Research On Early Diagnosis Of Alzheimer’s Disease Based On Machine Learning And Attention Mechanism

Alzheimer’s disease(AD)is one of the most common neurological diseases of the brain.The patient’s disease is specifically manifested as the deterioration of memory and thinking ability,as well as the deterioration of personal behavior and social skills,and people with AD usually accompanied by other physiological diseases.Therefore,the early recognition and diagnosis of AD is of great significance for slowing down the development of the disease.Aiming at the problem that it is difficult to distinguish between people at different stages of Alzheimer’s disease.This paper first uses convolutional neural networks(CNN)to identify and analyze structural magnetic resonance imaging(s MRI),and proposes a CNN recognition model based on slices grouping.In terms of reducing the computational cost,this model proposes a slices grouping strategy and uses a lightweight network model,which greatly reduces the computational cost of training multiple CNN models.By integrating the CNN models corresponding to multiple slices groups,good recognition results have been achieved.Although the CNN recognition model based on slices grouping has achieved good results in recognition,the method of slices processing still has the problem of weakening the spatial connection of the brain tissue structure of the subject and the overall computational cost of the recognition model.Therefore,this paper proposes a recognition model that incorporates attention mechanism.This method uses automated anatomical labeling(AAL)to divide the brain area of the subject,organizes voxels data with the same attribute value in the same brain area together,and constructs the base classifier corresponding to each brain area.At the same time,inspired by the work related to the attention mechanism of deep learning and computer vision,a direct mapping attention mechanism was proposed to improve the accuracy and stability of the recognition model.In the above-mentioned recognition model fused with attention mechanism,for the optimization of the base classifier,this paper proposes a recognition model based on transductive support vector machine(TSVM)transfer learning.This recognition model is inspired by work related to migration learning,and uses the TSVM classification algorithm in semi-supervised learning to optimize the design of the above-mentioned supervised base classifier.Full use of unlabeled sample data and labeled sample data improves the recognition performance of the recognition model again.Based on the recognition model of the fusion attention mechanism proposed in this paper and the recognition model based on TSVM migration learning,according to the experimental results,the high-quality brain regions selected by the recognition model that are highly correlated with AD disease are statistically analyzed,and combined with brain spectroscopy describe it to provide research reference for other researchers who study AD.