Dynamic MRI Reconstruction With Motion and Intensity Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing MRI reconstruction techniques struggle with motion corruption, particularly cardiac and respiratory motion, in dynamic imaging, leading to challenges in perfusion defect localization and image registration, especially in myocardial applications, due to irregular motion patterns and contrast changes, which are not adequately addressed by current regularization methods.
Innovation Solution
A method involving the acquisition of proton density weighted images before contrast agent injection, followed by motion estimation and correction, signal intensity correction, and registration with motion-corrected images, using spatial and temporal regularization to generate a dynamic series of motion-compensated magnetic resonance images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If very fast single-shot image acquisitions are used to capture dynamic tissue perfusion, then temporal resolution is improved, but motion corruption between acquisitions worsens
Solution Approach 1:
The method performs preliminary motion estimation and correction by registering each dynamically sampled image to a reference image before reconstruction. This preliminary alignment removes motion-induced misregistration artifacts, enabling subsequent high-speed acquisitions without motion corruption between frames
Solution Approach 2:
The method implements an iterative feedback loop where motion is estimated from dynamically sampled images, correction is applied to align images, and the process repeats. This feedback mechanism continuously refines motion compensation, allowing fast temporal sampling while maintaining image alignment across the dynamic series
2Manufacturing precision
If regularization through time is applied to use data redundancy across dynamic image series, then image resolution and morphologic coverage are improved, but performance deteriorates in the presence of motion
Solution Approach 1:
The method performs preliminary motion correction by registering all dynamically sampled images to a reference image before applying temporal regularization. This preliminary alignment ensures that subsequent regularization through time operates on motion-corrected data, allowing the exploitation of temporal redundancy without being degraded by motion artifacts
Solution Approach 2:
The reconstruction process is segmented into distinct stages: first motion estimation and correction, then temporal regularization. This segmentation separates the motion correction function from the regularization function, allowing each to optimize its specific task without interfering with the other, thereby maintaining both resolution and reliability
3Object-affected harmful factors
If breath holding is used for free-breathing acquisition with retrospective motion correction, then motion corruption is reduced, but achievable scan duration is limited
Solution Approach 1:
The method implements continuous feedback-based motion correction during free-breathing acquisition by iteratively estimating motion from dynamically sampled images and applying corrections. This real-time feedback mechanism allows the scan to continue throughout the entire breath cycle without interruption, achieving both motion correction and extended scan duration
Solution Approach 2:
The method performs preliminary motion estimation from the dynamically sampled images before final reconstruction. This preliminary motion characterization enables retrospective motion correction that can be applied to the entire free-breathing acquisition, removing the need to truncate scans to breath-hold durations
Data Source
Figure 1
Figure 2A~2C
Figure 3A~3B
AI summary
The invention discloses a computer-implemented method (100) of reconstructing a dynamic series of magnetic resonance images of a patient, comprising: acquiring (101, 103) first and second k-space data of the patient; reconstructing (104) at least one proton density weighted image based on the first k-space data; generating (105) a dynamic series of processing images based on the second k-space data and temporal regularization; applying (107) a motion correction to processing image based on an estimated motion; registering (107) the proton density weighted image to the motion corrected image; applying (108) a signal intensity correction to the dynamic series of motion corrected images based on the proton density weighted image.