Prediction Of Alzheimer’s Disease Based On Dense Convolutional Networks

As the aging of the population becomes more and more serious,more and more attention is now paid to the physical and mental health of the elderly.Alzheimer is one of the main culprits of endangering the health of the elderly,and its high risk and high disability rate have left countless families helpless.Although excellent researchers at home and abroad have conducted in-depth research on Alzheimer’s,so far no highly effective drugs have been developed that can cure.Therefore,it is necessary to intervene and treat patients in the early stages of Alzheimer’s.Doctors determine which patients are more likely to evolve into Alzheimer’s in the early stages of dementia,which is more conducive to preventive and interventional treatment of patients.For medical imaging,magnetic resonance imaging is a powerful tool used to detect Alzheimer’s.It is able to accurately provide the soft tissue structure of the brain,has a high spatial resolution,and is able to clearly display the structural information around the tissue,which can serve as important supporting evidence for Alzheimer’s clinical diagnosis.This paper uses a deep learning model with added attention mechanisms to classify Alzheimer’s and normal elderly people and predict the onset of early dementia patients.Because the attention module has the advantages of flexibility and small amount of computation,this method does not bring additional overhead to the network,and can greatly improve the recognition ability of the network,which is widely used in various types of deep learning research.In this paper,the coordinate attention mechanism and the compression excitation attention module are combined with the dense convolutional network,which uses both the coordinates of the image and the channel information of the image.The attention mechanism can make full use of the valid information in the network and inhibit the extraction of irrelevant information,thereby improving the recognition performance of dense convolutional networks.In this paper,we once again improve on the basis of the attention-intensive convolutional network,adding multi-scale modules to the network to provide different scale characteristics for the network.In this paper,the transformation layer of the original dense network is replaced by a multi-scale structure,and the conversion layer is replaced from the original convolutional layer to the parallel branch structure of three convolutional kernels of different sizes,so that the model can improve the sensing field of the network from each branch of the multi-scale scale,capture the characteristics of different levels of the image for the network,and improve the classification performance of the network.Multi-scale improves the feature extraction ability of the network,increases the sensory field of the network,and effectively uses the global information.The combination of these two modules allows the network to effectively extract multi-scale features and focus on the extraction of effective features,further improving the performance of the network.