4D-MRI super-resolution reconstruction method based on double-dictionary learning

Abstract

The invention discloses a 4D-MRI super-resolution reconstruction method based on double-dictionary learning, mainly for solving the problem of quite low special resolution of 4D-MRI reconstructed by use of a conventional method. The main steps comprise: performing retrospective ordering on multi-layer sagittal2D dynamic MRI, leading out 4D-MRI, and cutting a low-resolution image to be improved in a coronal direction; extracting a training image from multi-layer coronal2d dynamic MRI acquired in advance; then training the training image by use of a KSVD algorithm to obtain a high-resolution dictionary and a low-resolution dictionary; and performing super-resolution reconstruction on the low-resolution image to be improved by use of a relation between the high-resolution dictionary and the low-resolution dictionary. The method provided by the invention can effectively improve the special resolution of the 4D-MRI, and can be applied to MRI super-resolution reconstruction in multiple directions.

Description

technical field [0001] The invention belongs to the technical field of image processing, relates to a medical image processing method, and can be used for 4D-MRI super-resolution reconstruction. Background technique [0002] 4D-MRI, also known as three-dimensional dynamic MRI, will provide verifiable information for the current radiation therapy plan. Now, there are two dynamic MRI technologies, one is 3D scanning and the other is 2D multi-slice scanning. The 3D scanning method can provide real time information, but the image quality is not high. 2D multi-slice scanning techniques are better able to accommodate motion during scanning, however more post-processing is required to derive 3D information. [0003] Now, researchers mostly use multi-slice 2D dynamic MRI to generate 4D-MRI. Among them, Siebenthaletal repeatedly collected multiple layers of 2DbSSFP sequence data, and then evaluated the similarity between these sequence data and the navigation layer to sort the imag...

AI Technical Summary

Problems solved by technology

3D scanning methods can provide real time information, but the image quality is not high

This method is capable of producing high-continuity abdominal MRI stereograms, but has many disadvantages: first, this method requires specific MRI sequences that cannot be used in commercial systems; second, it will be used by It is hampered by a rather complex image processing process; third, how to choose an appropriate liver region to derive a representative average respiratory cycle is still an uncertain issue

[0006] Using the method described above, we were able to obtain a representative 4D-MRI, however, due to the low resolution of the multi-layer 2D dynamic MRI used to derive the 4D-MRI, each layer is about 10mm thick, so we obtained The serial 2D-MRI has low resolution

Images