Cine MRF Reconstruction with Neural Networks for Cardiac Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cardiac magnetic resonance imaging (MRI) techniques face challenges in efficiently acquiring co-registered maps of multiple tissue properties within a single acquisition, due to limitations such as low scan efficiency, mis-registration between maps, and sensitivity to confounding factors.
Innovation Solution
The implementation of a cine magnetic resonance fingerprinting (MRF) reconstruction system that employs an image reconstruction network (IRN) and a parameter estimation network (PEN), both of which are trained de novo, to generate cardiac phase-resolved maps without motion correction, and estimate an effective B1+ map to reduce errors from RF transmit inhomogeneities and motion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional cardiac MRI techniques are used to acquire multiple tissue property maps, then comprehensive tissue characterization can be achieved, but scan efficiency is low and mis-registration between maps occurs
Solution Approach 1:
The patent combines multiple tissue property mapping sequences (T1, T2, T2*) into a single integrated cine MRF acquisition protocol. This merging allows simultaneous collection of data for multiple tissue parameters during one scan, eliminating the need for separate acquisitions and preventing mis-registration between maps while maintaining comprehensive tissue characterization.
Solution Approach 2:
The MRF sequence serves multiple functions by encoding different tissue properties (T1, T2, T2*) within a single acquisition framework. The universal MRF approach allows extraction of multiple tissue parameters from one scan, improving scan efficiency while maintaining the precision needed for comprehensive tissue characterization.
2Measurement precision
If multiple separate MRI acquisitions are performed for different tissue properties, then detailed tissue characterization is achieved, but mis-registration between maps occurs
Solution Approach 1:
By merging multiple tissue property acquisitions into a single cine MRF scan, the patent ensures that all tissue maps are acquired simultaneously during the same breathhold and cardiac cycle phases. This eliminates temporal separation between acquisitions, preventing mis-registration and maintaining stable spatial alignment across different tissue property maps.
3Measurement precision
If conventional MRI sequences are used for tissue mapping, then standard tissue properties can be measured, but sensitivity to confounding factors such as motion and RF inhomogeneities remains high
Solution Approach 1:
The patent employs feedback mechanisms through iterative reconstruction algorithms that incorporate motion estimation and correction. The system continuously refines tissue parameter estimates by comparing acquired signals with simulated signals, adjusting for motion artifacts and RF inhomogeneities to improve measurement precision and reduce sensitivity to confounding factors.
Solution Approach 2:
The MRF sequence utilizes parameter changes by varying radiofrequency pulse flip angles and timing parameters throughout the acquisition. This dynamic parameter modulation creates unique signal evolutions that are less sensitive to motion and RF inhomogeneities, improving tissue parameter accuracy while reducing vulnerability to confounding factors.
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
This approach enables improved image quality and precision in cardiac T1, T2, and M0 mapping, as well as synthetic bright-blood and dark-blood cine imaging, during a short breathhold, thereby overcoming the limitations of previous techniques.
Implementation Method 1
MRI uses the nuclear magnetic resonance (NMR) phenomenon to produce images. When a substance such as human tissue is subjected to a uniform magnetic field (B0), the individual magnetic moments of the nuclei in the tissue attempt to align with this magnetic field, but precess about the field in random order at their characteristic Larmor frequency.
Implementation Method 2
When a substance such as human tissue is subjected to a uniform magnetic field (B0), the individual magnetic moments of the nuclei in the tissue attempt to align with this magnetic field
Implementation Method 3
If the tissue is subjected to an excitation magnetic field (B1) that is in the x-y plane and that is near the Larmor frequency, the net aligned moment may be rotated, or 'tipped,' into the x-y plane to produce a net transverse magnetic moment.
Implementation Method 4
A signal is emitted by the excited nuclei or 'spins,' after the excitation signal B1 is terminated, and this signal may be received and processed to form an image.
Data Source
AI summary
Methods and systems are provided for cine magnetic resonance fingerprinting (MRF). In one example, a method includes obtaining k-space data of an MRF scan of a subject, the k-space data acquired over a plurality of phases of at least one cardiac cycle of the subject, training an image reconstruction network (IRN) to output, for each phase, one or more subspace images of the subject using the k-space data, and training a parameter estimation network (PEN) to output, for each phase, a set of tissue parameter maps of the subject using the one or more subspace images output by the IRN for the corresponding phase. Upon training the IRN and the PEN, the method further includes obtaining (and displaying and/or saving in memory) a final set of tissue parameter maps of the subject for one or more or each of the plurality of phases.