DL-ESPIRiT Deep Neural Network for MRI Artifact Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Magnetic Resonance Imaging (MRI) acquisition processes are slow due to the large volume of data collected, and undersampling, while reducing scan times, often results in aliasing artifacts that obscure relevant anatomy, necessitating advanced reconstruction techniques to accelerate data collection without introducing artifacts.
Innovation Solution
A deep learning-based framework, DL-ESPIRiT, is employed to reconstruct MR images from undersampled k-space data using an extended coil sensitivity model, estimating multiple sets of coil sensitivity maps and iteratively reconstructing images with a trained deep neural network to reduce artifacts and computational effort.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If undersampling is used to reduce scan times, then productivity is improved, but image quality deteriorates due to aliasing artifacts
Solution Approach 1:
A deep neural network is introduced as an intermediary between the undersampled k-space data and the final image reconstruction. The network learns to map undersampled data to high-quality images by training on pairs of undersampled and fully-sampled data, effectively mediating the reconstruction process to produce artifact-free images from accelerated scans
Solution Approach 2:
The deep neural network is trained in advance using pairs of undersampled and fully-sampled k-space data before actual imaging. This preliminary training phase allows the network to learn the complex mapping relationships and reconstruction patterns, so that during actual scanning, high-quality images can be rapidly reconstructed from undersampled data without artifacts
2Manufacturing precision
If advanced reconstruction techniques are used to maintain image quality, then manufacturing precision is improved, but device complexity increases
Solution Approach 1:
Traditional iterative mathematical reconstruction algorithms are replaced with a deep neural network-based approach. The complex mathematical optimization processes are substituted with a trained neural network that performs reconstruction through learned patterns, simplifying the computational workflow while maintaining or improving image quality
Solution Approach 2:
The deep neural network creates a learned model (copy of the reconstruction knowledge) from training data that can be applied repeatedly to new undersampled data. Instead of performing complex iterative calculations for each reconstruction, the pre-trained network model is applied directly, reducing computational complexity during actual imaging while preserving reconstruction quality
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 DL-ESPIRiT method effectively reduces imaging artifacts and accelerates scan times by generating high-quality MR images with reduced aliasing, enabling more accurate diagnoses and efficient MRI procedures.
Implementation Method 1
iteratively reconstructing, with a trained deep neural network, multiple images by using the initial images and the multiple sets of coil sensitivity maps to generate multiple final images
Data Source
AI summary
Various methods and systems are provided for reconstructing magnetic resonance images from accelerated magnetic resonance imaging (MM) data. In one embodiment, a method for reconstructing a magnetic resonance (MR) image includes: estimating multiple sets of coil sensitivity maps from undersampled k-space data, the undersampled k-space data acquired by a multi-coil radio frequency (RF) receiver array; reconstructing multiple initial images using the undersampled k-space data and the estimated multiple sets of coil sensitivity maps; iteratively reconstructing, with a trained deep neural network, multiple images by using the initial images and the multiple sets of coil sensitivity maps to generate multiple final images, each of the multiple images corresponding to a different set of the multiple sets of sensitivity maps; and combining the multiple final images output from the trained deep neural network to generate the MR image.