Convolution Neural Network Application for Glioma Image Processing

This paper studies the convolutional neural network applied in medical image processing, and with glioma MRI figure as a specific research problem, the medical images using convolution neural network structure, change the traditional single convolution neural network structure, designed the multiple columns convolution neural networks, effectively solve the glioma grading difficult question. Experiments show that this method is effective in solving such medical problems.

I. Introduction

The convolutional neural network (CNN) has been applied to various fields of science, whether in the field of speech analysis or image classification, text processing, etc. CNN belongs to artificial neural network, it is a kind of supervised learning method. One of its key features is weight sharing, which reduces the complexity of the entire network and further reduces the number of required weights. Raw data such as images, text can directly input CNN for training, through a certain number of convolution and the sampling process, extract helps classification characteristics, such as the shape of the image, such as texture features[l]. The convolutional neural network is more advantageous to other networks in dealing with complicated geometries, and the treatment of CNN applied to medical image is a key point in the research of computer-aided diagnosis and treatment system. It is of great significance to improve the diagnostic efficiency, increase the accuracy of diagnosis, reduce the time of patients' examination, and reduce the work pressure of medical staff.