Real-time MR movie data reconstruction method and system

Abstract

A method includes: training an algorithm using fully sampled retrospective MR cine data; and applying the trained algorithm to the real-time MR cine data to produce a reconstructed MR image.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the benefit of U.S. Provisional Application No. 62 / 941,904, filed on November 29, 2019, and U.S. Provisional Application No. 17 / 060,988, filed on October 1, 2020, which are hereby incorporated by reference in their entirety. Technical Field

[0003] Aspects of the present disclosure relate generally to magnetic resonance imaging (MRI), and more particularly to using parallel imaging for real-time MR cine image reconstruction. Background Art

[0004] MRI is a widely used medical technology that uses magnetic and radio frequency energy to produce images of regions of interest. During an MRI scan, a volume coil (e.g., a body coil) and a local coil (e.g., a surface coil) can acquire MR signals generated by nuclear relaxation inside the object being examined. Cardiovascular MRI (cMRI) is the recognized gold standard for clinical assessment of cardiac structure and function, that is, it is widely recognized as a particularly effective procedure. Standard CMR applications such as retrospective movies rely on ECG gating plus several breath holds to provide high diagnostic image quality. This may be difficult because typical patients who may need to be evaluated for cardiac structure and function may have irregular heartbeat signals and have difficulty holding their breath. Real-time cardiac movie MRI utilizes relatively fast image acquisition, so compared to retrospective movie MRI, neither ECG gating nor breath holding is required during the data acquisition process. Therefore, real-time movies may be more useful for patients who may have difficulty holding their breath or may have irregular heart signals during MRI examinations. However, to achieve the required rapid image acquisition speeds, real-time cine may typically acquire highly undersampled data while utilizing parallel imaging techniques (typically using accelerations greater than 10X). This can present computational challenges for MRI image reconstruction when reconstructing undersampled data and reconstructing data from the multiple coils used in parallel imaging.

[0005] Compressed sensing-based methods have been proposed for real-time movie reconstruction. In addition, several deep learning-based methods have also been proposed for MRI reconstruction. For example, Qin et al. (Qin, Chen et al., "Convolutional recurrent neural networks for dynamic MR image reconstruction." IEEE transactions on Medical Imaging 38.1 (2018): pp. 280-290) have developed a convolutional recurrent neural network for cardiac MRI image reconstruction. However, these studies have several limitations. In traditional machine learning or deep learning frameworks, a gold standard or gold standard dataset is required to teach the deep learning model how to reconstruct images. However, given the sampling time and the number of coils, it is almost impossible to obtain fully sampled real-time movie data between heartbeats. It can be seen that the proposed method reconstructs simulated undersampled data from retrospective movie data instead of using actual real-time movie data for evaluation; the acceleration rate is less than 10X; the previous method is only designed for single-coil image reconstruction instead of multi-coil image reconstruction (i.e., parallel imaging); and the evaluation is performed only using image quality metrics instead of clinical usefulness. Summary of the invention

[0006] It would be advantageous to provide methods and systems that achieve high quality reconstruction of real-time cardiac cine MRI.The disclosed embodiments relate to real-time cardiac cine MR image reconstruction using an algorithm for parallel-imaged undersampled real-time cine data.

[0007] According to one aspect of the present disclosure, a method includes: training an algorithm using fully sampled retrospective cine data; and applying the trained algorithm to real-time MR cine data to produce a reconstructed MR image.

[0008] The method may include training an algorithm using one or more downsampled retrospective movie data and a downsampled mask.

[0009] The method may further include training the algorithm using retrospective cine data from individual coils of the multi-coil MR scanner.

[0010] The real-time MR cine data may include real-time MR cine data from individual coils of a multi-coil MR scanner.

[0011] The fully sampled retrospective movie data may be used to calculate a loss during training, where the loss may include one or more of a mean squared error loss, an L1 loss, a structural similarity index (SSIM) loss, or a Huber loss.

[0012] The algorithm may include a residual convolutional recurrent neural network.

[0013] The real-time MR cine data may include undersampled multi-coil real-time MR cine data.

[0014] The real-time MR cine data may include real-time MR cine data from an individual coil of a multi-coil MR scanner, and the algorithm may include a plurality of algorithms, each algorithm configured to be applied to data from a different individual coil of the multi-coil MR scanner.

