WO2019153654A1

WO2019153654A1 - Fractional-order model-based magnetic resonance fingerprinting method and device, and medium

Info

Publication number: WO2019153654A1
Application number: PCT/CN2018/096146
Authority: WO
Inventors: 王海峰; 梁栋; 邹莉娴; 刘新; 郑海荣
Original assignee: 深圳先进技术研究院
Priority date: 2018-02-08
Filing date: 2018-07-18
Publication date: 2019-08-15
Also published as: CN110133554B; CN110133554A

Abstract

A fractional-order Bloch model-based quantitative parameter magnetic resonance fingerprinting method, the steps thereof comprising: collecting data using different times of repetition, times of echo and flip angles in each excitation of pulse sequences, and reconstructing said data to obtain an image; generating a dictionary on the basis of a fractional-order Bloch model; and obtaining a multi-parameter image using pattern recognition. The method improves the sampling speed and imaging accuracy of quantitative parameter magnetic resonance fingerprinting. While improving the accuracy of quantitative parameter imaging, the present invention does not add any additional data and scanning time.

Description

Magnetic resonance fingerprint imaging method, device and medium based on fractional order model

Technical field

The present invention relates to the field of magnetic resonance imaging technologies, and in particular, to a magnetic resonance fingerprint imaging method.

Background technique

Magnetic resonance imagers use magnetic fields and radio wave pulses to generate images of body tissue and structure. Compared to traditional magnetic resonance imaging, Magnetic Resonance Fingerprinting (MRF) can obtain more information per measurement. As a new method for obtaining multi-parameter quantitative imaging quickly, the magnetic resonance fingerprint imaging technology mainly includes the following steps: 1. Using different repetition time (TR) and echo in each excitation of the pulse sequence. Time of Echo (TE) and Flip Angle (FA), and acquire data with multi-interleaf spiral and reconstruct an undersampled image sequence; 2. Parameters according to pulse sequence (TR) , TE and FA), based on the classic first-order Bloch model to calculate the dictionary; 3, the reconstructed image sequence and the elements in the dictionary are matched point by point, and multi-parameter quantitative imaging results can be obtained at the same time.

Traditional quantitative magnetic resonance imaging obtains the characteristic parameters of the tissue by nonlinear fitting of the signal at each pixel position of the image through the classical Bloch model (a model describing the physical process of magnetic resonance), such as: longitudinal relaxation time (T1), Lateral relaxation time (T2) and the like. However, the imaging method has a large error, and the quantitative parameter imaging result is also not satisfactory.

Summary of the invention

The technical problem to be solved by the present invention is to provide a new magnetic resonance fingerprint imaging method to solve the problem of large error.

In order to solve the above technical problem, the present invention firstly discloses a magnetic resonance fingerprint imaging method, the technical solution of which is implemented as follows:

A magnetic resonance fingerprint imaging method, the steps of the magnetic resonance fingerprint imaging method comprising:

Step S1, in each excitation of the pulse sequence, using different repetition time, echo time, flip angle, and using non-Cartesian trajectory in K space for data acquisition;

Step S2, using a Bloch model based on the fractional order model to simulate the calculation, and generating a dictionary based on the fractional Bloch model;

In step S3, the magnetic resonance image is reconstructed, and the signal of the corresponding element in the image is compared with the dictionary, and finally the multi-parameter quantitative imaging result is obtained.

Preferably, the step S2 is specifically: generating a dictionary by using a fractional Bloch model simulation calculation according to a parameter set of the adopted pulse sequence.

Preferably, in the step S3, the method for reconstructing the magnetic resonance image comprises: one of a non-uniform fast Fourier transform, a singular value decomposition back projection method and a low rank alternating direction multiplier algorithm.

Preferably, in the step S3, the point-by-point time series of the reconstructed image and the elements in the dictionary are matched one by one by the maximum inner product method to obtain a multi-parameter quantitative imaging result.

Preferably, the variables in the dictionary include a longitudinal relaxation time and a transverse relaxation time;

In the step S3, the multi-parameter quantitative imaging result is a quantitative imaging of the tissue characteristic parameters, including longitudinal relaxation time parameter imaging and transverse relaxation time parameter imaging.

Preferably, the longitudinal relaxation is represented by a fractional Bloch model as:

M _z (t)=M _z (0)+[M ₀ -M _z (0)][1-E _β (-(t/T ₁ ) ^β )].

