Fast magnetic resonance imaging reconstruction algorithm based on high-dimensional correlation prior information

Abstract

The invention provides a fast magnetic resonance imaging reconstruction method based on high-dimensional correlation prior information. The fast magnetic resonance imaging reconstruction method comprises the following steps: preparing an open source brain data set SIAT and dividing the open source brain data set SIAT into a training data set and a test data set; optimizing a plurality of network models under the condition of different noise levels in combination with a derived law of great numbers and a multi-channel high-dimensional thought, and forming correlation prior information based on high dimensions; designing an iterative solution algorithm for rapidly extracting prior information of the network model; and applying an iterative solution algorithm to the trained network model, testing the image test data set, and outputting to obtain a final reconstruction result. According to the method, full-sampling data and under-sampling data are combined to serve as correlation-based high-dimensional prior information, multiple models are trained by utilizing different noise levels, and an unsupervised network is trained through under-sampling and artifact-containing images, and the trained network parameters are very close to a supervised network trained by using a full-sampling image.

Description

technical field [0001] The invention belongs to the field of medical image processing research, and specifically relates to a fast magnetic resonance imaging reconstruction algorithm based on high-dimensional correlation prior information, which is mainly used in medical image reconstruction. Background technique [0002] Magnetic resonance imaging, as the most important medical auxiliary means in the contemporary era, has gradually become an indispensable part of modern medical imaging technology (X-ray imaging, ultrasound imaging, CT and MRI, etc.), and has unparalleled advantages in other biological imaging technologies . Although the technology has undergone considerable development over the decades, it still suffers from a major bottleneck of slow imaging, ie long scan times. Because the imaging time is too long, it will lead to a series of problems, such as artifacts in reconstructed images caused by patients moving their limbs; in addition, for cardiac dynamic imagin...

AI Technical Summary

Problems solved by technology

Early research on fast magnetic resonance imaging mainly focused on three aspects: one is to improve the performance of magnetic resonance image reconstruction hardware, and to enhance the main magnetic field strength and gradient switching speed of the magnetic resonance scanner. It is easy to cause blurred images. At the same time, the high-speed switching magnetic field will generate eddy current field, which will cause special artifacts and have certain nerve electrical stimulation to human muscles.

The second is to use a parallel imaging algorithm, but the parallel imaging algorithm is limited by the accurate measurement of the sensitivity distribution of the coil and the influence of various factors such as self-calibration data rows, down-sampling multiples, and the number of fitting blocks, and often cannot accurately measure its sensitivity. distribution, and the performance of these reconstruction algorithms will degrade significantly when the SNR is very low or the accelerated acquisition factor is relatively large

The third is to reduce the amount of data acquisition in the K-space of magnetic resonance images, but a large reduction in the amount of data acquisition will lead to a significant decline in imaging quality. Although researchers can improve the quality of undersampled image reconstruction through various reconstruction algorithms, it often requires more Long reconstruction time, difficult to meet the clinical needs of real-time imaging

But since its adaptation to MRI images is still in its early stages, there is still room for improvement

Images