System and method for simultaneous multi-layer imaging

Abstract

The present application provides a method for simultaneous multi-slice imaging, which can include acquiring target k-space data associated with a region of interest (ROI) of a subject. The method can also include generating at least two target images based on the target k-space data using a trained reconstruction model, wherein each target image corresponds to one of at least two target slices of the ROI and one of at least two target acquisitions. The reconstruction of the target k-space data using the machine learning model is a nonlinear approach that can improve image quality and reconstruction speed. The target k-space data inherently includes de-aliasing information. The machine learning model can automatically extract the de-aliasing information implicit in the target k-space data and output de-aliased images. In this way, the efficiency of the simultaneous multi-slice imaging reconstruction is improved without the need for additional operations to acquire a reference dataset (e.g., additional computational or scanning operations to acquire a reference dataset).

Description

[0001] Cross-reference to related applications

[0002] This application claims priority to U.S. Application No. 17 / 305,056, filed on June 30, 2021, the entire contents of which are incorporated herein by reference. Technical Field

[0003] This application relates generally to magnetic resonance imaging (MRI), and more particularly to systems and methods for simultaneous multi-slice imaging (SMS). Background Technology

[0004] Simultaneous multilayer imaging (SMS) has rapidly developed into a leading imaging technique for accelerating magnetic resonance imaging (MRI). Compared to imaging that excites only a single layer at a time, SMS can excite multiple layers simultaneously and acquire the magnetic resonance (MR) signals generated from these multiple excited layers concurrently. These MR signals are then filled into k-space to obtain k-space data. Because the received MR signals include contributions from multiple excited layers, directly reconstructing the k-space data using inverse Fourier transform may result in aliased images of multiple layers. Therefore, it is desirable to provide a SMS reconstruction system and / or method capable of generating dealiased images for each layer. Summary of the Invention

[0005] Some of the additional features of this application will be described in the following description. These additional features will be apparent to those skilled in the art from the study of the following description and corresponding drawings, or from an understanding of the production or operation of the embodiments. The features of this application can be implemented and obtained through the practice or use of various aspects of the methods, tools, and combinations set forth in the detailed examples discussed below.

[0006] According to one aspect of this application, a system for simultaneous multilayer imaging (SMS) is provided, the system comprising a magnetic resonance imaging (MRI) apparatus, a region of interest (ROI) configured to scan an object, one or more storage devices, and one or more processors communicating with the one or more storage devices. The one or more storage devices may include a set of instructions. When the one or more processors execute the set of instructions, they may be directed to perform one or more of the following operations: The one or more processors may acquire target k-space data associated with the ROI of the object. Based on the target k-space data, the one or more processors may generate at least two target images using a trained reconstruction model, wherein each target image corresponds to one of at least two target slices of the ROI and one target slice during at least two target acquisitions.

[0007] In some embodiments, the target k-space data may include at least two first k-space datasets, each of which is acquired in one of the at least two target slices and during one of the at least two target acquisition periods of the target slice.

[0008] In some embodiments, the target k-space data is undersampled.

[0009] In some embodiments, in order to generate the at least two target images using a trained reconstruction model based on the target k-space data, one or more processors may input the target k-space data into the trained reconstruction model; the one or more processors may output the at least two target images based on the target k-space data using the trained reconstruction model.

[0010] In some embodiments, to acquire target k-space data related to the region of interest (ROI) of the object, one or more processors may cause the magnetic resonance imaging (MRI) apparatus to apply one or more multi-band excitation radio frequency (RF) pulses to the ROI to simultaneously excite at least two target slices of the ROI once or more. The one or more processors may apply a transfer phase to at least one of the excited target slices by causing the MRI apparatus to apply phase modulation to the at least one excited target slice, the transfer phase varying with the at least two target acquisition periods and / or phase encoding direction. The one or more processors may acquire the target k-space data through the at least two excited target slices of the ROI during the at least two target acquisition periods.

[0011] In some embodiments, the transmission phases applied to two or more of the at least two target layers may be different.

[0012] In some embodiments, to generate at least two target aliased images based on the target k-space data using the trained reconstruction model, one or more processors can perform an inverse Fourier transform on the target k-space data, each of the at least two target aliased images corresponding to one of the at least two target acquisition periods and the at least two target slices. The one or more processors can input the at least two aliased images into the trained reconstruction model. The one or more processors can output the at least two target images based on the at least two target aliased images using the trained reconstruction model.

[0013] In some embodiments, in order to generate the at least two target images based on the target k-space data using the trained reconstruction model, the one or more processors may determine at least two reference datasets based on the target k-space data. Each of the at least two reference datasets may correspond to one of the at least two target slices and one of the at least two target acquisition periods. The at least two reference datasets may provide dealiasing information for dealiasing target slices on the target k-space data. The one or more processors may generate the at least two target images based on the target k-space data and the at least two reference datasets, using the trained reconstruction model.

[0014] In some embodiments, the trained reconstruction model may include a machine learning model.

[0015] In some embodiments, the trained reconstruction model is obtained through the following steps: acquiring at least two training datasets; and training an initial model based on the at least two training datasets to obtain the trained reconstruction model. Each of the at least two training datasets may include sample k-space data and at least two sample dealiasing images. Each of the at least two sample dealiasing images may correspond to one of at least two sample slices and one of at least two sample acquisition periods. For each of the at least two sample slices, the phase of the corresponding sample k-space data may vary based on the at least two sample acquisition periods and / or the phase encoding direction.

