One-stop cardiac magnetic resonance examination technology

Abstract

The invention relates to the field of medical image examination, and discloses a one-stop heart magnetic resonance examination technology which comprises the following steps: S1, high-integration-degree nuclear magnetic resonance scanning technology: acquiring original data of various heart magnetic resonance examinations through one-time scanning by adopting a high-integration-degree imaging sequence, the original data comprises images with different contrast ratios for measuring myocardial tissues T1, T2 and T2 * and film images for evaluating heart and ventricular functions; s2, expansion type deep learning reconstruction: reconstructing all different T1, T2 and T2 * weighted images and movie images in quantitative imaging and movie imaging data by adopting a reconstruction technology based on deep learning; and S3, quantitative image reconstruction: adopting a quantitative image reconstruction technology based on deep learning to reconstruct T1, T2 and T2 * quantitative images. According to the method, movie imaging for heart function examination and T1 quantitative imaging, T2 quantitative imaging and T2 * quantitative imaging for myocardial tissue characterization are completed in one-time scanning.

Description

Technical Field

[0001] The present invention relates to the technical field of medical imaging examination, and in particular to a one-stop cardiac magnetic resonance examination technology. Background Art

[0002] Magnetic resonance imaging (MRI) technology uses the two relaxation times generated by the interaction between hydrogen protons in the human body and the external magnetic field, and modulates multiple scanning parameters to obtain different contrast images of tissues. It can perform multi-dimensional examinations of structure, function, composition, blood flow and metabolites, and is an important medical imaging technology. Compared with imaging technologies such as CT, ultrasound and PET, MRI has the advantages of high sensitivity, high resolution, high spatial coverage, multiple angles, no radiation and good soft tissue contrast, and has been widely used in the examination of different diseases.

[0003] In heart disease, magnetic resonance imaging is the gold standard method for evaluating the structure, function and extent of myocardial infarction. Quantitative imaging techniques (English: mapping) developed in recent years measure the tissue magnetic resonance relaxation time of myocardial tissue under a fixed magnetic field, mainly the longitudinal relaxation time T 1 , transverse relaxation time T 2 and T 2 * Time. Compared with conventional qualitative examination techniques, this quantitative technique has the advantages of being objective, independent of the degree of lesions, and not requiring reference tissues. It overcomes the limitation of MRI that diffuse lesions cannot be examined, and has great potential clinical value in the early detection of heart disease. At the same time, this quantitative property helps to achieve accurate diagnosis and prognosis, as well as to conduct multicenter and long-term studies. 1 、T 2 and T 2 * The physical properties of the myocardial tissue determine that these parameters have different sensitivities to the microenvironment of the myocardial tissue. Therefore, fibrosis, edema, iron deposition, and fat infiltration in the tissue cause inconsistent changes in these tissue parameters. Therefore, when diagnosing a disease, it is often necessary to combine the information on changes in these tissue parameters to make an accurate assessment in clinical practice.

[0004] T 1 、T 2 or T 2 * Quantitative imaging technology requires the acquisition of multiple images of the heart in the same motion state with different contrasts. In image post-processing, the multiple signals of each pixel are fitted to the model of signal evolution during image acquisition to obtain the corresponding T 1 、T 2 and T 2 *. The above process is fitted for all pixels to obtain a cardiac image. The value of each pixel is the relaxation time of the corresponding tissue, which is called Map. The whole process is called quantitative image reconstruction. At present, cardiac magnetic resonance requires multiple independent scanning processes to complete the examination of these different quantitative results. There are the following problems: 1) There is a mismatch between images of different contrasts, which limits the further analysis of tissue characteristics based on the pixel level; 2) The execution of different scanning sequences, switching between sequences, and the preparation time of each sequence cause the patient to hold his breath frequently and the total scanning time is long, resulting in low scanning efficiency and heavy economic burden on patients; 3) There is data redundancy between different imaging methods. In the process of image reconstruction, the redundant information of time and space is not fully utilized to shorten the amount of data required for each examination; 4) The commonly used quantitative map reconstruction method (mainly the least squares method) is sensitive to the initial value and is prone to fitting errors.

