# US 11,079,453 B2 · System and method for ultrafast magnetic resonance spectroscopic imaging using learned spectral features

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 11,079,453 B2
- Document ID: US11079453B2
- Status: Granted
- Kind: B2
- Application number: 16/643,239
- Filed: 2018-08-23
- Published: 2020-12-31
- Granted: 2021-08-03
- Earliest claimed: 2017-08-30
- Assignees: The Board of Trustees of the University of Illinois
- Applicants: The Board of Trustees of the University of Illinois
- Inventors: Zhi-Pei Liang; Fan Lam
- Web page: https://patentgrep.com/patent/US11079453B2

## CPC

- G01R 33/485: based on chemical shift information {[CSI] or spectroscopic imaging, e.g. to acquire the spatial distributions of metabolites}
- A61B 5/0042: {for the brain}
- A61B 5/055: involving electronic [EMR] or nuclear [NMR] magnetic resonance, e.g. magnetic resonance imaging
- A61B 5/7253: {characterised by using transforms}
- G01R 33/246: {Spatial mapping of the RF magnetic field B1}
- G01R 33/34092: {RF coils specially adapted for NMR spectrometers}
- G01R 33/4625: {Processing of acquired signals, e.g. elimination of phase errors, baseline fitting, chemometric analysis}
- G01R 33/465: applied to biological material, e.g. in vitro testing
- G01R 33/4816: {NMR imaging of samples with ultrashort relaxation times such as solid samples, e.g. MRI using ultrashort TE [UTE], single point imaging, constant time imaging}
- G01R 33/4826: {in three dimensions}
- G01R 33/4828: {Resolving the MR signals of different chemical species, e.g. water-fat imaging}
- G01R 33/5608: {Data processing and visualization specially adapted for MR, e.g. for feature analysis and pattern recognition on the basis of measured MR data, segmentation of measured MR data, edge contour detection on the basis of measured MR data, for enhancing measured MR data in terms of signal-to-noise ratio by means of noise filtering or apodization, for enhancing measured MR data in terms of resolution by means for deblurring, windowing, zero filling, or generation of gray-scaled images, colour-coded images or images displaying vectors instead of pixels (image data processing or generation, in general G06T)}
- G01R 33/5615: {Echo train techniques involving acquiring plural, differently encoded, echo signals after one RF excitation, e.g. using gradient refocusing in echo planar imaging [EPI], RF refocusing in rapid acquisition with relaxation enhancement [RARE] or using both RF and gradient refocusing in gradient and spin echo imaging [GRASE]}
- G01R 33/5616: {using gradient refocusing, e.g. EPI}
- G01R: MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES (indicating correct tuning of resonant circuits H03J3/12)
- A61B: DIAGNOSIS; SURGERY; IDENTIFICATION (usefulness limited to only animals A61D)

## Abstract

A new method is developed for ultrafast, high-resolution magnetic resonance spectroscopic imaging (MRSI) using learned spectral features. The method uses Free Induction Decay (FID) based ultrashort-TE and short-TR acquisition without any solvent suppression pulses to generate the desired spatiospectral encodings. The spectral features for the desired molecules are learned from specifically designed “training” data by taking into account the resonance structure of each compound generated by quantum mechanical simulations. A union-of-subspaces model that incorporates the learned spectral features is used to effectively separate the unsuppressed water/lipid signals, the metabolite signals, and the macromolecule signals. The unsuppressed water spectroscopic signals in the data can be used for various purposes, e.g., removing the need of additional auxiliary scans for calibration, and for generating high quality quantitative tissue susceptiability mapping etc. Simultaneous spatiospectral reconstructions of water, lipids, metabolite and macromolecule can be obtained using a single 1H-MRSI scan.

## Claims

### Claim 1

A device to acquire spatiospectral distributions from an object, the device comprising:
a magnetic field generator that generates a static magnetic field, the object being positioned in the static magnetic field;
an RF coil assembly including a plurality of gradient coils;
an RF transceiver controlled by a pulse module that transmits RF signals to the RF coil assembly and that receives magnetic resonance (MR) signals; and
a computer which, responsive to executing instructions, performs operations, the operations comprising:
acquiring, based upon the MR signals, spatiospectrally encoded free induction decays (FIDs) that sparsely sample a (k,t)-space with ultrashort TE and short TR, wherein the spatiospectrally encoded FIDs are acquired without water suppression and without lipid suppression;
acquiring, based upon the MR signals, embedded navigator data for tracking system instability and for tracking motion of the object, wherein the navigator data are acquired simultaneously with the spatiospectrally encoded FIDs;
reconstructing spatiospectral functions from the spatiospectrally encoded FIDs using spectral features learned from training data acquired experimentally and generated using quantum mechanical simulations; and
separating, based upon the MR signals, signals from unsuppressed water molecules, signals from lipid molecules, signals from metabolites, and signals from macromolecules;
wherein one or more quantitative tissue susceptibility maps, one or more tissue relaxation constant maps, one or more metabolite maps, and one or more macromolecule maps can be obtained from a single MRSI data set that is based upon the MR signals.

