Research On Alzheimer’s Disease Classification Based On Algorithm Deep Learning

Alzheimer’s disease(AD)is a long-lasting,irreversible,and incurable neuropathy disease that commonly occurs in the elderly.It is commonly called Alzheimer’s disease.In recent years,the problem of China’s population "aging" has become increasingly serious.Effective research on the early diagnosis of AD can greatly alleviate the family and socioeconomic pressures brought by patients,and provide meaningful guidance for the patient group and medical diagnosis to a certain extent.Therefore,how to effectively diagnose early AD has become an important research direction.With the continuous deepening of Deep Learning(DL)and medical technology,more and more fields have switched from traditional manual analysis to computer-aided diagnosis.Magnetic resonance imaging(MRI),as one of the increasingly mature medical imaging technologies,has become a tool for studying important data sources of the brain due to the continuous updating and improvement of MRI and its high-dimensional and informative features.However,in the face of high-dimensional and massive medical data,how to use the DL network realize AD classification efficiently and conveniently,there are still certain challenges.In order to construct an early diagnosis framework for AD,which is easy to operate,and can be accurately diagnosed,and has a strong stability.This article uses the subject’s brain MRI image data as the research object,and combines traditional methods and DL-related methods to explore three types of AD,Mild Cognitive Impairment(MCI),and Normal Control(NC)Classification of the population.Its main contents are as follows:(1)A Hierarchical Ensemble Learning(HEL)framework is constructed for the classification of AD.The framework steps are as follows: In the first layer,the feature matrix of all MRI slices is extracted through a pre-trained DL network and input to the classifier to obtain the slice-level coarse classification result.In the second layer,the coarse classification result is classified by a classifier to obtain a single slice of the fine classification result.In the third layer,the result of the fine classification is input into the classifier,and the final classification result of the AD at the subject level is obtained.Experimental results show that the integration accuracy of the framework on slices and subjects reaches average accuracy of 0.9358 and 0.9912,respectively.(2)Combined with the DL method,a one-dimensional convolution neural network(1-D CNN)classification framework was constructed for AD classification.The framework first uses DL network to perform feature extraction on all pre-processed MRI slices,and then classifies them through 1-D CNN to achieve automatic classification of AD.A large amount of experimental data shows that the classification accuracy of the 1-D CNN framework for AD vs.MCI vs.NC,AD vs.MCI,AD vs.NC,MCI vs.NC has reached the average accuracy of 0.9550,0.9933,0.9945 and 0.9597,respectively.