# US 2024/0197196 A1 · Method and system for high-resolution multi-parametric quantitative magnetic resonance imaging

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

- Number: US 2024/0197196 A1
- Document ID: US20240197196A1
- Status: Published application
- Kind: A1
- Application number: 18/203,712
- Filed: 2023-05-31
- Published: 2024-06-20
- Earliest claimed: 2022-12-15
- Assignees: Xiamen University
- Applicants: Xiamen University
- Inventors: Shuhui CAI; Wenhua GENG; Qizhi YANG; Congbo CAI; Zhong CHEN
- Web page: https://patentgrep.com/patent/US20240197196A1

## CPC

- A61B 5/055: involving electronic [EMR] or nuclear [NMR] magnetic resonance, e.g. magnetic resonance imaging
- G01R 33/4835: {of multiple slices}
- G01R 33/54: Signal processing systems, e.g. using pulse sequences {; Generation or control of pulse sequences; Operator console}
- 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 33/5611: {Parallel magnetic resonance imaging, e.g. sensitivity encoding [SENSE], simultaneous acquisition of spatial harmonics [SMASH], unaliasing by Fourier encoding of the overlaps using the temporal dimension [UNFOLD], k-t-broad-use linear acquisition speed-up technique [k-t-BLAST], k-t-SENSE (structural details of arrays of sub-coils G01R33/3415)}
- G01R 33/5617: {using RF refocusing, e.g. RARE}
- G01R 33/565: Correction of image distortions, e.g. due to magnetic field inhomogeneities
- G06T 5/10: using non-spatial domain filtering
- G06T 7/0012: {Biomedical image inspection}
- G06V 10/82: using neural networks
- G06T 2207/10088: Magnetic resonance imaging [MRI]
- Y02A 90/30: Assessment of water resources
- A61B: DIAGNOSIS; SURGERY; IDENTIFICATION (usefulness limited to only animals A61D)
- G01R: MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES (indicating correct tuning of resonant circuits H03J3/12)
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- Y02A: TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE

## Abstract

A method for fast high-resolution multi-parametric quantitative magnetic resonance imaging comprises: designing a fast high-resolution multiple overlapping-echo imaging pulse sequence; determining sampling parameters of the pulse sequence; constructing a deep neural network for reconstructing high-resolution multi-parametric quantitative magnetic resonance images; generating training samples of the deep neural network; using the training samples to train the deep neural network to obtain trained deep neural networks; scanning a real imaging object using the pulse sequence under the sampling parameters to obtain k-space data of the real imaging object; pre-processing the k-space data of the real imaging object to obtain image domain data of the real imaging object; and inputting the image domain data of the real imaging object into the trained deep neural networks for the reconstructing to obtain the high-resolution multi-parametric quantitative magnetic resonance images of the real imaging object.

## Claims

### Claim 1

A method for fast high-resolution multi-parametric quantitative magnetic resonance imaging, comprising:
designing a fast high-resolution multiple overlapping-echo imaging pulse sequence;
determining sampling parameters of the fast high-resolution multiple overlapping-echo imaging pulse sequence;
constructing a deep neural network for reconstructing high-resolution multi-parametric quantitative magnetic resonance images;
generating training samples of the deep neural network;
using the training samples to train the deep neural network to obtain trained deep neural networks;
scanning a real imaging object using the fast high-resolution multiple overlapping-echo imaging pulse sequence under the sampling parameters to obtain first k-space data of the real imaging object;
pre-processing the first k-space data of the real imaging object to obtain image domain data of the real imaging object; and
inputting the image domain data of the real imaging object into the trained deep neural networks for the reconstructing to obtain the high-resolution multi-parametric quantitative magnetic resonance images of the real imaging object, wherein:
the fast high-resolution multiple overlapping-echo imaging pulse sequence comprises a signal excitation module and a data acquisition module,
the signal excitation module comprises N radio frequency (RF) excitation pulses with time intervals tn and flip angles αn, slice selection gradients Gss corresponding to the N RF excitation pulses, and echo shift gradients Gn,
n=1,2, . . . ,N, and N≥2,
each of the N RF excitation pulses is combined with a corresponding one of the slice selection gradients Gss in a slice selection dimension for slice selection,
the echo shift gradients Gn are applied after a corresponding RF excitation pulse of the N RF excitation pulses along a frequency encoding dimension and a phase encoding dimension,
the data acquisition module comprises a pre-phase gradient Gpre, M refocusing pulses with time intervals ESP and flip angles βm, crushing gradients Ger, phase encoding gradients Gpe,i,m, frequency encoding gradients Gro, and dephase encoding gradients Gpe,i,m′,
i represents an ith scanning,
m=1,2, . . . ,M, and M≥2,
the pre-phase gradient Gpre is applied along the frequency encoding dimension, and a size of the pre-phase gradient Gpre is half of a corresponding one of the frequency encoding gradients Gro,
the crushing gradients Ger are applied before and after each refocusing pulse of the M refocusing pulses, and sizes and directions of the crushing gradients Ger along the frequency encoding dimension, the phase encoding dimension, and the slice selection dimension are the same, and
the phase encoding gradients Gpe,i,m, the frequency encoding gradients Gro, and the dephase encoding gradients Gpe,i,m′ are applied after each refocusing pulse of the M refocusing pulses, and the phase encoding gradients Gpe,i,m and the dephase encoding gradients Gpe,i,m′ have same sizes but opposite directions.