### Claim 2 (depends on 1)

The device of claim 1, wherein the training data are acquired to learn the spectral features of different molecules in the object, wherein the spectral features that are learned are used to construct subspace representations of individual molecules and wherein the spectral features that are learned are used for general MRSI experiments.

### Claim 3 (depends on 2)

The device of claim 2, wherein the training data are acquired using: one or more single voxel spectroscopy scans; one or more ultrashort-TE, 2D or 3D chemical shift imaging sequences; or one or more ultrashort-TE, low-resolution, 2D or 3D echo-planar spectroscopic imaging sequences.

### Claim 4 (depends on 1)

The device of claim 1, wherein a FID-based, ultrashort-TE, very-short-TR acquisition is used to generate the spatiospectrally encoded FIDs, wherein the ultrashort-TE is less than 4 ms, wherein the very-short-TR is less than 300 ms, wherein power of the RF transceiver is optimized for a signal-to-noise-ratio (SNR) of a metabolite spatiospectral function, and wherein neither water nor lipid suppression is used.

### Claim 5 (depends on 4)

The device of claim 4, wherein the spatiospectral encoding strategy uses different gradient waveforms to generate (k,t)-space sampling trajectories, wherein the (k,t)-space sampling trajectories comprise one or more Cartesian trajectories, one or more echo-planar trajectories, or one or more other non-Cartesian trajectories, and wherein the non-Cartesian trajectories comprises a spiral trajectory, a radial trajectory, or a concentric ring trajectory.

### Claim 6 (depends on 4)

The device of claim 4, wherein the spatiospectral encoding strategy uses a corresponding (k,t)-space, with k denoting three spatial frequency dimensions and t denoting an additional temporal dimension, that is sampled sparsely in variable density, wherein a center k-space region is sampled with Nyquist density and the rest of the k-space is sampled below Nyquist density, and wherein the temporal dimension is sampled at or below Nyquist density.

### Claim 7 (depends on 4)

The device of claim 4, wherein the acquisition uses a spatiospectral encoding strategy in which acquisition of additional navigator signals with circular trajectories or linear trajectories is interleaved into a total acquisition period.

### Claim 8 (depends on 1)

The device of claim 1, wherein the learned spectral features of NMR-detectable molecules are used for processing acquired (k, t)-space data to obtain desired spatiospectral distributions.

### Claim 9 (depends on 1)

The device of claim 1, wherein a union-of-subspaces model is used to represent and separate the spatiospectral distributions originating from different types of molecules in the object, including the water molecules, the lipid molecules, the metabolites, and the macromolecules.

### Claim 10 (depends on 9)

The device of claim 9, wherein for 1H-MRSI, the union-of-subspaces model is expressed as:
ρ
⁡
(
r
,
t
)
=
∑
l
w
=
1
L
w
⁢
c
l
w
⁡
(
r
)
⁢
ϕ
l
w
⁡
(
t
)
+
∑
l
f
=
1
L
f
⁢
c
l
f
⁡
(
r
)
⁢
ϕ
l
f
⁡
(
t
)
+
∑
l
m
=
1
L
m
⁢
c
l
m
⁡
(
r
)
⁢
ϕ
l
m
⁡
(
t
)
+
∑
l
b
=
1
L
b
⁢
c
l
b
⁡
(
r
)
⁢
ϕ
l
b
⁡
(
t
)
,
where r denotes the spatial dimensions, the partially separable functions on the right hand side of the equation represent the water component, lipid component, metabolite component and macromolecule baseline component, respectively, and Lx is the model order for an individual component, wherein x denotes w, f, m, or b.

### Claim 11 (depends on 10)

The device of claim 10, wherein basis functions {ϕlx(t)}, are determined for individual signal components based on the spectral features learned from the training data acquired experimentally and generated using quantum mechanical simulations.

### Claim 12 (depends on 4)

The device of claim 4, wherein the acquisition uses a spatiospectral encoding strategy in which unsuppressed/companion water spectroscopic signals are used to determine B0 field inhomogeneity and coil receiver sensitivity profiles, to estimate and correct B0 field drifts, and to track and correct for head motion during a scan, without a need for acquiring additional auxiliary data for the same purposes.

### Claim 13 (depends on 12)

The device of claim 12, wherein the quantitative tissue susceptibility maps are extracted from B0 field maps estimated from the unsuppressed water spectroscopic signals.

### Claim 14 (depends on 12)

The device of claim 12, wherein quantitative T2* maps are obtained from the unsuppressed water spectroscopic signals.

### Claim 15 (depends on 12)

The device of claim 12, where B0 field maps and T2* maps are determined using all the temporal samples, or echoes, available.

