Convolution Neural Network Aided Diagnosis Method Of Alzheimer’s Disease

Alzheimer’s disease(AD)is one of the most common brain diseases at present.It will bring mental and physical burden to patients,and also bring great economic pressure to society.In clinical diagnosis,doctors analyze the changes of brain structure and function from brain images to diagnose patients,but there is a lot of subjectivity.Therefore,it is very important to use the computer to aid doctors to make objective and accurate diagnosis.With the rapid development of artificial intelligence,the application of machine learning technology in the auxiliary diagnosis of Alzheimer’s disease is also in full swing.Especially in the field of machine vision,the application of the convolution neural network(CNN)to make the computer understand the medical image data and help doctors make more objective diagnosis has become one of the hot topics.With the wide application of convolutional neural networks in medical image analysis,intelligent classification of non-invasive and high spatial resolution magnetic resonance imaging(MRI)data has become the focus of research.At present,researchers have found that a large number of Alzheimer’s assistance diagnosis work has data leakage problems,leading to the poor generalization of the model,and the diagnosis results are unreliable.In this paper,the free surfer is used to preprocess MRI data automatically,then the output data is clipped and resampled to standardize the input of the model.Finally,the coronary section was extracted from the pre-processing MRI data,and the random partition was compared with the independent partition of the subjects,which verified the negative effect of data leakage.In order to solve this kind of practical problem,this paper proposes an innovative algorithm based on slice voting to reliable Alzheimer’s disease aided diagnosis,and verifies the effectiveness of the model by visualizing areas of the network.In addition,the 3D convolutional neural network model is trained by using 3D MRI data to classify Alzheimer’s disease effectively.The comparative experiment shows that because of the negative influence of some non-lesion areas on model classification,the slice voting method is better than the 3D model,and it is more sensitive to the brain area of the lesion.Therefore,the experiment is carried out in different slice intervals to choose the best slice location for the slice voting method.The experiment shows that the method proposed in this paper is superior to the baseline experiment in many indexes,which provides a new method and idea for computer-aided diagnosis of Alzheimer’s disease.