[0015] The method may include combining reconstructed images from multiple algorithms using a root sum of squares method or coil sensitivity maps to generate a final combined image.

[0016] The real-time MR cine data may include real-time MR cine data from individual coils of a multi-coil MR scanner, and the algorithm may include a single algorithm configured to be applied to data from the individual coils of the multi-coil MR scanner.

[0017] According to another aspect of the present disclosure, a system includes: a real-time MR cine data source; and computing circuitry that implements an algorithm trained using fully sampled retrospective cine data, wherein the trained algorithm is configured to produce a reconstructed MR image when applied to the real-time MR cine data.

[0018] These and other aspects, implementations, and advantages of the exemplary embodiments will become apparent from the embodiments described herein considered in conjunction with the accompanying drawings. However, it should be understood that the description and drawings are designed for illustrative purposes only and not as a definition of the limitations of the disclosed invention, which should be referred to the appended claims. Additional aspects and advantages of the invention will be set forth in the following description and will be apparent in part from the description or may be understood by practice of the invention. Moreover, aspects and advantages of the invention may be realized and obtained by the means and combinations particularly pointed out in the appended claims. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In the following detailed section of the present disclosure, the present invention will be described in more detail with reference to the exemplary embodiments shown in the accompanying drawings. These embodiments are non-limiting exemplary embodiments, wherein the same reference numerals represent similar structures in several views of the accompanying drawings, wherein:

[0020] Figure 1 An exemplary process flow according to aspects of the disclosed embodiments is illustrated;

[0021] Figure 2 illustrates embodiments of exemplary systems incorporating aspects of the disclosed embodiments;

[0022] Figure 3 illustrates another exemplary system embodiment incorporating aspects of the disclosed embodiments;

[0023] Figure 4 illustrates a schematic block diagram of an exemplary multi-coil MRI data source according to the disclosed embodiments;

[0024] Figure 5A and Figure 5B illustrates different MRI multi-coil arrangements according to the disclosed embodiments;

[0025] Figure 6 illustrates an exemplary architecture of computing circuitry in which algorithms may be implemented according to the disclosed embodiments;

[0026] Figure 7 An example of an algorithm in the form of a deep residual convolutional recurrent neural network is illustrated; and

[0027] Figure 8 An example of a bidirectional convolutional recurrent neural network unit is illustrated. DETAILED DESCRIPTION

[0028] In the following detailed description, many specific details are set forth by way of example in order to provide a thorough understanding of the relevant disclosure. However, it should be apparent to those skilled in the art that the present disclosure can be practiced without such details. In other cases, in order to avoid unnecessarily obscuring aspects of the present disclosure, known methods, processes, systems, components and / or circuits have been described at a relatively high level without detailed description. Various modifications to the disclosed embodiments will be apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments and applications without departing from the spirit and scope of the present disclosure. Thus, the present disclosure is not limited to the embodiments shown, but should be given the widest scope consistent with the claims.

[0029] It should be understood that the terms "system", "unit", "module" and / or "block" used herein are a method of distinguishing different components, elements, parts, portions or assemblies at different levels in ascending order. However, if the terms and other expressions can achieve the same purpose, the terms can be replaced by other expressions.

[0030] It should be understood that when a unit, module or block is referred to as being "on," "connected to," or "coupled to" another unit, module or block, it can be directly on, directly connected to, or coupled to another unit, module or block, or there may be intervening units, modules or blocks, unless the context clearly indicates otherwise. As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0031] Generally, the words "module", "unit" or "block" as used herein refer to logic specifically implemented in hardware or firmware, or to a collection of software instructions. The modules, units or blocks described herein may be implemented as software and / or hardware, and may be stored in any type of non-transient computer-readable medium or another storage device. In some embodiments, software modules / units / blocks may be compiled and linked into an executable program. It should be understood that software modules may be called from other modules / units / blocks or from themselves, and / or may be called in response to detected events or interruptions. Software modules / units / blocks configured to be executed on a computing device may be provided on a computer-readable medium, such as an optical disc, a digital video disc, a flash drive, a disk, or any other tangible medium, or may be provided as digital download material (and may be initially stored in a compressed or installable format that requires installation, decompression or decryption prior to execution). Such software code may be partially or completely stored on a storage device that executes a computing device so as to be executed by the computing device. Software instructions may be embedded in firmware, such as an erasable programmable read-only memory (EPROM). It should also be understood that hardware modules / units / blocks may be included in connected logical components, such as gates and triggers, and / or may be included in programmable units, such as programmable gate arrays or processors. The modules / units / blocks or computing device functions described herein may be implemented as software modules / units / blocks, but may be represented in hardware or firmware. Typically, the modules / units / blocks described herein refer to logical modules / units / blocks that may be combined with other modules / units / blocks or divided into submodules / subunits / subblocks, regardless of their physical organization or storage. This description may be applicable to a system, an engine, or a portion thereof.

