Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-11921182-B2
Publication Date
2024-03-05
Expiration Date
Abstract
Systems and methods provide accelerated MR thermometry utilizing prior knowledge about the images to be reconstructed from incomplete k-space data, thereby facilitating accurate reconstruction. In various embodiments, missing data is computationally estimated using a machine learning algorithm such as a neural network, and an image is generated based on iteratively updated estimated missing information.
Core Innovation
The invention relates to accelerated magnetic resonance thermometry and other magnetic resonance reconstructions in MR-guided thermal therapies. It executes a first multi-echo pulse sequence to acquire fully sampled baseline k-space images of a target region and then executes a second multi-echo pulse sequence to acquire treatment k-space images in which at least one k-space image is under-sampled so that missing information is present compared to the baseline k-space images. A computation unit reconstructs a target image based at least in part on the baseline k-space images and the under-sampled second plurality of k-space images.
The reconstruction accelerates the treatment-time estimation of temperature/thermal maps and related map types by using prior knowledge tied to proton resonance frequency thermometry. Missing k-space data are iteratively estimated using constraints associated with proton resonance frequency thermometry, including the relationship between phase difference and echo time, and updates enforce data consistency by replacing reconstructed k-space with acquired samples while using the baseline multi-echo acquisitions to guide the estimation process.
The approach further supports extensions beyond thermometry and alternative missing-data estimation strategies. It optionally accounts for baseline magnitude/attenuation and other physical effects, and it optionally includes noise reduction using locally low-rank regularization. As an alternative or supplement to iterative estimation, missing data can be estimated using machine learning models such as neural networks configured to map incomplete k-space and baseline data to artifact-free k-space or directly to thermal maps, with loss functions that incorporate data consistency and, optionally, learned or physical k-trajectory correction.
Claims Coverage
The partial claim set provides two independent claims (clm-00001 and clm-00028). Both share the same core inventive structure: two multi-echo MRI acquisitions, a baseline set and a second treatment set with at least one under-sampled k-space image missing information, followed by reconstruction of a target image using baseline k-space data and the under-sampled second k-space data. Across the dependent content, additional inventive features specify echo-time-based association and update, optional neural-network-based estimation, specific target map modalities, and optional noise reduction and reconstruction refinements.
Two multi-echo acquisitions with under-sampled treatment k-space
An imaging system executes a first multi-echo pulse sequence to acquire a plurality of baseline k-space images associated with echoes, and executes a second multi-echo pulse sequence to acquire a second plurality of k-space images associated with echoes in which at least one second plurality k-space image is under-sampled so as to be missing information compared to the baseline k-space images.
Target image reconstruction using baseline and under-sampled second k-space
A computation unit reconstructs a target image based at least in part on at least one baseline k-space image and at least one of the second plurality of k-space images.
Reconstruction guided by echo time association
A computation unit generates a map using updated estimated missing information based at least in part on echo time, together with a second k-space image and a baseline k-space image, and can assign reconstructed images in a third image set to corresponding reconstructed baseline images based at least on the echo times of the associated echoes.
Neural network estimation for missing information
A computation unit runs an artificial neural network to estimate missing information or to perform reconstruction tasks in which incomplete k-space and baseline data are mapped to artifact-free k-space or directly to thermal maps.
Locally low-rank regularization for noise reduction
A computation unit applies locally low-rank regularization to at least two images in at least one of a third image set or a fifth image set to reduce noise therein.
Reconstructing specific target image modalities
The system is configured so that the target image is selected from a set of specified medical imaging types, including thermal maps, susceptibility-weighted images, T2* images, micro-bleeding images, and BOLD f-MRI images.
Across the independent claims, the coverage centers on accelerated reconstruction from two multi-echo MRI acquisitions where baseline k-space images provide reference information and at least one second-set k-space image is under-sampled with missing information. Dependent features further cover echo-time-associated update and mapping, optional artificial neural network-based estimation of missing information, selection of target map/modalities, and optional noise reduction using locally low-rank regularization.
Stated Advantages
Accelerated magnetic resonance thermometry during MR-guided thermal therapies by reconstructing target images using baseline k-space and under-sampled treatment k-space.
Documented Applications
MR-guided thermal therapies including MR-guided focused ultrasound (MRgFUS), using accelerated MR thermometry to reconstruct temperature/thermal maps.
Susceptibility-weighted imaging (SWI) reconstruction, including T2* mapping and micro-bleeding.
BOLD f-MRI reconstruction using similar multi-echo accelerated reconstruction.
Field map reconstruction using similar multi-echo accelerated reconstruction.
Multi-slice simultaneous excitation and other extensions beyond thermometry using similar multi-echo accelerated reconstruction.
Interested in licensing this patent?