Magnetic resonance image fusion method based on weight prediction network

Abstract

The invention provides a magnetic resonance image fusion method based on a weight prediction network. The magnetic resonance image fusion method comprises the steps: firstly designing an artifact removal network, and carrying out artifact recognition and suppression on an input multi-mode image; secondly, constructing a weight prediction network, generating fusion weight prediction graphs of images with different contrast ratios, and endowing regions with rich details with higher weights; and finally, establishing a fusion network, and inputting the image subjected to the artifact suppression and the corresponding weight prediction image to obtain a composite image with the advantage of each modal contrast ratio. According to the magnetic resonance image fusion method, the magnetic resonance image is subjected to artifact preprocessing, so that the image fusion applicability is effectively improved; by integrating the advantages of the images with different contrast ratios, the biological tissue structure can be described more comprehensively and accurately. Compared with a traditional method, complex feature extraction does not need to be designed, the obvious anti-artifact capability is achieved, the fusion effect is good, and medical diagnosis and treatment guided by images are facilitated.

Description

technical field [0001] The invention relates to the technical field of magnetic resonance image processing, in particular to a method for fusion of magnetic resonance images based on a weight value prediction network. Background technique [0002] With the rapid development of sensor and computer technology, medical imaging has become an irreplaceable component in various clinical applications including diagnosis, treatment planning and surgical navigation. Due to different imaging mechanisms, medical images with different modalities focus on different categories of organ / tissue information. In order to provide sufficient information to clinicians, it is often necessary to use medical images of multiple modalities, such as X-ray, computed tomography (CT), magnetic resonance (MR), positron emission tomography (PET), single photon emission computed Tomography (SPECT), etc. In the field of medical imaging, magnetic resonance imaging (MRI) is the best imaging method for head a...

AI Technical Summary

Problems solved by technology

However, these methods have the disadvantages of complex process and the need to manually design feature extraction rules.

In 2017, Liu Yu and others proposed medical image fusion based on convolutional neural network [10] , but it also uses traditional image processing methods, and its training data comes from natural images

The above fusion methods do not consider the possible existence of artifacts in the image during fusion, and there is no fusion method based on deep learning for multi-modal MRI.

Images