Multi-sequence MRI image segmentation method based on residual network

Abstract

The invention belongs to the crossing field of computer vision and machine learning, and relates to a multi-sequence MRI image segmentation method based on a residual network. The method comprises thefollowing steps: firstly, merging MRI multi-sequences in a data set as a plurality of channels of an input image, and separating segmentation sequences into different types of segmentation images. According to the invention, a residual network is introduced on the basis of a classic encoder decoder segmentation network, and an original encoding and decoding network is replaced by a residual unit.And more local features are extracted by utilizing skip connection in the residual error unit. Jumping connection between residual error units of the same level is increased, and extraction of globalfeatures is achieved. Besides, for the problem of unbalanced segmentation categories caused by large proportion difference of a target area and a background area in the MRI image, cross entropy lossand Dice loss are linearly combined, and the Dice loss is weighted to solve the problem.

Description

technical field [0001] The invention belongs to the intersection field of computer vision and machine learning, and relates to a multi-sequence MRI image segmentation method based on a residual network. Background technique [0002] With the development of computer vision, it is becoming more and more common to use computer technology to analyze and process images and videos. As an important part of image processing, medical image segmentation plays a key role in analyzing anatomy, locating diseases, and planning surgical procedures. In the field of biomedicine, with the increasing number of medical equipment and medical imaging data, it is difficult to rely on manual methods to analyze and process medical imaging data to cope with the rapid development of the medical field. Therefore, it has become particularly important to apply computer technology to accurately and quickly segment and detect tissue structures in medical images. It is of great significance to make full u...

AI Technical Summary

Problems solved by technology

[0007] Aiming at the problems of insufficient extraction of feature information in the traditional MRI image segmentation process, lack of effective use of multiple sequences of MRI images, and small sample size of medical data sets that are difficult to support deep network training, the present invention proposes a method with residual Multi-sequence MRI image segmentation method based on difference mechanism (Multi-ResUnet)

Images