[0016] In some embodiments, the sample k-space data may include real k-space data, which is acquired by simultaneously exciting the at least two sample slices. The at least two sample dealiasing images may be generated based on the sample k-space data.

[0017] In some embodiments, the sample k-space data may be undersampled.

[0018] In some embodiments, the sample k-space data may include synthetic k-space data. The synthetic k-space data can be obtained by performing a Fourier transform on the at least two sample dealiasing images to obtain at least two sample k-space datasets. Each of the at least two sample k-space datasets may correspond to one of the at least two sample dealiasing images. The synthetic k-space data can be further obtained by applying phase modulation to the sample k-space datasets such that, for at least one of the at least two sample slices, the sample phase of the corresponding sample k-space dataset varies based on the at least two sample acquisition periods and / or the phase encoding direction. The synthetic k-space data can be further obtained by combining the sample k-space datasets corresponding to the same sample acquisition period to obtain at least two second k-space datasets. Each of the at least two second k-space datasets may correspond to one of the at least two sample acquisition periods and the at least two sample slices. The sample k-space data includes the at least two second k-space datasets.

[0019] In some embodiments, the synthetic k-space data may be further obtained by the following steps: applying an undersampling strategy to the at least two second k-space datasets or the at least two sample k-space datasets by replacing a portion of the data in the at least two second k-space datasets or the at least two sample k-space datasets with zeros.

[0020] In some embodiments, training an initial model based on the at least two training datasets to obtain the trained reconstruction model may include: for each of the at least two training datasets, generating at least two sample aliasing images by performing an inverse Fourier transform on the sample k-space data of the training datasets, each of the at least two sample aliasing images corresponding to one of the at least two sample acquisition periods and the at least two sample slices; and training the initial model based on the at least two sample aliasing images to obtain the trained reconstruction model.

[0021] In some embodiments, training an initial model based on the at least two training datasets to obtain the trained reconstruction model may include: obtaining the trained reconstruction model by performing an iterative process comprising one or more iterations. At least one of the one or more iterations may include outputting at least two output images based on sample k-space data from one of the at least two training datasets using an intermediate model. When the current iteration is the first iteration of the iterative process, the intermediate model may include the initial model, or the intermediate model may include an updated model generated in a previous iteration of the current iteration. Each of the at least two output images may correspond to one of the at least two sample dealiasing images in the training dataset. At least one of the one or more iterations may further include updating the intermediate model based on the difference between the at least two output images and the at least two sample dealiasing images in the training dataset.

[0022] In some embodiments, each of the at least two target acquisition periods may correspond to a physiological motion phase of the region of interest. The at least two target images may form a physiological motion movie of the region of interest.

[0023] In some embodiments, the region of interest of the object may include at least a portion of the heart or lungs.

[0024] In some embodiments, the at least two target images can be used for perfusion analysis and indicate changes in contrast agent concentration over time in the at least two target slices.

[0025] According to another aspect of this application, a method for simultaneous multilayer imaging (SMS) is provided, the method comprising one or more of the following steps: One or more processors can acquire target k-space data associated with a region of interest (ROI) of an object. One or more processors can generate at least two target images based on the target k-space data using a trained reconstruction model, wherein each target image corresponds to one of at least two target slices of the ROI and one of at least two target acquisition periods.

[0026] According to another aspect of this application, a system for simultaneous multilayer imaging (SMS) is provided, the system including an acquisition module configured to acquire target k-space data associated with a region of interest (ROI) of an object. The system may further include a reconstruction module configured to generate at least two target images based on the target k-space data, using a trained reconstruction model, wherein each target image corresponds to one of at least two target slices of the ROI and one of at least two target acquisition periods.

[0027] According to another aspect of this application, a non-transitory computer-readable medium may include at least one set of instructions. The at least one set of instructions may be executed by one or more processors of a computing device. The one or more processors may acquire target k-space data associated with a region of interest (ROI) of an object. The one or more processors may, based on the target k-space data, use a trained reconstruction model to generate at least two target images, wherein each target image corresponds to one of at least two target slices of the ROI and one target slice during at least two target acquisition periods.

[0028] One aspect of this application relates to systems and methods for simultaneous multi-layer imaging, and more particularly to systems and methods for reconstructing autocalibrated acquisition data using machine learning models. The autocalibrated acquisition data may include the following characteristics: The autocalibrated acquisition data can be acquired from at least two slices of a region of interest (ROI) of simultaneously excited objects. The autocalibrated acquisition data can be acquired during at least two acquisition periods to reconstruct a series of dealiased images, wherein each dealiased image corresponds to one of the at least two slices and one of the at least two acquisition periods. The autocalibrated acquisition data can be phase-modulated by applying a transport phase. The transport phase varies along the phase encoding direction and / or the time dimension (e.g., during the at least two acquisition periods). Reconstructing the autocalibrated acquisition data using machine learning models is a non-linear method that can improve image quality and reconstruction speed.

[0029] At least two reference datasets can be determined based on automatically calibrated acquisition data (k-space data), which provide dealiasing information, indicating that the automatically calibrated acquisition data itself includes dealiasing information. Therefore, after the automatically calibrated acquisition data is input into a machine learning model, the model can automatically extract the implicit dealiasing information from the data and output a dealiased image. This eliminates the need for additional operations to acquire reference datasets in simultaneous multi-layer imaging reconstruction (e.g., additional computations or scanning operations required to acquire reference datasets), thus improving the efficiency of simultaneous multi-layer imaging reconstruction. Attached Figure Description

[0030] This application will be further described through exemplary embodiments. These exemplary embodiments will be described in detail with reference to the accompanying drawings. These embodiments are non-limiting exemplary embodiments, in which the same numbers in the figures denote similar structures, wherein:

[0031] Figure 1 These are schematic diagrams of exemplary MRI systems according to some embodiments of this application;

[0032] Figure 2 These are schematic diagrams of exemplary MRI scanning devices according to some embodiments of this application;

[0033] Figure 3 These are schematic diagrams of exemplary hardware and / or software components of a computing device according to some embodiments of this application;

[0034] Figure 4 These are schematic diagrams of exemplary hardware and / or software components of a mobile device according to some embodiments of this application;

[0035] Figure 5 This is an exemplary block diagram of a processing device according to some embodiments of this application.