[0005] Therefore, there is an urgent need to develop a fast, highly integrated cardiac magnetic resonance examination technology that can achieve one-stop cardiac function and multiple quantitative time examinations in a single scan, as well as a robust quantitative image reconstruction algorithm to solve the problems existing in the above-mentioned cardiac magnetic resonance quantitative technology and improve scanning efficiency and disease diagnosis capabilities. Summary of the invention

[0006] In view of the above-mentioned deficiencies of the prior art, the present invention proposes a one-stop cardiac magnetic resonance examination technology, which mainly solves the problems of low scanning efficiency, many times of breath holding by patients, long total breath holding time, and incomplete alignment between images when the examination is completed by different scanning methods in different scanning processes.

[0007] In order to achieve the above-mentioned purpose of the invention, the technical solution adopted by the present invention is: a one-stop cardiac magnetic resonance examination technology, comprising the following steps:

[0008] S1, high-integration magnetic resonance imaging sequence: A high-integration imaging sequence is used to collect raw data of different cardiac magnetic resonance examinations in one scan. These raw data include T 1 、T 2 、T 2 * quantitative imaging data and cine imaging data for assessment of ventricular function;

[0009] S2, Unfolded Deep Learning Reconstruction: Deep learning-based reconstruction techniques are used to reconstruct all different T 1 、T 2 、T 2 * weight images and movie images;

[0010] S3, Quantitative image reconstruction based on deep learning: Reconstruct T using quantitative image reconstruction technology based on deep learning 1 , T 2 and T 2 * Quantify images and improve image quality.

[0011] Furthermore, the one-stop cardiac magnetic resonance examination technique includes an imaging sequence comprising simultaneous T 1 and T 2 Quantitative Imaging Module, T 2 * The quantitative imaging module and the movie imaging module are divided into three parts, and the order of the three parts is designed as T 1 and T 2 Quantitative Imaging Module, T 2 * The order of the quantitative imaging module and the movie imaging module is executed in one sequence. 1 、T 2 、T 2 * The quantitative images and the movie images are fully registered in space; image reconstruction includes reconstruction of original images of different quantitative examinations, reconstruction of segmented movie images, and reconstruction of real-time movie images for detecting arrhythmia.

[0012] Furthermore, the one-stop cardiac magnetic resonance examination technology can also 1 and T 2 The quantitative imaging module switches between pulse and T 2 Prepare a combination of pulses to prepare T 1 and T 2 Simultaneously quantitatively image the signal; 1 and T 2 The data of the quantitative imaging module and the movie imaging module are collected by reading out the gradient echo signal of a single echo; T 2 * The data of the quantitative imaging module is collected by the gradient echo signal readout method of multiple echoes; T 1 and T 2 At the same time, the quantitative imaging signal can also be obtained by saturation pulse and T 2 Prepare a combination of pulses to prepare.

[0013] Furthermore, the one-stop cardiac magnetic resonance examination technology can also 1 and T 2 Quantitative Imaging Module and T 2 *The quantitative imaging module collects data in the middle or end of the diastole phase of the cardiac cycle through ECG gating. The movie imaging module continuously collects ECG signal data of multiple cardiac cycles and retrospectively classifies the data into different cardiac phases based on the ECG signals to avoid image artifacts caused by adjacent phase motion.

[0014] Furthermore, the one-stop cardiac magnetic resonance examination technology can also 1 and T 2 Quantitative Imaging Module, T 2 * The quantitative imaging module and the movie imaging module use different image acceleration multiples to achieve fast scanning; 1 and T 2 The acceleration factor of the quantitative imaging module is 2-4 times; T 2 * The acceleration factor of the quantitative imaging module is 3-5 times; the acceleration factor of the movie imaging module is 11-16 times.

[0015] Furthermore, the one-stop cardiac magnetic resonance examination technology uses the golden rotation angle of the original data to detect the simultaneous T 1 and T 2 For the quantitative imaging module and the movie imaging module, the spoke rotation angle between each two readout radio frequencies is a golden angle according to the radiation trajectory; for multiple echoes of T 2 * In the quantitative imaging module, the rotation angle between two readout radio frequencies is a golden angle of 124°, and the rotation angle between multiple echoes is 0°; the rotation angle of the data of the same echo between two adjacent cardiac cycles is a golden angle, so that the data of the same echo from multiple cardiac cycles can cover the entire K space.