### Claim 2 (depends on 1)

The method according to claim 1, wherein:
the determining sampling parameters of the fast high-resolution multiple overlapping-echo imaging pulse sequence comprises:
determining a number N of the N RF excitation pulses and a flip angle αn of each RF excitation pulse of the N RF excitation pulses;
determining a composition of echo signals to determine a time interval tn between two neighboring RF excitation pulses of the N RF excitation pulses;
determining a position of each of the echo signals in a k-space so as to determine a proportion of the echo shift gradients Gn after each RF excitation pulse of the N RF excitation pulses;
determining a number M of the M refocusing pulses and a flip angle βm of each refocused pulse of the M refocusing pulses;
determining a phase encoding manner to determine a direction and a proportion of each of the phase encoding gradients Gpe,i,m and each of the dephase encoding gradients Gpe,i,m′; and
determining a scanning number, an imaging field of view, an imaging matrix, a sampling frequency, a time interval ESP between two neighboring refocusing pulses of the M refocusing pulses, a sampling acceleration factor, and a repetition time TR of the fast high-resolution multiple overlapping-echo imaging pulse sequence so as to determine values of various gradients in the fast high-resolution multiple overlapping-echo imaging pulse sequence.

### Claim 3 (depends on 1)

The method according to claim 1, wherein:
the constructing a deep neural network for reconstructing high-resolution multi-parametric quantitative magnetic resonance images comprises:
constructing a network structure of the deep neural network;
constructing a number of input channels of the deep neural network and a number of output channels of the deep neural network; and
constructing a loss function of the deep neural network.

### Claim 4 (depends on 1)

The method according to claim 1, wherein:
the generating training samples of the deep neural network comprises:
preparing a simulated proton density image, a simulated T1 quantitative image, a simulated T2 quantitative image, and a simulated T2* quantitative image according to characteristics of the real imaging object to form a virtual imaging object;
establishing non-ideal factor models based on non-ideal factors existing in actual scanning;
performing Bloch simulation on the virtual imaging object using the fast high-resolution multiple overlapping-echo imaging pulse sequence and the non-ideal factor models to obtain first k-space data of the virtual imaging object;
separating odd and even rows of the first k-space data of the virtual imaging object to fill into two k-spaces to respectively obtain two second k-space data of the virtual imaging object, performing an inverse Fourier transform to obtain image domain data of the virtual imaging object to form a training sample of the training samples together with the virtual imaging object and the non-ideal factor models; and
repeating the aforementioned steps to generate a set number of the training samples.

### Claim 5 (depends on 1)

The method according to claim 1, wherein:
the pre-processing the first k-space data of the real imaging object to obtain the image domain data of the real imaging object comprises:
separating collected odd and even rows of the first k-space data of the real imaging object to respectively fill into two k-spaces according to the phase encoding manner to obtain two second k-space data of the real imaging object;
parallel reconstructing the two second k-space data of the real imaging object provided that a sampling acceleration factor is more than 1; and
performing an inverse Fourier transform for the two second k-space data to obtain the image domain data of the real imaging object.

### Claim 6

A system for fast high-resolution multi-parametric quantitative magnetic resonance imaging, comprising:
a pulse sequence design module for designing a fast high-resolution multiple overlapping-echo imaging pulse sequence;
a sampling parameter determination module for determining sampling parameters of the fast high-resolution multiple overlapping-echo imaging pulse sequence;
a deep neural network construction module for constructing a deep neural network for reconstructing high-resolution multi-parametric quantitative magnetic resonance images;
a training sample generation module for generating training samples of the deep neural network;
a deep neural network training module for using the training samples to train the deep neural network to obtain trained deep neural networks;
a k-space data acquisition module for scanning a real imaging object using the fast high-resolution multiple overlapping-echo imaging pulse sequence under the sampling parameters to obtain k-space data of the real imaging object;
an image domain data acquisition module for pre-processing the k-space data of the real imaging object to obtain image domain data of the real imaging object; and
a multi-parametric quantitative magnetic resonance image acquisition module for inputting the image domain data of the real imaging object into the trained deep neural networks for the reconstructing to obtain the high-resolution multi-parametric quantitative magnetic resonance images of the real imaging object, wherein:
the fast high-resolution multiple overlapping-echo imaging pulse sequence comprises a signal excitation module and a data acquisition module,
the signal excitation module comprises N radio frequency (RF) excitation pulses with time intervals tn and flip angles αn, slice selection gradients Gss corresponding to the N RF excitation pulses, and echo shift gradients Gn,
n=1,2, . . . ,N, and N≥2,
each of the N RF excitation pulses is combined with a corresponding one of the slice selection gradients Gss in a slice selection dimension for slice selection,
the echo shift gradients Gn are applied after a corresponding RF excitation pulse of the N RF excitation pulses along a frequency encoding dimension and a phase encoding dimension,
the data acquisition module comprises a pre-phase gradient Gpre, refocusing pulses with time intervals ESP and flip angles βm, crushing gradients Ger, phase encoding gradients Gpe,i,m, frequency encoding gradients Gro, and dephase encoding gradients Gpe,i,m′,
i represents an ith scanning,
m=1,2, . . . ,M, and M≥2,
the pre-phase gradient Gpre is applied along the frequency encoding dimension, and a size of the pre-phase gradient Gpre is half of a corresponding one of the frequency encoding gradients Gro,
the crushing gradients Ga are applied before and after each refocusing pulse of the M refocusing pulses, and sizes and directions of the crushing gradients Ger along the frequency encoding dimension, the phase encoding dimension, and the slice selection dimension are the same, and
the phase encoding gradients Gpe,i,m, the frequency encoding gradients Gro, and the dephase encoding gradients Gpe,i,m′ are applied after each refocusing pulse of the M refocusing pulses, the phase encoding gradients Gpe,i,m and the dephase encoding gradients Gpe,i,m′ have same sizes but opposite directions.