Preferably, the transverse relaxation is expressed by a fractional Bloch model as:

M _xy (t)=M _xy (0)[E _α (-(t/T ₂ ) ^α )]+M _xy (∞).

Preferably, assuming that the value of τ is small, the Mita-Lehler function can be approximated as

then

Or / and

The present invention also discloses a computer readable medium having a program stored therein for a computer to perform the magnetic resonance fingerprint imaging method.

The present invention also discloses a magnetic resonance image processing apparatus for using the magnetic resonance fingerprint imaging method, comprising a data acquisition module, a dictionary generation module, and an imaging result generation module;

The data acquisition module is configured to use different repetition times, echo times, flip angles in each excitation of the pulse sequence, and use the non-Cartesian trajectory in the K space for data acquisition;

The dictionary generation module simulates a calculation by a Bloch model based on a fractional order model, and generates a dictionary based on a fractional Bloch model;

The imaging result generating module is configured to reconstruct a magnetic resonance image, and compare a signal of a corresponding element in the image with the dictionary to finally obtain a multi-parameter quantitative imaging result.

Preferably, the dictionary generating module generates a dictionary by using a fractional Bloch model to simulate a parameter according to a parameter set of the adopted pulse sequence.

The imaging result generation module generates a reconstructed magnetic resonance image by using one of a non-uniform fast Fourier transform, a singular value decomposition back projection method, and a low rank alternating direction multiplier algorithm.

Preferably, the imaging result generation module identifies the point-by-point time series of the reconstructed image and the elements in the dictionary by the maximum inner product method one by one to obtain a multi-parameter quantitative imaging result.

Preferably, the multi-parameter quantitative imaging result generated by the imaging result generation module is a quantitative imaging of tissue characteristic parameters, including longitudinal relaxation time parameter imaging and transverse relaxation time parameter imaging.

Advantageously, said dictionary generation module generates said model by a longitudinally relaxed fractional Bloch model or/and a laterally relaxed fractional Bloch model;

The fractional-order Bloch model of longitudinal relaxation is:

M _z (t)=M _z (0)+[M ₀ -M _z (0)][1-E _β (-(t/T ₁ ) ^β )]

The fractional-order Bloch model of transverse relaxation is:

M _xy (t)=M _xy (0)[E _α (-(t/T ₂ ) ^α )]+M _xy (∞).

The beneficial effects of implementing the invention are:

1. The invention improves the sampling speed and imaging accuracy of magnetic resonance fingerprint quantitative parameter imaging;

2. The present invention does not add additional data and scan time while improving the accuracy of quantitative parameter imaging.

DRAWINGS

For a better understanding of the technical solutions of the present invention, reference may be made to the following drawings for illustrating the prior art or embodiments. These drawings will briefly show some of the embodiments or products or methods involved in the prior art. The basic information of these drawings is as follows:

Figure 1 is a diagram showing the flip angle variation of a pulse sequence employed in one embodiment;

Figure 2 is a diagram showing the repetition time and echo time variation of the pulse sequence employed in one embodiment;

3 is a timing diagram of a magnetic resonance fingerprint imaging pulse sequence in an embodiment;

4 is a view showing an imaging result of a longitudinal relaxation time parameter in one embodiment;

Figure 5 is a graph showing the results of imaging the transverse relaxation time parameters in one embodiment.

Detailed ways

The technical solutions or the beneficial effects of the embodiments of the present invention will be further described. It is obvious that the described embodiments are only partial, but not all, of the present invention.

It should be noted that the present invention has been made primarily to solve the problems of the corresponding prior art in the field of magnetic resonance image reconstruction, and therefore the present invention is particularly applicable to the subdivision, but does not mean that the present invention The scope of application of the technical solution is thus limited, and those skilled in the art can reasonably implement various specific applications in the field of magnetic resonance imaging as needed.

A magnetic resonance fingerprint imaging method, the steps of the imaging method comprising:

The setting of the pulse sequence can be referred to FIG. 1 to FIG. 3.

The invention replaces the classical first-order Bloch model with a more accurate fractional Bloch model to build a dictionary of fractional Bloch models, and then uses the generated dictionary elements for pattern recognition matching recognition to obtain quantitative Imaging of tissue characteristic parameters. This method improves the dictionary accuracy established by magnetic resonance fingerprint imaging and reduces the error of pattern recognition, making magnetic resonance fingerprint imaging closer to the gold standard quantitative parameter imaging.

In a preferred embodiment, the step S2 is specifically: generating a dictionary by using a fractional Bloch model simulation calculation according to a parameter set of the adopted pulse sequence.