[0016] Furthermore, the one-stop cardiac magnetic resonance examination technology adopts the breath-holding method to suppress respiratory movement: the number of breath-holding times is 1, at this time, T 1 and T 2 The acquisition time of the quantitative imaging module is 4-6 cardiac cycles, that is, 4-6 seconds; T 2 * The acquisition time of the image acquisition module is 4 heart cycles, i.e. 4 seconds; the acquisition time of the movie imaging module is 4 heart cycles, i.e. 4 seconds;

[0017] When using respiratory navigation to suppress respiratory motion: T 1 and T 2 Quantitative Imaging Module and T 2 * The quantitative imaging module uses diaphragmatic breathing navigation, which is placed before image acquisition or T 2Before the preparation pulse is executed; the movie imaging adopts the self-navigation mode, the self-navigation data acquisition direction is the head and foot direction, and the execution cycle is 100-300ms.

[0018] Furthermore, in the above-mentioned one-stop cardiac magnetic resonance examination technology, the gradient echo acquisition method includes two methods: a fast gradient echo sequence and a balanced free precession gradient echo sequence; when the fast gradient echo sequence is used, the readout angle is 6-10°; when the balanced free precession gradient echo sequence is used, the readout angle is 35°-40°.

[0019] Furthermore, in the above-mentioned one-stop cardiac magnetic resonance examination technology, when the imaging sequence is applied to three-dimensional imaging, the image readout adopts a multiple excitation method with a readout angle of 15°-20°; respiratory motion is suppressed by respiratory navigation; cardiac motion aggravated by free breathing is suppressed by reducing the acquisition window; saturation pulses and T are used to suppress the heart movement. 2 The combination of prepared pulses acts as a simultaneous T 1 With T 2 Signal preparation method for quantitative imaging reduces signal waiting time and reduces mixed T 1 / T 2 The number of weight images accelerates 3D imaging.

[0020] Furthermore, in the above-mentioned one-stop cardiac magnetic resonance examination technology, in S2, the unfolded deep learning reconstruction is reconstructed by a deep learning model of an unfolded convolutional neural network, and the deep learning model of the unfolded convolutional neural network includes a convolutional neural network for removing image noise and a data fidelity item, and the convolutional neural network and the data fidelity item are cascaded to simulate the iterative optimization process of the traditional algorithm.

[0021] Furthermore, in the one-stop cardiac magnetic resonance examination technology, in S3, the quantitative image reconstruction is reconstructed by a deep neural network, and the deep neural network includes a recurrent neural network and a multi-layer fully connected neural network, and the recurrent neural network includes a plurality of nodes, each node including a grayscale signal, a flip recovery time, a T 2 Prepare the echo time of the pulse and the echo time of the multiple echoes.

[0022] Furthermore, in the one-stop cardiac magnetic resonance examination technology, the quantitative image reconstruction method includes the least square method, which is specifically: using the flip pulse and T 2 When the combination of the preparation pulses is 1 and T 2 The signal formula is as follows:

[0023]

[0024] Where A is the image signal corresponding to the longitudinal magnetization vector entering a steady state after multiple image acquisitions; TINV i is the flip recovery time corresponding to the i-th graph; TEprepi is T 2 Echo time of the preparation pulse;

[0025] T 2 * The calculation formula of the quantitative image is as follows:

[0026]

[0027] Mss is the signal value of the magnetization vector flipped to the horizontal direction after the longitudinal magnetization vector enters the steady state after the excitation pulse;

[0028] Use saturation pulse with T 2 When the combination of the preparation pulses is 1 and T 2 The signal formula is as follows:

[0029]

[0030] Where A is the image signal corresponding to the complete recovery of the longitudinal magnetization vector; T SAT i is the saturation recovery time corresponding to the i-th graph; TEprepi is the 2 Prepare the echo time of the pulse.

[0031] The beneficial effects of the present invention are as follows: the method can complete the movie imaging for cardiac function examination and T for myocardial tissue characterization in the conventional clinical magnetic resonance examination scanning protocol in one fast scan. 1 Quantitative imaging, T 2 Quantitative synthesis and T 2 * Quantitative imaging solves the problems of low scanning efficiency, high number of breath-holding times, long total breath-holding time and economic burden on patients in current clinical scanning technology; it provides a technology that can obtain multiple fully registered cardiac examination images, solving the problem of misalignment of results obtained when different sequences are used in different scans to complete these examinations in clinical practice, which limits the recognition of subtle abnormalities from images with different contrasts. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 A flowchart of the method is shown in FIG.