[0032] The terms used herein are used only for the purpose of describing specific examples and embodiments and are not intended to be limiting. As used herein, the singular forms "a", "an", and "the" may also be intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the term "comprising" when used in the present disclosure specifies the presence of integers, devices, behaviors, described features, steps, elements, operations, and / or parts, but does not exclude the presence or addition of one or more other integers, devices, behaviors, features, steps, elements, operations, parts, and / or groups thereof.

[0033] These and other features and characteristics of the present disclosure, as well as the methods of operation and functions of the relevant elements of the structure and the economy of combination and manufacture of the parts, may become more apparent when considering the following description with reference to the accompanying drawings, all of which form a part of the present disclosure. However, it should be expressly understood that the drawings are for illustration and description purposes only and are not intended to limit the scope of the present disclosure. It should be understood that the drawings are not to scale.

[0034] The disclosed embodiments relate to a method comprising: training an algorithm using fully sampled retrospective cine data; and applying the trained algorithm to multi-coil real-time MRI cine data to produce a reconstructed MRI image.

[0035] The disclosed embodiments also relate to a system comprising: a source of fully sampled retrospective cine MR data; an algorithm configured to be trained using the fully sampled retrospective cine MR data; and a source of multi-coil real-time MR cine data, wherein the trained algorithm can be applied to the multi-coil real-time MR cine data to produce a reconstructed MR image.

[0036] See also Figure 1 , illustrates a schematic block diagram of an exemplary system 100 incorporating aspects of the disclosed embodiments. The system may include a source 102 of multi-coil real-time MR cine data, an algorithm 104, and a source 106 of fully sampled retrospective cine MR data. The fully sampled retrospective cine data 106 may be used to train the algorithm 104, and the trained algorithm 104 may be applied to the multi-coil real-time MR cine data 102 to produce a reconstructed MR image 108.

[0037] Figure 2 and Figure 3 Embodiments of exemplary systems 200 , 300 incorporating aspects of the disclosed embodiments are illustrated. Figure 2 An exemplary embodiment of providing independent multi-coil reconstruction by providing an algorithm in the form of a deep learning model for each coil of a multi-coil MR system is illustrated. An inverse Fourier transform (iFT) can be performed on downsampled k-space real-time multi-coil data 202 to produce a multi-coil image 204 including aliasing and artifacts due to downsampling. The image 204 with aliasing and artifacts can be separated into individual images 2061-206 n . Each image 2061-206 n Application of associated algorithms 2081-208 n , to generate corresponding reconstructed images 2101-210 n , and the images 2101-210 can be combined to reconstruct n , to produce a combined reconstructed image 212. The reconstructed images from each individual coil may be combined using the root sum of squares (RSS) method or the coil sensitivity map to generate the final combined reconstructed image 212. The root sum of squares (RSS) method refers to a calculation method of first squaring, then summing, and then taking the square root.

[0038] Figure 3 An exemplary embodiment is illustrated that provides parallel multi-coil reconstruction by applying a single algorithm to images from coils. An inverse Fourier transform (iFT) can be performed on downsampled k-space real-time multi-coil data 302 to produce a multi-coil image 304 that includes aliasing and artifacts due to downsampling. The image 304 with aliasing and artifacts can be separated into individual images 3061-306 according to the individual coils from which they were obtained. n . Images 3061-306 n This may be provided as input to a single algorithm 308 to produce a reconstructed image 310 .

