Nuclear magnetic resonance image processing device and method based on convolutional neural network

Abstract

The invention provides a nuclear magnetic resonance image processing device and method based on a convolutional neural network, and relates to a method for carrying out computer-assisted identification on a hippocampal head of a hippocampus, CA1, CA2, CA3, CA4 / dentate gyrus, hippocampal fimbria and a hippocampus tail, and a classification method for realizing three classification of three groups of a normal old age group, a clinically-diagnosed amnestic mild cognitive impairment group and a clinically-diagnosed Alzheimer's disease group through the deep learning convolution neural network of artificial intelligence. The method determines a training set, a verification set and a test set for deep learning of the artificial intelligence by utilizing nuclear magnetic resonance image data of the normal old age group, the clinically-diagnosed amnestic mild cognitive impairment group and the clinically-diagnosed Alzheimer's disease group, and can convert nuclear magnetic resonance three-dimensional data into two-dimensional data by utilizing selected two-dimensional slices.

Description

technical field [0001] The present invention relates to a nuclear magnetic resonance image processing device and method based on a convolutional neural network, in particular, to a convolutional neural network based on deep learning for Alzheimer's disease and amnesiac mild recognition using cranial nuclear magnetic resonance imaging. Apparatus and method for image processing and analysis of pairwise discrimination between impaired and normal elderly groups. Background technique [0002] Alzheimer's disease is a neurodegenerative disease. Existing research work has confirmed that in the MRI image files of the head, although there are individual differences in the MRI image data of each subject, in the sense of biostatistics, Alzheimer's disease, Alzheimer's disease, Amnestic mild cognitive impairment and the three categories of normal elderly group exist differences between the three categories of MRI imaging data. The deep learning method of machine learning using artific...

AI Technical Summary

Problems solved by technology

However, this document only processes the three-dimensional voxel data of the MRI of the head, and the technical solution used is a three-dimensional convolutional autoencoder

Since it is necessary to process all the three-dimensional voxel data of the magnetic resonance imaging of each subject's head, this technical solution requires a huge amount of calculation. Even if a high-performance computer or a high-performance computer cluster is used, this technical solution will not Difficult or impossible to use clinically

In addition, the image data of the MRI of the clinical common scanning head is the image data of 20 to 40 two-dimensional slices, the method in the above-mentioned arXiv:1607.00556v1 document cannot be used for the MRI of the clinical common scanning head Imaging image data

Images