[0036] Figure 6A This is a flowchart illustrating an exemplary process for simultaneous multi-layer imaging according to some embodiments of this application.

[0037] Figure 6B and Figure 6C This is an exemplary schematic diagram illustrating the acquisition of target k-space data according to some embodiments of this application;

[0038] Figure 6D These are exemplary schematic diagrams of trained reconstruction models according to some embodiments of this application;

[0039] Figure 7 This is a flowchart illustrating an exemplary process for obtaining a trained reconstruction model according to some embodiments of this application.

[0040] Figure 8 This is a flowchart illustrating an exemplary training process for obtaining a trained reconstruction model, according to some embodiments of this application.

[0041] Figure 9 These are flowcharts illustrating exemplary model training and exemplary model application phases according to some embodiments of this application;

[0042] Figure 10 This is a flowchart illustrating an exemplary process for MR imaging according to some embodiments of this application. Detailed Implementation

[0043] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. However, those skilled in the art should understand that this application can be implemented without these details. In other instances, to avoid unnecessarily obscuring various aspects of this application, well-known methods, processes, systems, components, and / or circuits have been described at a higher level. It will be apparent to those skilled in the art that various changes can be made to the disclosed embodiments, and the general principles defined in this application can be applied to other embodiments and application scenarios without departing from the principles and scope of this application. Therefore, this application is not limited to the embodiments shown, but conforms to the broadest scope consistent with the scope of the claims.

[0044] The terminology used in this application is for the purpose of describing particular exemplary embodiments only and is not restrictive. The singular forms “a,” “an,” and “the” used in this application may also include the plural forms unless the context explicitly indicates otherwise. The terms “at least two” and “two or more” used in this application may be replaced with “multiple” to indicate the plural form. It should also be understood that the terms “comprising” and “including” as used in this specification only indicate the presence of the stated features, integers, steps, operations, components, and / or parts, but do not exclude the presence or addition of one or more other features, integers, steps, operations, components, parts, and / or combinations thereof.

[0045] It should be understood that the terms “system,” “unit,” “module,” and / or “block” used in this application are a way of distinguishing different components, elements, parts, sections, or assemblies at different levels in ascending order. However, these terms may be replaced with other expressions if the same purpose can be achieved.

[0046] These and other features, characteristics, functions and operating methods of related structural elements, as well as component assembly and manufacturing economics, will become more apparent from the following description of the accompanying drawings, which form part of this application specification. However, it should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of this application. It should also be understood that the drawings are not drawn to scale.

[0047] This application provides systems and components for medical imaging and / or medical treatment. In some embodiments, the medical system may include an imaging system. The imaging system may include a single-modal imaging system and / or a multimodal imaging system. A single-modal imaging system may include, for example, a magnetic resonance imaging (MRI) system. For example, an MRI system may include a superconducting MRI system, a non-superconducting MRI system, etc. A multimodal imaging system may include, for example, a computed tomography-magnetic resonance imaging (MRI-CT) system, a positron emission tomography-magnetic resonance imaging (PET-MRI) system, a single-photon emission computed tomography-magnetic resonance imaging (SPECT-MRI) system, a digital subtraction angiography-magnetic resonance imaging (DSA-MRI) system, etc. In some embodiments, the medical system may include a treatment system. The treatment system may include a treatment planning system (TPS), an image-guided radiotherapy (IGRT) system, etc. An image-guided radiotherapy (IGRT) system may include a treatment device and an imaging device. The treatment device may include a linear accelerator, a cyclotron accelerator, a synchrotron, etc., configured to perform radiotherapy on a subject. Treatment devices may include accelerators for various particles, including, for example, photons, electrons, protons, or heavy ions. Imaging devices may include MRI scanners, CT scanners (e.g., cone-beam computed tomography (CBCT) scanners), digital radiography (DR) scanners, electron field imaging devices (EPID), etc.

[0048] Figure 1 This is a schematic diagram of an exemplary MRI system according to some embodiments of this application. As shown, the magnetic resonance imaging system 100 may include a scanning device 110, a network 120, a user device 130, a processing device 140, and a storage device 150. The components of the magnetic resonance imaging system 100 may be connected in one or more ways. This is merely an example. Figure 1 As shown, scanning device 110 can be connected to processing device 140 via network 120. As another example, scanning device 110 can be directly connected to processing device 140 (as shown by the dashed double-headed arrow connecting scanner 110 and processing device 140). As another example, storage device 150 can be connected to processing device 140 directly or via network 120. As yet another example, user device 130 (e.g., 131, 132, 133, etc.) can be directly (as shown by the dashed double-headed arrow connecting user device 130 and processing device 140) or via network 120 to processing device 140.

