Construction method for CS-MRI image with noise

Abstract

The invention discloses a construction method for a CS-MRI image with noise, and the method comprises the steps: S101, employing a K-means clustering algorithm to carry out the clustering of CS-MRI image signals with noise, and forming M classes; S102, employing a DCT base to build an initial dictionary for each class, enabling the columns of the initial dictionaries to be unitized, and obtaining a dictionary of each class; S103, employing an OMP algorithm according to the dictionary of each class, and obtaining a sparse vector corresponding to each class through solving; S104, employing a K-SVD algorithm, and updating each column of the dictionary of each class and the corresponding sparse vectors; S105, carrying out the repeated iteration of steps S103 and S104 till the number of iteration times reaches a preset threshold value, and obtaining the final dictionary of each class and the corresponding sparse vectors; S106, reconstructing the MRI image according to the final dictionary of each class and the corresponding sparse vectors. The method is higher in precision.

Description

technical field [0001] The invention relates to the fields of compressed sensing and medical image processing, in particular to a noisy CS-MRI image reconstruction method. Background technique [0002] Magnetic resonance imaging (magnetic resonance imaging, MRI) is a kind of imaging of human organs in a uniform static magnetic field using the principle of nuclear magnetic resonance. It has many advantages such as multiple imaging parameters, no ionizing radiation, and tomographic imaging in any direction. widely used. However, conventional magnetic resonance imaging takes a long time, costs a lot, and has a lot of noise, which limits its practical clinical application. [0003] Factors affecting the speed of MRI mainly include two aspects: (1) raw data acquisition speed; (2) k-t spatial data acquisition quantity. Researchers have increased raw data acquisition speed through improved MRI hardware, rapid sequence design studies, and efficient sampling trajectories. However,...

AI Technical Summary

Problems solved by technology

[0007] 1. Images are often two-dimensional signals with complex structural features. Using only one orthogonal transformation basis function cannot effectively represent the structural features of the image, so that the sparsest representation of the image cannot be formed.

[0008] 2. There are certain limitations in image noise processing and edge preservation

The existing dictionary-based learning method learns a global dictionary that is difficult to effectively represent the local structure of various magnetic resonance MRI images

Images