Multiple Type Of MR Image Feature Fusion Algorithm In Early Diagnosis Of Alzheimer’s Disease

Alzheimer’s disease(AD) is a kind of irreversible neurodegenerative disease, it is a common variety of neurological dementia with serious implications. The key to preventative treatment is non-invasive early diagnosis. The relevant research shows that volume and texture features revealed through MRI scans are closely related to early development of AD, however the relationships between these features are complex, which requires that MRI feature fusion methods are used. This research has selected multiple shapes and texture features based on MRI scans, designed a feature selection classification ensemble model in order to select and filter appropriate MRI features; this process enables the obtaining of higher and more consistently accurate rates of differentiation. The main research goals of this article are as follows:① Investigate and realize a pipeline program for the preliminary processing and feature selection of brain MRI scans. This program can be used on MRI images for processing procedures such as noise reduction, image registration, skull stripping, image segmentation, selection of structural and textural features, and can be applied on any MRI brain scans. At present there appear to be no single software or algorithms that can provide these processing procedures, this program provides an effective tool for brain MRI- scan based research. This provides a reliable database for classification and cognition based on features.② Based on hypothesis testing methods, we carried out feature fusion algorithms on the MRI scans, used P-value methods to design thresholds for feature fusion selection. Using this method we obtained a useful set of feature subsets in order to distinguish three different kinds of pathological changes shown on the MRI scans. Following this, based on a hypothesis testing method, we performed feature fusion selection on a number of hypothesized features and used an SVM to performed classification, differentiation and testing; finally performing effect analysis on the test results.③ Research a MRI packaging feature fusion algorithm based on a feature selection classification and integration model. This algorithm uses a packaging feature selection model, uses SVM to assess the feature subsets; it uses genetic algorithms to carry out full-scale optimization. The result of this process is the obtaining of the most accurate set of optimized subsets through the SVM process. From this, we were able to design a feature selection classification and integration model.This research realize a pipeline program for the preliminary processing and feature selection of brain MRI scans. Through these two separate methods: hypothesis testing and genetic algorithms, we have designed a new pipeline program based on feature fusion algorithm in order to differentiate three different kinds of MRI brain scan. This has enabled us to improve the accuracy of the system currently used for early diagnosis of Alzheimer Disease and has also improved the stability and adaptability of the system. This has provided a new practical basis for progress in the field of MRI scan-based early diagnosis of Alzheimer’s Disease.