Multiple coil sensitivity maps of coils of a receiver array of a magnetic resonance imaging apparatus and sense reconstruction
Inventors
Benkert, Thomas • Dominik, Marcel Nickel
Assignees
Publication Number
US-12329507-B2
Publication Date
2025-06-17
Expiration Date
2043-04-13
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Abstract
Various examples relate to SENSitivity Encoding (SENSE) reconstruction of Magnetic Resonance Imaging (MRI) images. Multiple coil sensitivity maps per coil of a receiver coil array are used, e.g., obtained from an Eigenvalue-based Spatially Constrained Iterative Reconstruction Technique (ESPIRIT) autocalibration protocol.
Core Innovation
The invention relates to SENSitivity Encoding (SENSE) reconstruction of Magnetic Resonance Imaging (MRI) images using multiple coil sensitivity maps per coil of a receiver coil array. These multiple coil sensitivity maps are obtained, for example, from an Eigenvalue-based Spatially Constrained Iterative Reconstruction Technique (ESPIRIT) autocalibration protocol that densely samples a part of k-space. The method includes stitching together these multiple coil sensitivity maps along a given direction in k-space to form coil sensitivity data structures and acquiring undersampled k-space data that is rearranged on an adjusted k-space trajectory stretched by a stitching factor corresponding to the number of maps used per coil. An iterative optimization based on a SENSE reconstruction algorithm is performed using these structures to reconstruct MRI images, involving regularization and data-consistency operations.
The problem addressed is that conventional SENSE reconstruction algorithms, which typically use a single coil sensitivity map per coil, suffer from aliasing, wrap-around, and fold-over artifacts, particularly when reconstructing images from undersampled k-space data. While approaches like GRAPPA reduce some of these artifacts, they do not explicitly use coil sensitivity maps in image-domain reconstruction. Existing methods that employ multiple coil sensitivity maps per coil often require modifying the SENSE algorithm or retraining neural networks, which pose practical challenges.
This invention solves this by enabling the use of multiple coil sensitivity maps per coil, determined for example by the ESPIRIT autocalibration protocol, as input to an unmodified or minimally modified SENSE reconstruction algorithm. This is achieved by stitching the coil sensitivity maps and adjusting the k-space trajectory accordingly, allowing the exploitation of higher-rank coil sensitivity information while maintaining compatibility with conventional SENSE reconstruction. The approach further extends to advanced reconstruction tasks such as water-fat separation using Dixon methods by augmenting coil sensitivity data with phase and dephasing information.
Claims Coverage
The claims disclose two independent computer-implemented methods for reconstructing MRI images using multiple coil sensitivity maps and adjusted undersampling trajectories in a SENSE reconstruction framework. The inventive features focus on the determination, structuring, and use of coil sensitivity maps and measurement data structures, as well as the iterative optimization procedures.
Multiple coil sensitivity maps structured in a stitched arrangement
For each coil of a receiver coil array, a respective set of two or more coil sensitivity maps is determined using an ESPIRIT autocalibration protocol that densely samples part of the k-space. These maps are combined into a coil sensitivity data structure arranged in a stitched manner along a predetermined direction in k-space, with a stitching factor corresponding to the number of maps.
Adjusted undersampling k-space trajectory and measurement data rearrangement
For each coil, k-space data is acquired using an undersampling trajectory and rearranged on an adjusted k-space trajectory stretched by the stitching factor along the predetermined direction, resulting in measurement data structures compatible in dimension with the stitched coil sensitivity data.
SENSE reconstruction iterative optimization using stitched data
An iterative optimization based on an iterative SENSE reconstruction algorithm is performed using the coil sensitivity data structures and measurement data structures. The optimization includes a regularization operation and a data-consistency operation based on differences between measurement data and synthesized data from prior image estimates, adjusted trajectories, and coil sensitivities.
Reconstruction algorithm configured for multiple stitching factors and unstitched data
The reconstruction algorithm is designed to accept input data comprising coil sensitivity data and measurement data structures for a range of stitching factors and to handle non-stitched coil sensitivity maps and non-rearranged k-space data.
Extension for multi-echo acquisitions enabling water-fat separation
For k-space data acquired at multiple echo times, the coil sensitivity data structures are augmented with predetermined dephasing maps and factors associated with spin species dephasing, enabling multi-species image reconstruction using the same SENSE framework.
Properties of the undersampling trajectory and stitching direction
The undersampling trajectory includes sampling points arranged in a Cartesian pattern along the predetermined (e.g., phase-encoding or readout) direction, and may include non-Cartesian patterns in perpendicular directions such as stack-of-stars or stack-of-spirals. The stretching of k-space trajectory is symmetrical about an axis through the center of k-space.
Use of coil sensitivity maps including wrap-in information from outside field-of-view
At least one of the multiple coil sensitivity maps per coil includes wrap-in information from outside the predetermined field-of-view for which k-space data is acquired.
Alternative method using a single coil sensitivity map for a larger field-of-view
A method determining a single coil sensitivity map per coil using an autocalibration protocol for a first larger field-of-view, while acquiring k-space data for a second smaller field-of-view that is an integer fraction of the first. The k-space data is rearranged by stretching the undersampling trajectory by the inverse of the integer fraction for SENSE reconstruction.
The independent claims cover methods for MRI reconstruction using multiple coil sensitivity maps arranged in stitched data structures combined with k-space data rearranged via adjusted undersampling trajectories, enabling SENSE reconstruction through iterative optimization procedures. These techniques allow the use of existing SENSE reconstruction algorithms without major modification, extend to multi-echo acquisitions, and include embodiments for different field-of-view scaling.
Stated Advantages
Improved SENSE reconstruction by reducing aliasing and wrap-around artifacts through use of multiple coil sensitivity maps per coil.
Compatibility with conventional SENSE reconstruction algorithms without requiring modification or retraining for multiple coil sensitivity maps.
Capability to handle different stitching factors, allowing flexibility in the number of coil sensitivity maps used.
Extension to multi-echo acquisitions for water-fat separation using Dixon methods within the same reconstruction framework.
Efficient handling of undersampled k-space data acquired with varied sampling patterns, including Cartesian and non-Cartesian sampling.
Documented Applications
Reconstruction of MRI images from undersampled k-space data acquired using a receiver coil array in a Magnetic Resonance Imaging apparatus.
Water-fat separation in MRI using multi-echo acquisitions and Dixon methods.
Simultaneous multi-slice MRI reconstruction and three-dimensional or time-resolved MRI acquisition scenarios.
Interested in licensing this patent?