[0039] Figure 4 A schematic block diagram of an exemplary multi-coil MRI data source in the form of an MRI device 402 for providing multi-coil MRI data according to the disclosed embodiments is shown. The MRI device 402 may include an MRI scanner 404, a control circuit 406, and a display 408. The function, size, type, geometry, location, number, or magnitude of the MRI scanner 404 may be determined or changed according to one or more specific conditions. For example, the MRI scanner 404 may be designed to surround a subject (or a region of a subject) to form a tunnel MRI scanner (referred to as a closed-bore MRI scanner) or an open-bore MRI scanner (referred to as an open-bore MRI scanner). As another example, the MRI scanner may be portable and may be transported to a patient along a corridor and through a doorway, thereby providing MR scanning services to the patient, as opposed to transporting the patient to the MRI scanner. In some examples, the portable MRI scanner may be configured to scan a region of interest of the subject, such as the subject's brain, spinal cord, limbs, heart, blood vessels, and internal organs.

[0040] like Figure 4 As shown in the cross-sectional view in FIG. 4 , the MRI scanner 404 may include a magnetic field generator 410, a gradient magnetic field generator 412, and a radio frequency (RF) generator 414, all of which surround a table 416 on which a subject under study may be positioned. The MRI scanner 404 may also include: one or more coil arrays 418; an ECG signal sensor 420 for capturing MRI data in the form of ECG signals from a subject under study during an MRI scan; and a camera 422 for capturing MRI data in the form of video images of a subject under study during an MRI scan.

[0041] In some embodiments, MRI scanner 404 can perform scanning on the subject or the region of the subject. The subject can be, for example, a human body or other animal body. In some embodiments, the subject can be a patient. The region of the subject can include a part of the subject. For example, the region of the subject can include the tissue of the patient. The tissue can include, for example, lungs, prostate, breast, colon, rectum, bladder, ovary, skin, liver, spine, bone, pancreas, cervix, lymph, thyroid, spleen, adrenal gland, salivary gland, sebaceous gland, testis, thymus, penis, uterus, trachea, skeletal muscle, smooth muscle, heart, etc. In some embodiments, scanning can be a pre-scan for calibrating an imaging scan. In some embodiments, scanning can be an imaging scan for generating an image.

[0042] The main magnetic field generator 410 can generate a static magnetic field B0 and can include, for example, a permanent magnet, a superconducting electromagnet, a resistive electromagnet, or any magnetic field generating device suitable for generating a static magnetic field. The gradient magnetic field generator 412 can use a coil to generate a magnetic field in the same direction as B0 but with a gradient in one or more directions (e.g., along the X, Y, or Z axis in the coordinate system of the MRI scanner 404).

[0043] In some embodiments, the RF generator 414 can use an RF coil to transmit RF energy through a subject or a region of interest of the subject to induce electrical signals in the region of interest. The resulting RF field is generally referred to as a B1 field and is combined with a B0 field to generate MR signals spatially localized and encoded by a gradient magnetic field. The coil array 418 can generally be operated to sense the RF field and deliver a corresponding output to the control circuit 406. In some embodiments, the coil array can be operated to transmit and receive RF energy, while in other embodiments, the coil array can be operated to receive only.

[0044] Figure 5A and Figure 5B Different MRI multi-coil arrangements are illustrated. The multi-coil arrangements may include phased array coil arrangements and parallel array coil arrangements. Figure 5A An exemplary phased array coil arrangement is shown in which coils overlap and are coupled together to enhance gain and signal-to-noise characteristics. Figure 5B An exemplary parallel array arrangement is shown in which the coils are separated and optimized for parallel imaging. The coil arrangement may include any number of coils, depending on the particular application. Exemplary numbers of coils may include 12, 16, 24, 32, 64, or more.

[0045] Back to Figure 4, the control circuit 406 can control the overall operation of the MRI scanner 404, specifically, the overall operation of the magnetic field generator 410, the gradient magnetic field generator 412, the RF generator 414, and the coil array 418. For example, the control circuit 406 can control the magnetic field gradient generator to generate a gradient field along one or more of the X, Y, and Z axes, and control the RF generator to generate an RF field. In some embodiments, the control circuit 406 can receive commands from, for example, a user or another system, and control the magnetic field generator 410, the gradient magnetic field generator 412, the RF generator 414, and the coil array 418 accordingly.