In a preferred embodiment, in the step S3, the method for reconstructing the magnetic resonance image comprises: one of a non-uniform fast Fourier transform, a singular value decomposition back projection method and a low rank alternating direction multiplier algorithm.

In a preferred embodiment, in step S3, the point-by-point time series of the reconstructed image and the elements in the dictionary are matched one by one by the maximum inner product method to obtain a multi-parameter quantitative imaging result.

In a preferred embodiment, the variables in the dictionary include a longitudinal relaxation time and a transverse relaxation time;

In a preferred embodiment, the longitudinal relaxation is represented by a fractional Bloch model as:

M _z (t)=M _z (0)+[M ₀ -M _z (0)][1-E _β (-(t/T ₁ ) ^β )].

In a preferred embodiment, the transverse relaxation is represented by a fractional Bloch model as:

M _xy (t)=M _xy (0)[E _α (-(t/T ₂ ) ^α )]+M _xy (∞).

In the above formula, the meaning of each character is:

The β-order differential operator of Riemann-Liouville in Caputo form

M ₀ : initial magnetization vector

M _z (t): longitudinal magnetization vector at time t

M _xy (t): transverse magnetization vector at time t

Β-order T ₁ relaxation

Α-order T ₂ relaxation

E _β (-(t/T ₁ ) ^β ): β ₁ stretch Mittag-Leffler function of T ₁

E _α (-(t/T ₂ ) ^α ): α ₂ stretch Mittag-Leffler function of T ₂

ω ₀ : resonance frequency

Riemann-Liouville's (1-α) order integral operator

The Caputo and Riemann–Liouville formulas can be found in “Chaotic, Control and Synchronization of Fractional Dynamic Systems” by Gao Xin, Liu Xingwen, and Shao Shiquan, or other related literature.

The Mittag-Leffler correlation function can be found in the Complex Variable Function Theory (updated edition) or other related literature compiled by Qing Hanlai.

In a preferred embodiment, it is specifically assumed that when the value of τ is small, the Mita-Levre function is approximated as

then

Or / and

In a preferred embodiment, the dictionary generation module generates the dictionary by using a fractional Bloch model to simulate calculations according to a parameter set of the employed pulse sequence.

In a preferred embodiment, the imaging result generation module identifies the point-by-point time series of the reconstructed image and the elements in the dictionary by the maximum inner product method one by one to obtain a multi-parameter quantitative imaging result.

In a preferred embodiment, the multi-parameter quantitative imaging result generated by the imaging result generation module is a quantitative imaging of tissue characteristic parameters, including longitudinal relaxation time parameter imaging and transverse relaxation time parameter imaging.

In a preferred embodiment, the dictionary generation module generates the model by a longitudinally relaxed fractional Bloch model or/and a laterally relaxed fractional Bloch model;

The fractional-order Bloch model of longitudinal relaxation is:

M _z (t)=M _z (0)+[M ₀ -M _z (0)][1-E _β (-(t/T ₁ ) ^β )]

The fractional-order Bloch model of transverse relaxation is:

M _xy (t)=M _xy (0)[E _α (-(t/T ₂ ) ^α )]+M _xy (∞).

Simulations were performed using the method of the present invention to obtain Figures 4 and 5, and the average difference (avg.diff.?) of each image is shown. 4 is a longitudinal relaxation time parameter imaging (T1 mapping) obtained by using a conventional method and the method of the present invention to generate a dictionary, and finally by a singular value decomposition back projection method and a low rank alternating direction multiplier algorithm. Fig. 5 is a transverse relaxation time parameter imaging (T2 mapping) obtained by generating a dictionary using the conventional method and the method of the present invention, respectively, and finally by a singular value decomposition back projection method and a low rank alternating direction multiplier algorithm.

The reference mean square error of the image and the comparative reference image is reduced compared to the classical first-order Bloch model, while no additional data and scan time are added. At present, experiments show that the T1 mapping and T2 mapping generated by the fractional Bloch model of the present invention are more than 50% accurate than the quantitative parameter imaging of the original first-order Bloch model.

It should be noted that the above-exemplified embodiments are preferred and preferred embodiments of the present invention, and are only used to explain and explain the technical solutions of the present invention in order to facilitate the reader's understanding and are not intended to limit the present invention. The scope of protection or application. Therefore, the technical solutions obtained by any modification, equivalent replacement, improvement, etc., made within the spirit and principle of the present invention are intended to be included in the scope of the present invention.