[0049] The scanning device 110 can scan an object located within its detection area and generate at least two imaging data related to the object. In this application, the terms "object" and "target" are used interchangeably. For example, an object may include a patient, an artificial object, etc. As another example, an object may include a specific part, organ, and / or tissue of a patient. For instance, an object may include the head, brain, neck, body, shoulder, arm, chest, heart, stomach, blood vessels, soft tissue, knee joint, foot, etc., or any combination thereof.

[0050] In some embodiments, the scanning device 110 may include a magnetic resonance imaging scanner, a multimodal device, etc. Exemplary multimodal devices may include MRI-CT devices, PET-MRI devices, etc. In some embodiments, the magnetic resonance imaging scanner may be a closed-aperture scanner or an open-aperture scanner. In this application, as... Figure 1 As shown, the X-axis, Y-axis, and Z-axis can form an orthogonal coordinate system. Figure 1 The X and Z axes shown can be horizontal, and the Y axis can be vertical. As shown in the figure, the positive direction of the X axis can be the direction from the right side to the left side of the scanning device 110 when looking at the front of the scanning device 110. Figure 1 The positive direction of the Y-axis shown can be from the bottom to the top of the scanner 110; Figure 1 The positive direction of the Z-axis shown can refer to the direction in which the object moves out of the scanning channel (or scanning aperture) of the scanning device 110. Further description of the scanner 110 can be found elsewhere in this application. See, for example, [link to relevant documentation]. Figure 2 And its description.

[0051] Network 120 may include any suitable network that can facilitate the exchange of information and / or data between the MRI system 100. In some embodiments, one or more components of the MRI system 100 (e.g., scanning device 110, user device 130, processing device 140, or storage device 150) may communicate information and / or data with one or more other components of the MRI system 100 via network 120. For example, processing device 140 may obtain magnetic resonance (MR) data (also referred to as MR signals, echo signals, or echo data) from scanning device 110 via network 120. As another example, user device 130 and / or storage device 150 may obtain one or more images from processing device 140. In some embodiments, network 120 may be a wired network or a wireless network, or any combination thereof. Network 120 may be and / or include public networks (e.g., the Internet), private networks (e.g., local area networks (LANs), wide area networks (WANs), etc.), wired networks (e.g., Ethernet), wireless networks (e.g., 802.11 networks, Wi-Fi networks, etc.), cellular networks (e.g., Long Term Evolution (LTE) networks), Frame Relay networks, Virtual Private Networks (VPNs), satellite networks, telephone networks, routers, hubs, switches, server computers, and / or any combination thereof. By way of example only, network 120 may include cable networks, wired networks, fiber optic networks, telecommunications networks, intranets, wireless local area networks (WLANs), local area networks (MANs), public switched telephone networks (PSTNs), Bluetooth, etc. TM Network, ZigBee TM A network, a near-field communication (NFC) network, or any combination thereof. In some embodiments, network 120 may include one or more network access points. For example, network 120 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 system 100 may connect to network 120 to exchange data and / or information.

[0052] User equipment 130 may include mobile device 131, tablet computer 132, laptop computer 133, desktop computer (not shown), workstation (not shown), etc., or any combination thereof. In some embodiments, mobile device 131 may include smart home devices, wearable devices, smart mobile devices, virtual reality devices, augmented reality devices, etc., or any combination thereof. In some embodiments, smart home devices may include smart lighting devices, smart appliance control devices, smart monitoring devices, smart TVs, smart cameras, walkie-talkies, etc., or any combination thereof. In some embodiments, wearable devices may include smart bracelets, smart shoes, smart glasses, smart helmets, smartwatches, smart clothing, smart backpacks, smart accessories, etc., or any combination thereof. In some embodiments, smart mobile devices may include smartphones, personal digital assistants (PDAs), gaming devices, navigation devices, point-of-sale (POS) devices, etc., or any combination thereof. In some embodiments, virtual reality devices and / or augmented reality devices may include virtual reality helmets, virtual reality glasses, virtual reality goggles, augmented reality helmets, augmented reality glasses, augmented reality goggles, etc., or any combination thereof. For example, virtual reality devices and / or augmented reality devices may include Google... TM Glass, Oculus Rift, HoloLens, Gear VR, etc. In some embodiments, user device 130 can remotely operate scanning device 110 and / or processing device 140. In some embodiments, user device 130 can operate scanning device 110 and / or processing device 140 via a wireless connection. In some embodiments, user device 130 can receive user-inputted information and / or instructions, and send the received information and / or instructions to scanning device 110 or processing device 140 via network 120. For example, a user of the magnetic resonance imaging system 100 (e.g., a doctor, technician, or engineer) can set a scanning protocol through user device 130. User device 130 can send a scanning protocol to processing device 140 to instruct processing device 140 to cause scanning device 110 (e.g., a magnetic resonance imaging scanner) to operate according to the scanning protocol. In some embodiments, user device 130 can receive data and / or information from processing device 140 and / or storage device 150. For example, user device 130 can obtain one or more images from processing device 140 and / or storage device 150.

[0053] Processing device 140 can process data and / or information obtained from scanning device 110, user device 130, and / or storage device 150. For example, processing device 140 can obtain magnetic resonance data from scanning device 110 and determine one or more images based on the magnetic resonance data. As another example, processing device 140 can receive one or more instructions from user device 130 and cause scanning device 110 to operate according to the one or more instructions. In some embodiments, processing device 140 can be a single server or a group of servers. The server group can be centralized or distributed. In some embodiments, processing device 140 can be local or remote. For example, processing device 140 can access information and / or data stored in or obtained by scanning device 110, user device 130, and / or storage device 150 via network 120. As another example, processing device 140 can be directly connected to scanning device 110 (e.g., ...). Figure 1 The dashed double-headed arrows connecting the processing device 140 and the scanner 110 indicate that the user equipment 130 (as shown in the image) is connected to the processing device 140 and the scanner 110. Figure 1 The processing device 140 (shown by a dashed double-headed arrow connecting the processing device 140 and the user device 130) and / or storage device 150 access stored or retrieved information and / or data. In some embodiments, the processing device 140 may be implemented on a cloud platform. By way of example only, the cloud platform may include a private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, internal cloud, multi-tiered cloud, etc., or any combination thereof. In some embodiments, the processing device 140 may be implemented in a cloud environment as described in this disclosure. Figure 3 Implemented on a computing device 300 containing one or more of the components shown.