[0033] Figure 2 A deployment of the fast and highly integrated cardiac scanning method of the present invention in a breath-holding mode;

[0034] Figure 3 A deployment of the fast and highly integrated cardiac scanning method of the present invention using diaphragmatic breathing navigation and self-navigation under free breathing;

[0035] Figure 4 The fast and highly integrated cardiac scanning method of the present invention utilizes diaphragmatic breathing navigation and self-navigation under free breathing and uses saturation pulse as T 1 A deployment of weight preparation pulses;

[0036] Figure 5 Reconstruct the network model for the unfolded deep learning based on convolutional neural network in this method;

[0037] Figure 6 is the T based on recurrent neural network and fully connected neural network in this method. 1 ,T 2 and T 2 * Quantitative image reconstruction network model;

[0038] Figure 7 These are various examination result images obtained by this method in one breath-hold cardiac magnetic resonance scan. DETAILED DESCRIPTION

[0039] The specific implementation modes of the present invention are described below to facilitate those skilled in the art to understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the scope of the present invention defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0040] like Figures 1 to 7 As shown, this embodiment provides a one-stop cardiac magnetic resonance examination technology, including the following steps:

[0041] S1, high-integration magnetic resonance imaging sequence: A high-integration imaging sequence is used to collect raw data of different cardiac magnetic resonance examinations in one scan. These raw data include T 1 、T 2 、T 2 * quantitative imaging data and cine imaging data for assessment of ventricular function;

[0042] S2, Unfolded Deep Learning Reconstruction: Deep learning-based reconstruction techniques are used to reconstruct all different T 1 、T 2 、T 2 * weight images and movie images;

[0043] S3, Quantitative image reconstruction based on deep learning: Reconstruct T using quantitative image reconstruction technology based on deep learning 1 , T2 and T 2 * Quantify images and improve image quality.

[0044] The imaging method includes simultaneous T 1 、T 2 and T 2 * Quantitative imaging and movie imaging: In terms of examination content, this method can evaluate both myocardial tissue composition and ventricular function; in terms of function, this method can examine myocardial fibrosis, fat infiltration, edema, iron deposition and inflammation, quantitative parameters of ventricular contraction and relaxation function (such as ejection fraction and output), etc. In addition, the movie imaging part of this method can be used to reconstruct real-time movie images to examine problems such as arrhythmia.

[0045] The method comprises a fast and highly integrated imaging sequence, which comprises simultaneous T 1 -T 2 Quantitative Imaging Module, T 2 * The quantitative imaging module and the cine imaging module are executed sequentially.

[0046] In order to be compatible with the image acquisition method, this imaging method 1 -T 2 Quantitative imaging, T 2 * Quantitative measurement and movie scanning use gradient echo with phase dispersion effect. Gradient echo is less affected by the inhomogeneity of B0 field and B1 field, ensuring T 1 and T 2 The excitation angle range of the gradient echo acquisition sequence should be 6°-10°. The specific angle can be determined based on the signal-to-noise ratio of the image and the mixed T 1 / T 2 The signal of the weight image is T 2 * Optimization is based on whether the data enters a steady state before collection.

[0047] In this method, T 1 -T 2 The images of the quantitative imaging module and the movie imaging module adopt the signal readout method of single excitation and single echo, the acceleration factor is 2-4, and the data acquisition window size is 190-220 milliseconds; T 2 * Quantitative imaging uses multiple echoes, with an acceleration factor of 3-5, an echo number of 8-10, a repeated acquisition time of 27-30 milliseconds, an echo time increment interval of 2 milliseconds, and a data acquisition window size of 190-220 milliseconds. 2* The signal formula of the weight image is: Here Mss is the component of the longitudinal magnetization vector flipped to the horizontal plane after entering the steady state after multiple imaging, and TEi is the echo center time of the i-th echo. 2 * The data of quantitative images are collected in a segmented manner over multiple cardiac cycles.