[0046] The control circuit 406 may be connected to the MRI scanner 404 via a network 424. The network 424 may include any suitable network that may facilitate information and / or data exchange with the MRI scanner 404. The network 424 may be and / or include a public network (e.g., the Internet), a private network (e.g., a local area network (LAN), a wide area network (WAN), etc.), a wired network (e.g., Ethernet), a wireless network (e.g., an 802.11 network, a Wi-Fi network, etc.), a cellular network (e.g., a long term evolution (LTE) network), a frame relay network, a virtual private network (“VPN”), a satellite network, a telephone network, a router, a hub, a switch, a server computer, and / or any combination thereof. By way of example only, the network 424 may include a cable network, a wired network, a fiber optic network, a telecommunications network, an intranet, a wireless local area network (WLAN), a metropolitan area network (MAN), a public switched telephone network (PSTN), a Bluetooth network, a ZigBee network, a near field communication (NFC) network, etc., or any combination thereof. In some embodiments, the network 424 may include one or more network access points. For example, the network 424 may include wired and / or wireless network access points, such as base stations and / or Internet exchange points, through which one or more components of the MRI scanner 402 may connect to the network 424 to exchange data and / or information.

[0047] According to some embodiments, the algorithm may be implemented in computing circuitry of control circuitry 406 , while in other embodiments, the algorithm may be implemented in computing circuitry remote from control circuitry 406 .

[0048] Figure 6An exemplary architecture of a computing circuit 600 according to the disclosed embodiment is illustrated. The computing circuit 600 may include a computer readable program code stored on at least one computer readable medium 602 for performing and executing the processing steps described herein. The computer readable program code for performing the operations of the disclosed aspects may be written in any combination of one or more programming languages, including object-oriented programming languages ​​(such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET, Python, etc.), conventional programming languages ​​(such as "C" programming language, Visual Basic, Fortran 2103, Perl, COBOL 2102, PHP, ABAP), dynamic programming languages ​​(such as Python, Ruby and Groovy), or other programming languages. The computer readable program code may be executed entirely on the computing circuit 600, partially on the computing circuit 600, as a stand-alone software package, partially on the computing circuit 600 and partially on a remote computer or server, or entirely on a remote computer or server. In the latter scenario, the remote computer may be connected to computing circuitry 600 via any type of network, including the networks mentioned above with respect to network 424 .

[0049] The computer readable medium 602 may be a memory of the computing circuit 600. In alternative aspects, the computer readable program code may be stored in a memory external to or remote from the computing circuit 600. The memory may include a magnetic medium, a semiconductor medium, an optical medium, or any medium that is computer readable and executable. The computing circuit 600 may also include a computer processor 604 for executing the computer readable program code stored on at least one computer readable medium 602. In some embodiments, the computer processor may be a graphics processing unit (GPU). In at least one aspect, the computing circuit 600 may include one or more input or output devices, generally referred to as a user interface 606, which may be operated to allow input to the computing circuit 600 or provide output from the computing circuit 600, respectively. The computing circuit 600 may be implemented in hardware, software, or a combination of hardware and software.

[0050] Figure 7 An example of an algorithm in the form of a Deep Residual Convolutional Recurrent Neural Network (Res-CRNN) 700 is illustrated, and Figure 8An example of a bidirectional convolutional recurrent neural network unit 800 is illustrated. One feature of a deep residual convolutional recurrent neural network is that the state neurons of a conventional recurrent neural network (RNN) can be divided into a part responsible for the positive time direction (forward state) and a part responsible for the negative time direction (reverse state). In addition, the output from the forward state is not connected to the input of the reverse state, and the output from the reverse state is not connected to the input of the forward state.

[0051] The bidirectional convolutional RNN can model the dynamic information of the cardiac movie data, a data consistency layer that ensures that the reconstructed data is consistent with the observed data, and residual connections that promote the network to learn high-frequency details and increase the stability of the training process. Res-CRNN 700 may include three such building blocks 7021-7023 and an additional residual connection 7041-7043. Complex values ​​are represented as two-channel tensors and sent to the network. Deep residual convolutional recurrent neural networks can be trained with the same algorithm as regular unidirectional RNNs because there is no interaction between the two types of state neurons.