[0054] Storage device 150 may store data and / or instructions. In some embodiments, storage device 150 may include a database, image archive and communication system, file system, etc., or any combination thereof. In some embodiments, storage device 150 may store data obtained from scanning device 110, user device 130, and / or processing device 140. For example, storage device 150 may store magnetic resonance data acquired by scanning device 110. As another example, storage device 150 may store medical images (e.g., magnetic resonance images) generated by processing device 140 and / or user device 130. As another example, storage device 150 may store preset scanning parameters (e.g., preset scanning protocol) of magnetic resonance imaging system 100. In some embodiments, storage device 150 may store data and / or instructions that processing device 140 may execute or use to perform the exemplary methods described in this application. For example, storage device 150 may store instructions that processing device 140 may execute to cause scanning device 110 to acquire magnetic resonance data based on a pulse sequence including a steady-state sequence and an acquisition sequence. As another example, storage device 150 may store instructions executable by processing device 140 and / or user device 130 to generate one or more images based on magnetic resonance data. In some embodiments, storage device 150 may include a mass storage device, a removable storage device, a volatile read-write memory, a read-only memory (ROM), or any combination thereof. Exemplary mass storage devices may include disks, optical disks, solid-state drives, etc. Exemplary removable storage devices may include flash drives, floppy disks, optical disks, memory cards, compact disks, magnetic tapes, etc. Exemplary volatile read-write memory may include random access memory (RAM). Exemplary RAM may include dynamic random access memory (DRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), static random access memory (SRAM), thyristor random access memory (T-RAM), and zero-capacitance random access memory (Z-RAM), etc. Exemplary ROM may include mask ROM (MROM), programmable ROM (PROM), erasable programmable ROM (PEROM), electrically erasable programmable ROM (EEPROM), optical disk ROM (CD-ROM), and digital multifunction disk ROM, etc. In some embodiments, the storage device 150 can be implemented on a cloud platform. As an example only, the cloud platform may include private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, internal cloud, multi-tiered cloud, etc., or any combination thereof.

[0055] In some embodiments, storage device 150 may be connected to network 120 to communicate with one or more components of magnetic resonance imaging system 100 (e.g., scanning device 110, processing device 140, user device 130, etc.). One or more components of magnetic resonance imaging system 100 may access data or instructions stored in storage device 150 via network 120. In some embodiments, storage device 150 may be directly connected to or communicate with one or more components of magnetic resonance imaging system 100 (e.g., scanning device 110, processing device 140, user device 130, etc.). In some embodiments, storage device 150 may be part of processing device 140.

[0056] In some embodiments, the magnetic resonance imaging system 100 may further include one or more power supplies connected to one or more components (e.g., scanning device 110, processing device 140, user device 130, storage device 150, etc.) of the magnetic resonance imaging system 100. Figure 1 (Not shown in the image).

[0057] Figure 2 This is a schematic diagram of an exemplary MRI scanning apparatus according to some embodiments of this application. As shown, a main magnet 201 can generate a first magnetic field (or main magnetic field) applicable to an object (also called a target) exposed in the field. The main magnet 201 may include a resistive magnet or a superconducting magnet, both of which require a power source (not shown) to operate. Alternatively, the main magnet 201 may include a permanent magnet. The main magnet 201 may include a scanning aperture for placing the object. The main magnet 201 can also control the uniformity of the generated main magnetic field. Several shimming coils may be located within the main magnet 201. The shimming coils placed in the gaps of the main magnet 201 can compensate for non-uniformity of the magnetic field of the main magnet 201. The shimming coils can be charged by a shimming power source.

[0058] Gradient coil 202 can be located within main magnet 201. Gradient coil 202 can generate a second magnetic field (or gradient field, including gradient fields Gx, Gy, and Gz). The second magnetic field can be superimposed on the main magnetic field generated by main magnet 201 and distort the main magnetic field, so that the magnetic orientation of the protons of the object can change with their position in the gradient field, thereby encoding spatial information into a magnetic resonance signal generated by the imaged object. Gradient coil 202 may include an X coil (e.g., for generating a magnetic field gradient Gx corresponding to the X direction), a Y coil (e.g., for generating a magnetic field gradient Gy corresponding to the Y direction), and / or a Z coil (e.g., for generating a magnetic field gradient Gz corresponding to the Z direction). Figure 2(Not shown in the image). In some embodiments, the Z coil may be based on a Maxwell coil design, while the X and Y coils may be based on a Golay coil configuration. The three sets of coils can generate three different magnetic fields for position encoding. Gradient coil 202 allows spatial encoding of the magnetic resonance signal for image reconstruction. Gradient coil 202 may be connected to one or more of the X gradient amplifier 204, Y gradient amplifier 205, or Z gradient amplifier 206. One or more of the three amplifiers may be connected to waveform generator 216. Waveform generator 216 can generate gradient waveforms applied to X gradient amplifier 204, Y gradient amplifier 205, and / or Z gradient amplifier 206. The amplifiers can amplify the waveforms. The amplified waveforms can be applied to one of the coils in gradient coil 202 to generate magnetic fields along the X, Y, or Z axes, respectively. Gradient coil 202 may be designed for use with closed-aperture MRI scanners or open-aperture MRI scanners. In some cases, all three sets of coils in gradient coil 202 may be energized, thereby generating three gradient fields. In some embodiments of this application, the X coil and Y coil can be energized to generate gradient fields in the X and Y directions. As used herein, Figure 2 The description of the X-axis, Y-axis, Z-axis, X direction, Y direction, and Z direction is consistent with... Figure 1 The same or similar as described in the text.