[0048] In the magnetic resonance data acquisition of this method, a radiation trajectory with a rotation angle of a golden angle is designed to fill the K space data. The radiation trajectory has the ability to achieve high-multiple accelerated acquisition. For a single echo sequence, the rotation angle between each two excitations is designed to be 124°; T 1 、T 2 In cine imaging, the repetition time (TR) was 3 ms and the echo time was 2-4 ms; for T 2 * The multiple echo acquisition method of quantitative imaging has 8-10 echoes, a repeated acquisition time (TR) of 27-30 milliseconds, and an echo time increment interval of 2 milliseconds; the rotation angle between echoes is 0. However, for the image of the same echo, the rotation angle between two close readouts is 124°. The field of view (FOV) of the image in the present invention is designed to be 228mm×228mm, and the resolution is 2×2mm 2 , so the full sampling data of each image is 114 radiation lines.

[0049] In order to reduce the influence of cardiac motion, the present invention adopts ECG gating. Specifically, during the image acquisition process, ECG gating checks the R wave of each cardiac cycle and triggers the start time of subsequent image acquisition, so that the image of a single excitation comes from the middle or end of the diastolic phase of the cardiac cycle, and the image acquisition window width of a single excitation is 190-220 milliseconds to ensure that the heart motion is small. During the data acquisition process of the movie imaging module, the ECG gate is not used for triggering, but the ECG-gated signal will be recorded and used in retrospective data processing.

[0050] Specifically, the method proposed in this paper is 1 -T 2 The quantitative imaging module collects 4-6 mixed T images over 4-6 cardiac cycles. 1 / T 2 The repetition time of the weighted image and single-shot image is designed to be 3 milliseconds. Therefore, in one cardiac cycle, the number of radiation lines in the K space of each image is 72, and the acceleration is 2 times.

[0051] At the same time T 1 and T 2 Quantitative imaging of the mixed weight signal was performed using T 1Prepare pulse and T 2 Prepare the pulse combination, where the flip pulse is used as T 1 Prepare pulse; T 2 The weights are determined by the T with Carr-Purcell-Meiboom-Gill signal characteristics. 2 Prepare pulse preparation. In order, first execute T 1 Prepare the pulse to allow the longitudinal vector to start relaxation recovery, and then collect T 1 Weight image; starting from the second image, T is applied before acquisition 2 Prepare the pulse and introduce T 2 Weight, the longitudinal magnetization vector before image acquisition is: Here M ZSS is the value of the longitudinal magnetization vector entering a steady state after multiple image acquisitions, T INV To flip the pulse to before image acquisition or T 2 Time between preparation pulses, TE prep It is T 2 Prepare the echo time of the pulse. Flip the pulse and T 2 The preparation pulses are designed as adiabatic pulses, T 2 The echo time of the preparation pulse is optimized in the range of 25 ms to 60 ms.

[0052] The T of this method 2 * The quantitative imaging part uses multiple echoes to collect the signal values ​​at equal intervals on the free induction decay curve (FID) to obtain different T 2 * In this method, T is collected over four consecutive cardiac cycles. 2 * Weighted image, 8 excitations are performed in each cardiac cycle, and the distance between two excitations is 25-30 milliseconds. Therefore, after each excitation, the signals of 9 echoes are continuously collected at equal intervals, and the step length of the echo is 2.5-3 milliseconds, ensuring that the image acquisition window in each cardiac cycle does not exceed 210 milliseconds. The data of the same echo from different cardiac cycles are complementary in K space, so as to ensure that the image of each echo has 32 radiation lines, and the acceleration is 4 times, ensuring that the image can be reconstructed.

[0053] The movie imaging module of this method collects data on four consecutive cardiac cycles, and the time resolution of each cardiac phase is designed to be 42 milliseconds. The number of K-space radiation lines collected in each cardiac cycle is 12, and the acquisition time of each line is 3 milliseconds. During the acquisition process, the patient's ECG signal is recorded. In the image post-processing, based on the recorded ECG signal and the window length of the phase, the data of the same phase on the four cardiac cycles are retrospectively merged into one K-space, and the acceleration factor is 2 times.