[0052] like Figure 7 As shown, real-time cine data 706 from the multi-coil downsampled MRI scanner 404 may be provided to the algorithm 700. Reconstructed cine images from each stage may be provided to the next stage until a final reconstructed cine image 708 is achieved.

[0053] The algorithm 700 may be trained using retrospective cine data and images reconstructed by the algorithm 700. The fully sampled retrospective cine data may be downsampled using sampling masks that are similar to the sampling masks used in real-time cine during MRI image acquisition. One or more of the downsampled retrospective cine data, the sampling mask, and the fully sampled data retrospective cine data may also be used to train the algorithm 700. The retrospective cine training data may include fully sampled images from a single coil of the MRI scanner 404. To save memory consumption, for example where the algorithm computation circuitry includes a graphics processing unit, the retrospective cine images may be cropped to a smaller size for training.

[0054] During training, fully sampled retrospective movies can be used to calculate the loss. The training loss can be implemented as one or any combination of mean squared error loss, L1 loss, structural similarity index (SSIM) loss, Huber loss, or any loss used for image quality assessment.

[0055] Once trained, algorithm 700 can be applied to undersampled real-time cine data for image construction, including images from multi-coil acquisitions and full-size uncropped images.

[0056] Thus, although the basic novel features of the present invention as applied to the exemplary embodiments of the present invention have been shown, described and pointed out, it should be understood that those skilled in the art may make various omissions, substitutions and changes in the form and details of the exemplified devices and methods and their operation without departing from the spirit and scope of the presently disclosed invention. Further, it is expressly contemplated that all combinations of those elements that perform substantially the same functions in substantially the same manner to achieve the same results are within the scope of the present invention. Moreover, it should be recognized that the structures and / or elements shown and / or described in conjunction with any disclosed form or embodiment of the present invention may be incorporated into any other disclosed or described or suggested form or embodiment as a general matter of design choice. Therefore, the present invention is intended to be limited only by the scope indicated by the scope of the appended claims.

Claims

1. A real-time MR movie data reconstruction method, comprising: The algorithm is trained using fully sampled retrospective movie data; wherein the retrospective movie data is acquired with electrocardiogram gating and breath holding; wherein the algorithm comprises a residual convolutional recurrent neural network, the residual convolutional recurrent neural network comprising at least one bidirectional convolutional recurrent neural network unit; and The trained algorithm is applied to real-time MR cine data acquired without electrocardiogram gating and without breath holding to produce a reconstructed MR image, wherein the real-time MR cine data includes undersampled multi-coil real-time MR cine data.

2. The method of claim 1, further comprising training the algorithm using one or more downsampled retrospective movie data and downsampled masks. 3 . The method of claim 1 , further comprising training the algorithm using retrospective cine data from individual coils of a multi-coil MR scanner.

4. The method according to claim 1, wherein: The undersampled multi-coil real-time MR cine data includes real-time MR cine data from individual coils of a multi-coil MR scanner.

5. The method according to claim 4, wherein: The algorithm includes a plurality of algorithms, each algorithm being configured to be applied to data from a different individual coil of the multi-coil MR scanner.

6. The method of claim 5, comprising combining the reconstructed images from the plurality of algorithms using a root sum square method to generate a final combined image.

7. The method of claim 5, comprising combining reconstructed images from the plurality of algorithms using coil sensitivity maps to generate a final combined image.

8. The method according to claim 4, wherein: The algorithm comprises a single algorithm configured to be applied to data from the individual coils of the multi-coil MR scanner.

9. A real-time MR movie data reconstruction system, comprising: Real-time MR movie data source; and A computational circuit implementing an algorithm trained using fully sampled retrospective cine data acquired with electrocardiogram gating and breath holding, the algorithm comprising a residual convolutional recurrent neural network comprising at least one bidirectional convolutional recurrent neural network unit, the trained algorithm being configured to produce a reconstructed MR image when applied to real-time MR cine data acquired without electrocardiogram gating and without breath holding, wherein the real-time MR cine data comprises undersampled multi-coil real-time MR cine data.

Images