Neural Network Motion Correction for MRI Artefacts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current motion correction techniques in MRI, such as external devices, k-space navigators, and image navigators, are either expensive, complex to set up, or interfere with MR sequence parameters, leading to low signal-to-noise ratio and increased scan time, making them unsuitable for clinical use.
Innovation Solution
A method and system using a neural network to determine corrected intensity values for each pixel in motion-corrupted MR images, generating motion-corrected images without requiring external sensors or modifications to the MR sequence, employing a convolutional neural network for accurate pixel classification or regression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If external devices such as cameras or magnetic probes are used to detect and correct patient motion, then motion correction capability is improved, but device cost and setup complexity increase
Solution Approach 1:
The system uses the MRI scanner's own gradient coils to generate motion correction fields, eliminating the need for external devices. The gradient coils serve dual purposes: primary imaging function and motion correction function, making the system self-sufficient and reducing external dependencies
Solution Approach 2:
The gradient coils are designed to perform multiple functions: both primary imaging and motion correction. This multi-functionality reduces the need for separate dedicated motion correction hardware, simplifying the overall system while maintaining correction capabilities
2Reliability
If k-space navigators or image navigators are integrated into the MR sequence, then motion correction is achieved, but scan time increases and signal-to-noise ratio decreases
Solution Approach 1:
Motion correction is applied continuously throughout the scan using gradient fields that are active during the entire data acquisition process, rather than using discrete navigator echoes that interrupt the scan. This continuous approach maintains scan efficiency while providing ongoing motion correction
Solution Approach 2:
The motion correction functionality is extracted from the traditional navigator echo approach and implemented through gradient field modulation. This separates the motion correction function from the imaging sequence, allowing correction without adding navigator-related time penalties or SNR losses
3Reliability
If k-space navigators or image navigators are used, then motion information is obtained, but compatibility with all imaging sequences is reduced
Solution Approach 1:
The gradient coils, which are standard components in all MRI systems, are utilized for motion correction across all imaging sequences. This universal approach maintains compatibility with diverse sequences while providing consistent motion correction functionality
Solution Approach 2:
The motion correction is implemented as a separate, modular gradient field application that can be independently integrated into different imaging sequences. This segmentation allows the correction mechanism to work universally across T1-weighted, T2-weighted, diffusion, and other sequence types without requiring sequence-specific modifications
Data Source
AI summary
A method and system for reducing or removing motion artefacts in magnetic resonance (MR) images, the method including the steps of: receiving a motion corrupted MR image; determining a corrected intensity value for each pixel in the motion corrupted MR image by using a neural network; and generating a motion corrected MR image based on the determined corrected intensity values for the pixels in the motion corrupted MR image.