### Claim 16 (depends on 4)

The device of claim 4, wherein the acquisition uses a spatiospectral encoding strategy in which unsuppressed water and lipid signals are removed from sparsely sampled data using a union-of-subspaces model that is expressed as:
ρ
⁡
(
r
,
t
)
=
∑
l
w
=
1
L
w
⁢
c
l
w
⁡
(
r
)
⁢
ϕ
l
w
⁡
(
t
)
+
∑
l
f
=
1
L
f
⁢
c
l
f
⁡
(
r
)
⁢
ϕ
l
f
⁡
(
t
)
+
∑
l
m
=
1
L
m
⁢
c
l
m
⁡
(
r
)
⁢
ϕ
l
m
⁡
(
t
)
+
∑
l
b
=
1
L
b
⁢
c
l
b
⁡
(
r
)
⁢
ϕ
l
b
⁡
(
t
)
,
where r denotes the spatial dimensions, the partially separable functions on the right hand side of the equation represent the water component, lipid component, metabolite component and macromolecule baseline component, respectively, and Lx is the model order for an individual component, wherein x denotes w, f, m, or b,
and wherein metabolite and macromolecule spatiospectral distributions are reconstructed and separated from water/lipid-removed MRSI data.

### Claim 17 (depends on 16)

The device of claim 16, wherein metabolite and macromolecule reconstructions are determined by solving the following optimization problem
C
^
m
,
C
^
b
=
arg
⁢
⁢
min
C
m
,
C
b
⁢

d
-
Ω
⁢
{
FB
⊙
(
C
m
⁢
Ψ
m
+
C
b
⁢
Ψ
b
)
}

2
2
+
λ
m
⁢
R
m
⁡
(
C
m
,
Ψ
m
)
+
λ
b
⁢
R
b
⁡
(
C
b
,
Ψ
b
)
,
where d is a vector containing all the data with water and lipid signals removed, 12 denotes a (k,t)-space sampling operator, F denotes a Fourier transform operator, B models the B0 inhomogeneity effects including spatially dependent frequency shifts and experiment-dependent spectral distortions, ψm and ψb are matrix representations of the metabolite basis and macromolecule basis, respectively, Cm and Cb are matrix representations of {clm(r)} and {clb(r)}, respectively, the ∥⋅∥22 term measures data consistency, Rm(⋅) and Rb(⋅) impose spatial regularization on the desired spatiospectral reconstruction, and λm and λb are the regularization parameters.

### Claim 18 (depends on 17)

The device of claim 17, where Rm(⋅) and Rb(⋅) are quadratic regularization that enforces edge-preserving smoothness or non-quadratic regularization that enforces sparsity constraints on spatiospectral variations of individual signal components.

### Claim 19

A method of acquiring spatiospectral distributions from an object, the method comprising:
generating a static magnetic field, the object being positioned in the static magnetic field;
controlling an RF transceiver to transmit RF signals to an RF coil assembly including a plurality of gradient coils;
controlling the RF transceiver to receive imaging data;
acquiring, based upon the imaging data, spatiospectrally encoded free induction decays (FIDs) that sparsely sample a (k,t)-space with ultrashort TE and short TR;
acquiring, based upon the imaging data, embedded navigator data for tracking system instability and for tracking motion of the object, wherein the navigator data are acquired simultaneously with the spatiospectrally encoded FIDs;
reconstructing spatiospectral functions from the spatiospectrally encoded FIDs using spectral features learned from training data acquired experimentally and generated using quantum mechanical simulations;
separating, based upon the imaging data, spatiospectral distributions of water, lipids, metabolites and macromolecules; and
generating one or more quantitative tissue susceptibility maps, one or more T2* maps, one or more B0 inhomogeneity maps, and one or more maps of various metabolites and macromolecules from a single MRSI data set that is based upon the imaging data;
wherein the ultrashort TE is less than 4 ms and the short TR is less than 300 ms, and wherein the method does not require a water suppression pulse or a lipid suppression pulse for 1H-MRSI.

### Claim 20

A non-transitory computer-readable storage device comprising executable instructions which, responsive to being executed by a processor, cause the processor to perform operations, the operations comprising:
acquiring spatiospectrally encoded FID data covering (k, t)-space in variable density and variable SNR with ultrashort-TE and short-TR, without water and lipid suppression, suitable for image reconstruction using a union-of-subspaces model or a linear combination of low-rank matrix/tensor models; and
reconstructing a spatiospectral function from the spatiospectrally encoded FID data that is sparsely sampled using the union-of-subspaces model or the linear combination of low-rank matrix/tensor models;
wherein the acquiring of the spatiospectrally encoded FID data is performed utilizing an RF coil assembly including a plurality of gradient coils, and an RF transceiver controlled by a pulse module to transmit RF signals to the RF coil assembly, and
wherein an ultrashort TE less than 4 ms and a short TR less than 300 ms are utilized and no water suppression pulse is used and no lipid suppression pulse is used.