SMRI image classification method and device based on multi-input convolutional neural network

Abstract

The invention discloses a schizophrenia sMRI image classification method based on a multi-input convolutional neural network. The method comprises the steps of collecting structural magnetic resonanceimaging image data of schizophrenia patient and a normal person; preprocessing sMRI image data to obtain grey matter density images which are not subjected to smoothing processing and spatial smoothing respectively, and constructing an original data set; and sending the original data set into a multi-input convolutional neural network model for training, finally inputting sMRI image data into themulti-input convolutional neural network model, and finally outputting a classification result. According to the invention, the images which are not subjected to the spatial smoothing processing andimages which are subjected to spatial smoothing are input into a multi-input convolutional neural network, high-frequency information in an sMRI image lost in spatial smoothing can be compensated, andmeanwhile, the generalization ability of the convolutional neural network can be enhanced due to the noise contained in the images which are not subjected to smoothing processing, so that the problemof low accuracy in schizophrenia sMRI image classification can be effectively solved.

Description

technical field [0001] The invention relates to an sMRI image classification method and device based on a multi-input convolutional neural network, belonging to the technical fields of image processing and artificial intelligence. Background technique [0002] Magnetic resonance imaging (Magnetic Resonance Imaging, MRI) is a relatively new medical imaging technology. It uses static magnetic field and radio frequency magnetic field to image human tissue. During the imaging process, neither ionizing radiation nor contrast agent can be obtained Sharp images with high contrast. With the wide application of magnetic resonance imaging, higher requirements are put forward for the establishment of accurate and detailed magnetic resonance imaging model library. The establishment of the traditional MRI model library relies on the manual retrieval and classification of professional technicians, but the retrieval efficiency is low, the accuracy is poor, and the workload is increasing d...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the current problems of low classification accuracy of sMRI images of schizophrenia and slow model training speed, the present invention proposes a method and device for classifying sMRI images based on multi-input convolutional neural networks, and uses multi-input CNN to solve the problem of sMRI caused by spatial smoothing. The problem of loss of high-frequency information in video images has significantly improved the recognition rate of identifying structural images between patients with first-episode schizophrenia and normal people

Images