Claims

A magnetic resonance fingerprint imaging method based on fractional order model, characterized in that:

The steps of the magnetic resonance fingerprint imaging method include:

Step S1, in each excitation of the pulse sequence, using different repetition time, echo time, flip angle, and using non-Cartesian trajectory in K space for data acquisition;

Step S2, using a Bloch model based on the fractional order model to simulate the calculation, and generating a dictionary based on the fractional Bloch model;

In step S3, the magnetic resonance image is reconstructed, and the signal of the corresponding element in the image is compared with the dictionary, and finally the multi-parameter quantitative imaging result is obtained.
The magnetic resonance fingerprint imaging method according to claim 1, wherein:

The step S2 is specifically: generating a dictionary by using a fractional Bloch model simulation calculation according to a parameter set of the adopted pulse sequence.
The magnetic resonance fingerprint imaging method according to claim 2, wherein:

In the step S3, the method for reconstructing the magnetic resonance image comprises: one of a non-uniform fast Fourier transform, a singular value decomposition back projection method and a low rank alternating direction multiplier algorithm.
The magnetic resonance fingerprint imaging method according to claim 3, wherein:

In the step S3, the point-by-point time series of the reconstructed image and the elements in the dictionary are matched one by one by the maximum inner product method to obtain a multi-parameter quantitative imaging result.
The magnetic resonance fingerprint imaging method according to any one of claims 4, wherein:

The variables in the dictionary include longitudinal relaxation time and transverse relaxation time;

In the step S3, the multi-parameter quantitative imaging result is a quantitative imaging of the tissue characteristic parameters, including longitudinal relaxation time parameter imaging and transverse relaxation time parameter imaging.
The magnetic resonance fingerprint imaging method according to claim 5, wherein:

The longitudinal relaxation is expressed by the fractional Bloch model as:

M z (t)=M z (0)+[M 0 -M z (0)][1-E β (-(t/T 1 ) β )].
The magnetic resonance fingerprint imaging method according to claim 6, wherein:

The transverse relaxation is expressed by the fractional Bloch model as:

M xy (t)=M xy (0)[E α (-(t/T 2 ) α )]+M xy (∞).
The magnetic resonance fingerprint imaging method according to claim 7, wherein:

It can be assumed that the Mita-Levre function is approximated as

then
Or / and
A computer readable medium having a program stored therein for a computer to perform the magnetic resonance fingerprint imaging method of any one of claims 1-8.
A magnetic resonance image processing apparatus for using the magnetic resonance fingerprint imaging method according to any one of claims 1 to 8, characterized in that:

The utility model comprises a data acquisition module, a dictionary generation module and an imaging result generation module;

The data acquisition module is configured to use different repetition times, echo times, flip angles in each excitation of the pulse sequence, and use the non-Cartesian trajectory in the K space for data acquisition;

The dictionary generation module simulates a calculation by a Bloch model based on a fractional order model, and generates a dictionary based on a fractional Bloch model;

The imaging result generating module is configured to reconstruct a magnetic resonance image, and compare a signal of a corresponding element in the image with the dictionary to finally obtain a multi-parameter quantitative imaging result.
The device of claim 10 wherein:

The dictionary generating module simulates and calculates with a fractional Bloch model according to a parameter set of the adopted pulse sequence, and generates the dictionary;

The imaging result generation module generates a reconstructed magnetic resonance image by using one of a non-uniform fast Fourier transform, a singular value decomposition back projection method, and a low rank alternating direction multiplier algorithm.
The device of claim 10 wherein:

The imaging result generation module identifies the point-by-point time series of the reconstructed image and the elements in the dictionary by the maximum inner product method one by one to obtain a multi-parameter quantitative imaging result.
The device of claim 10 wherein:

The multi-parameter quantitative imaging result generated by the imaging result generation module is a quantitative imaging of tissue characteristic parameters, including longitudinal relaxation time parameter imaging and transverse relaxation time parameter imaging.
The device of claim 10 wherein:

The dictionary generation module generates the model by a longitudinally relaxed fractional Bloch model or/and a laterally relaxed fractional Bloch model;

The fractional-order Bloch model of longitudinal relaxation is:

M z (t)=M z (0)+[M 0 -M z (0)][1-E β (-(t/T 1 ) β )];

The fractional-order Bloch model of transverse relaxation is:

M xy (t)=M xy (0)[E α (-(t/T 2 ) α )]+M xy (∞).