Alzheimer's disease classification method based on improved 3D CNN network

Abstract

The invention discloses an Alzheimer's disease classification method based on an improved 3D CNN network, and the method carries out the training based on 3D nuclear magnetic resonance data, and achieves the precise brain disease classification. The method mainly comprises the following steps: 1) modeling by using a 3D CNN: replacing a 2D CNN part in a VGG network by using the 3D CNN; 2) optimizing a model structure: firstly adding a batch normalization layer in the model, and then introducing jump connection; 3) experimental data processing: preprocessing experimental data, and dividing a training set and a verification set; 4) model training: inputting the data into a model, training network parameters, and reserving an optimal model. According to the method, multiple jump connections are introduced into the 3DVGG model for Alzheimer's disease diagnosis for the first time, the performance of the method is superior to that of an existing method, and the method has the advantages of being high in universality, high in robustness and the like.

Description

technical field [0001] The invention relates to the field of computer medical image analysis, in particular to an improved 3D CNN network-based classification method for Alzheimer's disease. Background technique [0002] Alzheimer's disease (AD) is a temporarily incurable neurodegenerative disease. It is the most common case of Alzheimer's disease. Its pre-state is mild cognitive impairment (MCI). human memory and cognition. Mild cognitive impairment (MCI) is an intermediate state between normal aging and dementia, and is a syndrome of cognitive impairment. As early as in the 2011 AD diagnostic criteria and guidelines, MCI cases were considered as the early stage of AD. In recent years, neuroimaging has been widely used as an important biomarker for AD diagnosis. Among them, magnetic resonance imaging is a non-invasive, low-cost imaging technique that can clearly describe the three-dimensional anatomical structure of the human brain. Therefore, using the characteristics ...

AI Technical Summary

Problems solved by technology

Although it has been verified that this method can effectively avoid the high similarity between classes and overfitting problems, it is still affected by the size of the feature scale extracted by 3D CNN

In order to prevent the problem of model degradation caused by deep convolutional neural networks, Karasawa H et al. (Karasawa H, LiuCL, Ohwada H. Deep 3d convolutional neural network architectures for alzheimer's disease diagnosis[C] / / Asian Conference on Intelligent Information and Database Systems. Springer , Cham, 2018: 287-296.) A 3DCNN framework similar to the ResNet structure was designed for Alzheimer's disease classification, in which the convolutional layer has as many as 35 layers, which may make the model more difficult to train, thus affecting the accuracy of the model

Images