Study on MRI Medical Image Segmentation Technology Based on CNN-CRF Model

Naiqin Feng,Lijuan Qin,Xiuqin Geng

doi:10.1109/access.2020.2982197

Abstract

Image segmentation is an important technique for segmenting images without overlapping each other and having their own features. It has been rapidly developed in the field of medical imaging, but there is currently a difference between classification accuracy and segmentation accuracy for medical image segmentation. In this paper, the deep convolutional neural network is combined with the cascading structure, and a uniform learning framework is established with the use conditional random field. This paper first adds a cascading structure under the deep convolutional neural networks (DCNN) framework to more effectively simulate the direct dependencies between spatial closure tags. Secondly, the conditional random field (CRF) is used for post-segmentation processing, which effectively solves the contradiction between the segmentation accuracy and the network depth and the number of pooling times in the traditional convolutional network. Secondly, the CRF is used for post-segmentation processing, which effectively solves the contradiction between the segmentation accuracy and the network depth and the number of pooling times in the traditional convolutional network.

Highlights

In recent years, relevant research institutions have successively proposed algorithms such as normalized cut [1], graph cut [2], mean shift [3], and level set [4] to segment images, but the above algorithms are used for natural image segmentation
SEGMENTATION METHOD BASED ON CNN-conditional random field (CRF) Based on the novel segmentation method of MRI segmentation of brain tumors proposed in this paper, combined with feature extraction and convolutional neural network and applied to the diagnosis of brain tumors, the automatic segmentation of MRI images is realized, aiming at the size, shape and position of brain tumors
The brain tumor segmentation method is upgraded in the convolutional neural network segmentation compared to the original method structure, and the conditional random field model is used in the post-processing part to avoid mis-segmentation of the previous fusion results

Summary

INTRODUCTION

Relevant research institutions have successively proposed algorithms such as normalized cut [1], graph cut [2], mean shift [3], and level set [4] to segment images, but the above algorithms are used for natural image segmentation. The method of convolutional neural network combined with cascading structure can effectively and accurately realize the automatic segmentation of MRI images of brain tumors, saving doctors’ working time and improving diagnosis efficiency. The fused results are post-processed based on GMM model, and the segmentation results are generated [28], [29] This method is more accurate and better than the conventional method, but at the same time it uses the basic convolutional neural network, that is, CNN has obvious deficiencies, the network structure is backward and can not effectively simulate the direct dependence between the spatial closure tags. The brain tumor segmentation method is upgraded in the convolutional neural network segmentation compared to the original method structure, and the conditional random field model is used in the post-processing part to avoid mis-segmentation of the previous fusion results. We evaluate the segmentation accuracy and classification accuracy of the four segmentation methods

SEGMENTATION PERFORMANCE EVALUATION

CONCLUSION