced data processing.Then,the static and dynamic Bayesian network modeling methods are used to establish the model,and the built model is analyzed and compared with the C5.0 decision tree and support vector machine model.The results show that the Bayesian network model established by CDR-SB can visualize the strong and weak relationship between symptoms and provide clinicians with early diagnosis of AD.This article discovers the patient’s hippocampus,asks whether the patient has had a mental illness,and performs various scale tests on the patient as much as possible,which can provide clinicians with information and data that need to be focused on in assisting judgments in daily diagnosis and treatment. |