[0054] The execution order of the image acquisition process of this method is T 1 and T 2 Quantitative imaging, T 2 * Quantitative imaging and Cine imaging. This sequence ensures simultaneous T 1 and T 2 After quantitative imaging, the longitudinal magnetization vector enters a steady state and does not affect the subsequent T 2 * Quantitative imaging. 2 * The number of radio frequency pulses executed in quantitative imaging is sparse, about 8-9 times, so the disturbance to the longitudinal magnetization vector is small, so that the longitudinal magnetization vector is retained more before cine imaging, ensuring that the cine imaging has a higher signal-to-noise ratio.

[0055] When breath holding is used to suppress respiratory movement: the acquisition process requires 14 cardiac cycles to ensure that the single breath holding time does not exceed the patient's average breath holding ability. 1 -T 2 Quantitative imaging, T 2 * The optimized breath-hold lengths for quantitative imaging and Cine imaging were 40%, 30%, and 30%.

[0056] The image reconstruction of this method adopts an unfolded deep learning reconstruction network model based on a convolutional neural network. The reconstruction network model consists of a U-shaped convolutional neural network and a data consistency module. These two modules are repeated multiple times in a cascade manner to simulate the iterative optimization process in parallel imaging. Here, the convolutional neural network is used to remove aliasing artifacts caused by high acceleration in the image, and the data consistency module ensures that the original data remains unchanged before and after image denoising. The present invention uses volume data to train the neural network. Specifically, the image of the Cartesian acquisition trajectory is resynthesized into a data set of the radiation trajectory, and the fully acquired radiation trajectory data will be used to reconstruct the reference image; the fully acquired data is simulated and down-acquired to obtain training data.

[0057] In this method, a deep learning model based on recurrent neural network and fully connected network is constructed to reconstruct quantitative images. Each input endpoint in the recurrent neural network contains a grayscale signal S and corresponding time information t (flip recovery time, T 2 Prepare the echo time of the pulse or T 2 * The central time of multiple echoes), the recurrent neural network contains multiple hidden layers and is designed to be bidirectional, which is used to mine the relationship between the collected signal and the corresponding time. The input of the fully connected network is the output of the recurrent neural network, and the output of the fully connected network is T 1 、T 2 or T2 * The fully connected neural network contains 8-10 hidden layers, each with multiple nodes. The activation function between each hidden layer of the recurrent neural network and the fully connected neural network is based on the modified linear unit function. The recurrent neural network and the fully connected neural network of the present method are trained through simulation, phantom and in vivo data. In the process of generating phantom and simulated data, the influence of different B1, off-resonance, noise level and heart rate are added to improve the robustness of the network model and the quality of the image.

[0058] When using free breathing to collect data, free breathing will introduce motion artifacts, so it is necessary to optimize the length of each module and the number of repeated acquisitions. 1 / T 2 The preparation method of the weight signal sets the corresponding idle time to ensure that the longitudinal magnetization vector is consistent between different repeated acquisitions. When using diaphragm breathing navigation gating, there is a process of data rejection and repeated acquisition, so it is necessary to adjust the T 1 / T 2 The preparation of the weight signal is optimized to ensure that the data collected again after rejection is consistent with the rejected data in terms of signal and T 1 / T 2 When using breathing navigation, different breathing navigation methods can be used according to the tasks of different modules; for T 1 / T 2 / T 2 * Module, using diaphragmatic breathing navigation; for movie data acquisition, using the cardiac self-navigation method. When using diaphragmatic breathing navigation, the diaphragmatic breathing navigation is placed at the junction of the liver and lungs; in terms of timing, the diaphragmatic breathing gating is placed before image acquisition or T 2 Before the pulse, respiratory navigation can be used to determine whether the subsequent image data should be accepted or re-acquired; it can also be used to track the real-time position of the heart and adjust the imaging plane. When using respiratory navigation, for movie imaging, cardiac self-navigation is used, and the execution cycle of self-navigation is 100-500ms.

[0059] The imaging method proposed in this method can use Cartesian trajectory for K space filling. When the trajectory is Cartesian trajectory, the accelerated image acquisition method is designed as compressed sensing or parallel imaging; at the same time, T 1 and T 2 The acceleration factor of quantitative imaging is designed to be 2-3; T 2 * The acceleration factor is 4.5 times for quantitative imaging; 14 times for real-time cine imaging, and 3.5 times for cine imaging via multi-segment reconstruction.

