Cascaded Recurrent Neural Network for MRI Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for reconstructing magnetic resonance images are time-consuming and often compromise on image quality, leading to long acquisition times and potentially distorted or blurred images that may hinder medical diagnosis.
Innovation Solution
A system and method utilizing a trained cascaded recurrent neural network to process sub-sampled k-space data, applying an inverse fast Fourier transform to generate preliminary images, which are then reconstructed to produce high-quality magnetic resonance images in shorter times without compromising image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If compressed sensing or deep-learning techniques are used to reconstruct magnetic resonance images from under-sampled data, then image acquisition time is reduced, but image quality deteriorates with artifacts, blurring, or distortion
Solution Approach 1:
The patent divides the reconstruction process into multiple iterative stages, where each stage refines the image quality by processing the output of the previous stage. This segmented approach allows progressive improvement of image quality while maintaining reduced acquisition time benefits.
Solution Approach 2:
The patent implements a feedback mechanism where the reconstructed image from each iteration is fed back into the next iteration for further refinement. This feedback loop continuously improves image quality by correcting artifacts and distortions while preserving the time efficiency gained from under-sampled data.
2Manufacturing precision
If traditional iterative algorithms are used to reconstruct magnetic resonance images, then image quality is maintained, but image acquisition time increases considerably
Solution Approach 1:
The patent performs preliminary reconstruction using under-sampled data to generate an initial image estimate before applying iterative refinement. This preliminary action provides a starting point that reduces the computational burden of subsequent iterative processing, thereby maintaining image quality while reducing overall acquisition time.
Solution Approach 2:
The patent applies a limited number of iterative refinement stages rather than exhaustive processing, providing just enough refinement to achieve acceptable image quality without the full time cost of traditional iterative algorithms. This partial action approach balances quality and time efficiency.
3Productivity
If sub-sampled k-space data is used for reconstruction, then scanning time is reduced, but fine image detail becomes invisible or distorted
Solution Approach 1:
The patent introduces an intermediary processing stage that acts as a bridge between the under-sampled data and the final high-quality image. This intermediary refinement process recovers fine image details that would otherwise be lost, while preserving the time efficiency benefits of using sub-sampled data.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The approach significantly reduces image acquisition time while maintaining high image quality, comparable to images reconstructed from fully-sampled data, thus overcoming the limitations of existing methods.
Implementation Method 1
apply an inverse fast fourier transform to the sub-sampled k-space data to generate a preliminary image
Data Source
AI summary
A system for reconstructing magnetic resonance images includes a processor that is configured to obtain, from a magnetic resonance scanner, sub-sampled k-space data; apply an inverse fast fourier transform to the sub-sampled k-space data to generate a preliminary image; and process the preliminary image via a trained cascaded recurrent neural network to reconstruct a magnetic resonance image.