[0059] In some embodiments, the radio frequency (RF) coil 203 may be located within the main magnet 201 and function as a transmitter, receiver, or both. The RF coil 203 may be connected to an RF electronics device 209, which may be configured or used as one or more integrated circuits (ICs) serving as waveform transmitters and / or waveform receivers. The RF electronics device 209 may be connected to an RF power amplifier (RFPA) 207 and an analog-to-digital converter (ADC) 208.

[0060] When used as a transmitter, the RF coil 203 can generate an RF signal that provides a third magnetic field, which is used to generate an RF signal related to the object being imaged. The third magnetic field can be perpendicular to the main magnetic field. The waveform generator 216 can generate RF pulses. These RF pulses can be amplified by the RF power amplifier 207, processed by the RF electronics 209, and applied to the RF coil 203 to generate an RF signal in response to a strong current generated by the RF electronics 209 based on the amplified RF pulses.

[0061] When used as a receiver, the radio frequency (RF) coil may be responsible for detecting the magnetic resonance signal (e.g., echo). After excitation, the RF coil 203 can sense the magnetic resonance signal generated by the object. The sensed magnetic resonance signal can then be received from the RF coil 203, amplified, and provided to the analog-to-digital converter (ADC) 208. The ADC 208 can convert the magnetic resonance signal from an analog signal to a digital signal. The digital magnetic resonance signal can then be filled into the k-space.

[0062] In some embodiments, gradient coil 202 and RF coil 203 may be positioned circumferentially relative to the object. Those skilled in the art will understand that the main magnet 201, gradient coil 202, and RF coil 203 may be located in various configurations around the object.

[0063] In some embodiments, the RF power amplifier 207 can amplify RF pulses (e.g., the power of the RF pulses, the voltage of the RF pulses), and the amplified RF pulses generated are used to drive the RF coil 203. In some embodiments, the RF power amplifier 207 may include one or more RF power amplifiers.

[0064] Magnetic resonance imaging (MRI) systems (e.g., MRI system 100 of this invention) are generally used to acquire internal images of an object (e.g., a patient) from a specific region of interest, which can be used for purposes such as diagnosis, treatment, or a combination thereof. An MRI system includes a master magnet assembly (e.g., master magnet 201) for providing a master magnetic field to align the individual magnetic moments of hydrogen atoms within the patient's body. In this process, hydrogen atoms oscillate around their magnetic poles at their characteristic Larmor frequencies. If the object is subjected to an additional magnetic field tuned to the Larmor frequency, the hydrogen atoms absorb additional energy, causing the net alignment torque of the hydrogen atoms to rotate. The additional magnetic field can be provided by a radio frequency excitation signal (e.g., a radio frequency signal generated by radio frequency coil 203). When the additional magnetic field is removed, the magnetic moments of the hydrogen atoms rotate back to their alignment with the master magnetic field, thereby emitting a magnetic resonance signal. The acquired magnetic resonance signal can be digitized and filled into k-space. One or more images can be generated based on the k-space data. In this application, the terms "magnetic resonance data," "magnetic resonance signal," "echo," "echo data," and "echo signal" are used interchangeably.

[0065] If the main magnetic field is uniformly distributed throughout the patient's body, the radio frequency excitation signal may non-selectively excite all hydrogen atoms in the object. Therefore, to image a specific part of the patient's body, X, Y, and Z directions with specific timing, frequency, and phase can be superimposed on the magnetic field (e.g., with...). Figure 1The magnetic field gradients Gx, Gy, and Gz (e.g., generated by gradient coil 202) on the X, Y, and Z axes (which are the same or similar in the image) cause the radio frequency excitation signal to excite hydrogen atoms in one or more target sheets of the patient's body, and encode unique phase and frequency information in the magnetic resonance signal according to the position of the hydrogen atoms in the "image sheet".

[0066] Figure 3 These are schematic diagrams of exemplary hardware and / or software components of a computing device according to some embodiments of this application. Figure 3 As shown, the computing device 300 may include a processor 310, a memory 320, an input / output (I / O) 330, and a communication port 340.

[0067] Processor 310 can execute computer instructions (program code) and perform the functions of processing device 140 according to the techniques described in this application. Computer instructions may include routines, programs, objects, components, signals, data structures, procedures, modules, and functions that perform specific functions described in this application. For example, processor 310 may generate one or more images based on magnetic resonance data. In some embodiments, processor 310 may include a microcontroller, microprocessor, reduced instruction set computer (RISC), application-specific integrated circuit (ASIC), application-specific instruction set processor (ASIP), central processing unit (CPU), graphics processing unit (GPU), physical processing unit (PPU), microcontroller unit, digital signal processor (DSP), field-programmable gate array (FPGA), advanced RISC machine (ARM), programmable logic device (PLD), any circuit or processor capable of performing one or more functions, or any combination thereof.