[0060] This method also has T 1 and T 2 Quantitative imaging using saturation pulses with T 2 When preparing the mixed weight signal, the saturation pulse and T are not used before the first image acquisition. 2 Preparation pulse; the two preparation pulses are executed simultaneously before the subsequent image acquisition, more specifically, before the image acquisition starts, T 2 The preparation pulse and the saturation pulse are placed at each T 2 Before the preparation pulse, the saturation pulse and T 2 The time between the ready pulses is set to T SAT The signal of the image is: Here Mz0 is the value after the longitudinal magnetization vector is fully restored.

[0061] This method is in T 1 and T 2 In terms of calculation, the fitting method of the least squares algorithm can also be used. Specifically, the collected signal, T 1 Relaxation recovery time T INV and T 2 Echo time TE of the preparation pulse prep Fit to If a saturation pulse and T 2 When prep prepares the pulse, the signal formula is: For T 2 * Quantitative calculation, the fitting formula is:

[0062] When the scanning technology in this method is deployed in three dimensions, multiple excitations are used to complete the scanning of the entire heart. In addition to the 2x or higher acceleration used in two-dimensional data acquisition, three-dimensional data requires multiple excitations to complete the acquisition of an image; respiratory movement is synchronized using diaphragm navigation; in addition, the acquisition duration of each excitation is controlled at about 100ms; different T 1 / T 2 The number of weight images can be reduced to 3-4.

[0063] The above are only implementation methods of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A one-stop cardiac magnetic resonance examination technology, characterized in that: The following steps are involved: S1, high-integration magnetic resonance imaging sequence: A high-integration imaging sequence is used to collect raw data of different cardiac magnetic resonance examinations in one scan. These raw data include T1, T2, and T2 for measuring myocardial tissue. * quantitative imaging data and cine imaging data for assessment of ventricular function; S2, Expanded Deep Learning Reconstruction: Deep learning-based reconstruction technology is used to reconstruct all different T1, T2, T2 in quantitative imaging data and movie imaging data. * weight images and movie images; S3, Quantitative image reconstruction based on deep learning: Reconstruct T1, T2 and T2 using quantitative image reconstruction technology based on deep learning * Quantify images and improve image quality.

2. The one-stop cardiac magnetic resonance examination technology according to claim 1, characterized in that: The imaging sequence includes simultaneous T1 and T2 quantitative imaging modules, T2 * The three parts are divided into three parts: quantitative imaging module and movie imaging module; the order of the three parts is designed as T1 and T2 quantitative imaging modules, T2 * The order of the quantitative imaging module and the cine imaging module is executed in one sequence.

3. The one-stop cardiac magnetic resonance examination technology according to claim 2, characterized in that: The simultaneous T1 and T2 quantitative imaging module prepares T1 and T2 simultaneous quantitative imaging signals by combining a flip pulse with a T2 preparation pulse; the data of the simultaneous T1 and T2 quantitative imaging module and the movie imaging module are collected by a single-echo gradient echo signal readout method; T2 * The data of the quantitative imaging module is collected by a gradient echo signal readout method of multiple echoes; the signals of simultaneous quantitative imaging of T1 and T2 can also be prepared by a combination of a saturation pulse and a T2 preparation pulse.

4. The one-stop cardiac magnetic resonance examination technology according to claim 2, characterized in that: Simultaneous T1 and T2 quantitative imaging module and T2 * The quantitative imaging module collects data in the middle or end of the diastole phase of the cardiac cycle through ECG gating. The movie imaging module continuously collects ECG signal data of multiple cardiac cycles and retrospectively classifies the data into different cardiac phases based on the ECG signals to avoid image artifacts caused by adjacent phase motion.

5. The one-stop cardiac magnetic resonance examination technology according to claim 2, characterized in that: The simultaneous T1 and T2 quantitative imaging module, T2 * The quantitative imaging module and the movie imaging module use different image acceleration multiples to achieve rapid scanning; the acceleration multiple of the simultaneous T1 and T2 quantitative imaging modules is 2-4 times; T2 * The acceleration factor of the quantitative imaging module is 3-5 times; the acceleration factor of the movie imaging module is 11-16 times.

