MRI Interventional Image Reconstruction Method Based on Recurrent Neural Network

A cyclic neural network and image reconstruction technology, applied in the field of image processing, can solve the problems of low resolution of image reconstruction, poor image quality, slow image reconstruction speed of compressed sensing method, etc., to achieve fast acquisition and reconstruction, and good reconstruction effect. , the effect of high downsampling rate

Active Publication Date: 2022-03-15
SHANGHAI JIAOTONG UNIV
View PDF6 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] However, the shortcomings of the above-mentioned technologies include that the acceleration factor of conventional parallel imaging and short TR acquisition methods cannot meet the requirements of real-time imaging, the image reconstruction resolution of partial k-space acquisition and key-hole methods is low, the image quality is not good, and non-Cartesian coordinates In the case of high downsampling, the reconstructed image quality of the K-space acquisition system is poor, the image reconstruction speed of the compressed sensing method is slow, and cannot meet the real-time requirements, and most of the machine learning-based algorithms are used in the reconstruction of magnetic resonance structural images Meet the requirements for fast reconstruction of interventional images

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • MRI Interventional Image Reconstruction Method Based on Recurrent Neural Network
  • MRI Interventional Image Reconstruction Method Based on Recurrent Neural Network
  • MRI Interventional Image Reconstruction Method Based on Recurrent Neural Network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0014] Due to the undersampling acquisition signal y=F in MRI image reconstruction u x+e, where: x is a fully sampled image, F u is the downsampled Fourier encoding matrix, e is the noise at the time of acquisition, so recovering x from y is an ill-posed problem, and reconstruction by zero-padding y will result in an aliased undersampled image x unI .

[0015] According to prior knowledge, between different under-sampled images, because the background information is related to each other, there is temporal consistency, and the neural network based on convolution-long short-term memory module (Conv-LSTM) Networks can take advantage of temporal consistency. Therefore, the cyclic neural network adopted in this embodiment reconstructs the undersampled magnetic resonance images of five consecutive frames, and the input of the trained cyclic neural network is the fully sampled preoperative reference image x ref and five consecutive frames of undersampled images The output is th...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

A MRI interventional image reconstruction method based on a recurrent neural network, which uses a recurrent neural network to reconstruct five consecutive frames of subsampled magnetic resonance images, and the input of the trained recurrent neural network is a fully sampled preoperative reference image and continuous Five frames of undersampled images, output as reconstructed images. The invention can realize fast acquisition and real-time reconstruction of images, and obtain images with quality satisfying navigation requirements.

Description

technical field [0001] The present invention relates to a technology in the field of image processing, in particular to a method for reconstructing images of magnetic resonance intervention based on a cyclic neural network. Background technique [0002] Magnetic resonance images have good soft tissue contrast and various imaging methods, which provide an important way for current image-guided surgery and interventional operations. However, the acquisition time of magnetic resonance imaging is long, and it is difficult to carry out real-time image acquisition and image reconstruction in the process of intraoperative operation. Existing MRI fast acquisition and reconstruction methods include: parallel imaging method and short TR acquisition method, partial k-space acquisition, such as key-hole method, non-Cartesian coordinate system k-space acquisition, such as radial acquisition and helical acquisition method, Compressed sensing methods and machine learning methods. [0003...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & AuthorityPatents(China)
IPC IPC(8): G06T11/00G06T5/10G06N3/04G06N3/08
CPCG06T11/005G06T11/006G06T5/10G06N3/08G06T2207/10088G06T2207/20056G06T2207/20081G06T2207/20084G06N3/045
Inventor冯原赵睿洋杜一平梁志培
OwnerSHANGHAI JIAOTONG UNIV