[0068] For illustrative purposes only, only one processor is described in computing device 300. However, it should be noted that computing device 300 in this application may also include multiple processors, and therefore, the operation of the method executed by one processor as described in this application may also be executed jointly or individually by multiple processors. For example, if the processors of computing device 300 in this application simultaneously execute operations A and B, it should be understood that operations A and B may also be executed jointly or individually by two different processors in computing device 300 (e.g., the first processor executes operation A, the second processor executes operation B, or the first and second processors jointly execute operations A and B).

[0069] Memory 320 may store data / information obtained from scanning device 110, user equipment 130, storage device 150, or any other component of magnetic resonance imaging system 100. In some embodiments, memory 320 may include mass storage devices, removable storage devices, volatile read-write memory, read-only memory (ROM), etc., or any combination thereof. For example, mass storage devices may include disks, optical disks, solid-state drives, etc. Removable storage devices may include flash drives, floppy disks, optical disks, memory cards, compact disks, magnetic tapes, etc. Volatile read-write memory may include random access memory (RAM). RAM may include dynamic RAM (DRAM), dual date rate synchronous dynamic RAM (DDRSDRAM), static RAM (SRAM), thyristor RAM (T-RAM), and zero-capacitance RAM (Z-RAM), etc. Read-only memory may include mask read-only memory, programmable read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, optical disk read-only memory, and digital universal disk read-only memory, etc. In some embodiments, memory 320 may store one or more programs and / or instructions to perform the exemplary methods described in this application. For example, memory 320 may store a program for processing device 140 to generate one or more images based on magnetic resonance data. In some embodiments, memory 320 may store one or more reconstructed magnetic resonance images.

[0070] Input / output (I / O) 330 can input or output signals, data, or information. In some embodiments, input / output 330 allows a user to interact with processing device 140. In some embodiments, input / output 330 may include input devices and output devices. Exemplary input devices may include a keyboard, mouse, touchscreen, microphone, trackball, etc., or combinations thereof. Exemplary output devices may include display devices, speakers, printers, projectors, etc., or combinations thereof. Exemplary display devices may include liquid crystal displays (LCDs), light-emitting diodes (LEDs), screen-based displays, flat panel displays, curved screens, television equipment, cathode ray tubes (CRTs), etc., or combinations thereof.

[0071] Communication port 340 can be connected to a network (e.g., network 120) to facilitate data communication. Communication port 340 can establish a connection between processing device 140 and scanning device 110, user equipment 130, or storage device 150. This connection can be a wired connection, a wireless connection, or a combination of both to enable data transmission and reception. Wired connections can include cables, optical fibers, telephone lines, etc., or any combination thereof. Wireless connections can include Bluetooth, Wi-Fi, WiMAX, WLAN, ZigBee, mobile networks (e.g., 3G, 4G, 5G, etc.), etc., or combinations thereof. In some embodiments, communication port 340 can be a standardized communication port, such as RS232, RS485, etc. In some embodiments, communication port 340 can be a specially designed communication port. For example, communication port 340 can be designed according to the Digital Imaging and Medical Communications (DICOM) protocol.

[0072] Figure 4 These are schematic diagrams illustrating exemplary hardware and / or software components of a mobile device according to some embodiments of this application. Figure 4 As shown, the mobile device 400 may include a communication platform 410, a display 420, a graphics processing unit (GPU) 430, a central processing unit (CPU) 440, an input / output 450, memory 460, and a storage 490. In some embodiments, any other suitable components, including but not limited to a system bus or controller (not shown), may also be included in the mobile device 400. In some embodiments, a mobile operating system 470 (e.g., iOS, Android, Windows Phone, etc.) and one or more applications 480 may be loaded from storage 490 into memory 460 for execution by the CPU 440. Application 480 may include a browser or any other suitable mobile application for receiving and presenting information related to image processing or other information from processing device 140. User interaction with the information stream can be achieved through input / output 450 and provided to processing device 100 and / or other components of magnetic resonance imaging system 100 via network 120.

[0073] In some embodiments, input / output 450 may include input devices and output devices. Exemplary input devices may include a keyboard, mouse, touchscreen, microphone, trackball, etc., or combinations thereof. Exemplary output devices may include display devices, speakers, printers, projectors, etc., or combinations thereof. Exemplary display devices may include liquid crystal displays (LCDs), light-emitting diodes (LEDs), screen-based displays, flat panel displays, curved screens, television equipment, cathode ray tubes (CRTs), etc., or combinations thereof.

[0074] To implement the various modules, units, and functions described in this application, a computer hardware platform may be used as the hardware platform for one or more of the components described herein. The hardware components, operating system, and programming language of such a computer are conventional and presumably well-known to those skilled in the art in adapting those technologies to the blood pressure monitoring described herein. A computer with user interface elements may be used to implement a personal computer (PC) or another type of workstation or terminal device, but if the computer is properly programmed, it may also act as a server. It is understood that those skilled in the art should be familiar with the structure, programming, and general operation of this computer device; therefore, the figures should be readily apparent to them.

[0075] Figure 5 This is an exemplary block diagram of a processing device according to some embodiments of this application. In some embodiments, the processing device 140 may include an acquisition module 520 and a reconstruction module 530. In some embodiments, the processing device 140 may also include a control module 510.

[0076] The acquisition module 520 can acquire target k-space data related to the region of interest of an object.

[0077] The reconstruction module 530 can generate one or more target images based on target k-space data and using a trained reconstruction model. The trained reconstruction model can be a machine learning model.