6. The one-stop cardiac magnetic resonance examination technology according to claim 2, characterized in that: The raw data adopts the radiation track of the golden rotation angle. For the simultaneous T1 and T2 quantitative imaging modules and the movie imaging module through the two-dimensional single echo, the spoke rotation angle of the radiation track between each two readout radio frequencies is a golden angle; for the T2 of multiple echoes * In the quantitative imaging module, the rotation angle between two readout radio frequencies is a golden angle of 124°, and the rotation angle between multiple echoes is 0°; the rotation angle of the data of the same echo between two adjacent cardiac cycles is a golden angle, so that the data of the same echo from multiple cardiac cycles can cover the entire K space.

7. The one-stop cardiac magnetic resonance examination technology according to claim 2, characterized in that: When holding breath to suppress respiratory movement: the number of breath holdings is 1, and the acquisition time of the T1 and T2 quantitative imaging modules is 4-6 cardiac cycles, that is, 4-6 seconds; T2 * The acquisition time of the image acquisition module is 4 heart cycles, i.e. 4 seconds; the acquisition time of the movie imaging module is 4 heart cycles, i.e. 4 seconds; When respiratory motion is suppressed using respiratory navigation: Simultaneous T1 and T2 quantitative imaging module and T2 * The quantitative imaging module uses diaphragmatic breathing navigation, which is placed before image acquisition or before the T2 preparation pulse is executed; Movie imaging adopts self-navigation mode, the direction of self-navigation data collection is head and foot direction, and the execution cycle is 100-300ms.

8. The one-stop cardiac magnetic resonance examination technology according to claim 3, characterized in that: The gradient echo acquisition methods include two methods: a rapid gradient echo sequence and a balanced free precession gradient echo sequence; when the rapid gradient echo sequence is used, the readout angle is 6-10°; when the balanced free precession gradient echo sequence is used, the readout angle is 35°-40°.

9. The one-stop cardiac magnetic resonance examination technology according to claim 2, characterized in that: When the imaging sequence is applied to three-dimensional imaging, the image readout adopts a multiple excitation method, and the readout angle is 15°-20°; respiratory movement is suppressed by respiratory navigation; and cardiac movement aggravated by free breathing is suppressed by reducing the acquisition window; A combination of saturation pulses and T2 preparation pulses is used as a signal preparation method for simultaneous T1 and T2 quantitative imaging to reduce signal waiting time; reducing the number of mixed T1 / T2 weighted images accelerates three-dimensional imaging.

10. The one-stop cardiac magnetic resonance examination technology according to claim 1, characterized in that: In S2, the expanded deep learning reconstruction is reconstructed through a deep learning model of an expanded convolutional neural network, wherein the deep learning model of the expanded convolutional neural network includes a convolutional neural network for removing image noise and a data fidelity item, and the convolutional neural network and the data fidelity item are cascaded to simulate the iterative optimization process of the traditional algorithm.

11. The one-stop cardiac magnetic resonance examination technology according to claim 1, characterized in that: In S3, quantitative image reconstruction is performed through a deep neural network, wherein the deep neural network includes a recurrent neural network and a multi-layer fully connected neural network, wherein the recurrent neural network includes multiple nodes, each node including a grayscale signal, a flip recovery time, an echo time of a T2 preparation pulse, and an echo time of multiple echoes.

12. The one-stop cardiac magnetic resonance examination technology according to claim 3, characterized in that: The quantitative image reconstruction method includes the least square method, which is specifically: when a combination of a flip pulse and a T2 preparation pulse is used, the signal formulas of T1 and T2 are simultaneously as follows: Where A is the image signal corresponding to the longitudinal magnetization vector entering a steady state after multiple image acquisitions; T INV i is the inversion recovery time corresponding to the i-th image; TEprepi is the echo time of the T2 preparation pulse; T2 * The calculation formula of the quantitative image is as follows: Mss is the signal value of the magnetization vector flipped to the horizontal direction after the longitudinal magnetization vector enters the steady state after the excitation pulse; When using the combination of saturation pulse and T2 preparation pulse, the signal formula of T1 and T2 is formula (3): Where A is the image signal corresponding to the complete recovery of the longitudinal magnetization vector; T SAT i is the saturation recovery time corresponding to the ith image; TEprepi is the echo time with the T2 preparation pulse.

Images