[0078] In some embodiments, the reconstruction module 530 may generate a target image corresponding to a target slice of the region of interest based on target k-space data using a trained reconstruction model. In some embodiments, the reconstruction module 530 may generate at least two target images (time frames) based on target k-space data using a trained reconstruction model, each target image (time frame) corresponding to the same target slice of the region of interest and a different target acquisition period. In some embodiments, the reconstruction module 530 may generate at least two target images based on target k-space data using a trained reconstruction model, each target image corresponding to one of at least two target slices of the region of interest. In some embodiments, the reconstruction module 530 may generate at least two target images (time frames) based on target k-space data using a trained reconstruction model, each target image (time frame) corresponding to one of at least two target slices of the region of interest and one of at least two target acquisition periods.

[0079] The control module 510 enables the scanning device 110 to scan the region of interest (ROI) of the object. In some embodiments, the control module 510 enables the scanning device 110 to apply one or more multi-band excitation radio frequency (RF) pulses to the ROI to simultaneously excite at least two target slices in the ROI once or more. In some embodiments, the control module 510 can apply a transfer phase to at least one excited target slice by the magnetic resonance imaging (MRI) device, causing the MRI device to perform phase modulation on the at least one excited target slice, wherein the transfer phase varies with the acquisition period and / or phase encoding direction of the at least two targets.

[0080] The acquisition module 520 can acquire target k-space data from the at least two excitation target slices of the region of interest during at least two target acquisitions.

[0081] The modules in processing device 140 can connect to or communicate with each other via wired or wireless connections. Wired connections can include metal cables, optical fibers, hybrid cables, etc., or any combination thereof. Wireless connections can include local area networks (LANs), wide area networks (WANs), Bluetooth, Wi-Fi, near field communication (NFC), etc., or any combination thereof. Two or more modules can be combined into one module, and any module can be split into two or more units. For example, reconstruction module 530 can be divided into a first unit configured as a trained reconstruction model and a second unit configured to perform image reconstruction using the trained reconstruction model.

[0082] It should be noted that the above description is provided for illustrative purposes only and is not intended to limit the scope of this application. Various changes and modifications can be made by those skilled in the art based on the description herein. However, these changes and modifications do not depart from the scope of this application. For example, the processing device 140 may also include a storage module ( Figure 5 (Not shown in the image). The storage module can be configured to store data generated during any processing performed by any component of the processing device 140. As another example, each component of the processing device 140 may include a storage device. Optionally, the components of the processing device 140 may share a common storage device.

[0083] Figure 6A This is a flowchart illustrating an exemplary process for simultaneous multi-layer imaging according to some embodiments of this application. In some embodiments, process 600 can be performed... Figure 1 This is implemented in the magnetic resonance imaging system 100 shown. For example, process 600 can be stored as instructions in a storage device (e.g., storage device 150 or memory 320 of processing device 140), and can be processed by processing device 140 (e.g., processor 310 of processing device 140, or...). Figure 5The process 600 shown below is invoked and / or executed by one or more modules of the processing device 140. The operation of the process 600 is presented below for illustrative purposes. In some embodiments, process 600 may be accomplished using one or more additional operations not described and / or one or more operations not discussed. Additionally, as... Figure 6A The order of operations of process 600 shown and described below is not intended to be restrictive.

[0084] In 610, the processing device 140 (e.g., control module 510) can cause a magnetic resonance imaging device (e.g., scanning device 110) to apply one or more excitation radio frequency (RF) pulses to a region of interest (ROI) of a subject (e.g., a patient) to simultaneously excite at least two (or more) target slices of the ROI once or more.

[0085] In some embodiments, the RF coil 203 can generate one or more excitation RF pulses, wherein each excitation RF pulse is applied to the region of interest to simultaneously excite at least two target layers of the region of interest once. The excitation RF pulses can be applied in the presence of a layer selection gradient to generate lateral magnetization in at least two target layers of the region of interest.

[0086] In some embodiments, the excitation RF pulse may be a composite RF pulse, comprising at least two excitation bands, each excitation band being used to excite one of the at least two target layers. For example, the excitation bands corresponding to different layer positions and / or different thicknesses of the target layers may have different frequency values ​​and bandwidths in order to excite the corresponding target layers.

[0087] In 620, for at least one excited target slice, processing device 140 (e.g., control module 510) can phase modulate the at least one excited target slice by causing magnetic resonance imaging device to apply a transmit phase to the at least one excited target slice, the transmit phase varying with at least two target acquisition periods and / or phase encoding direction.

[0088] In some embodiments, the target acquisition period can refer to the period of acquiring target k-space data corresponding to an aliased image. The aliased image is an aliased image of at least two target slices of the region of interest. In some embodiments, in simultaneous multi-layer imaging, k-space data corresponding to at least two target slices can be acquired simultaneously. Therefore, the target acquisition period can also refer to the period of acquiring target k-space data corresponding to a target dealiased image (a detailed description of a “target dealiased image” can be found in the description of operation 640). The target dealiased image corresponds to a target slice. For example, the process 600 for simultaneous multi-layer imaging can be used to generate a physiological motion movie of the region of interest (e.g., a movie of cardiac motion or lung respiratory motion). In this case, a series of dealiased images (e.g., time frames) can be generated, where each image corresponds to a target slice and depicts a motion phase (e.g., a cardiac phase, such as diastole or systole, or a respiratory phase, such as the end of expiration or the end of inspiration). A target acquisition period can correspond to a motion phase. As another example, the procedure 600 for simultaneous multilayer imaging can be used for perfusion analysis, reflecting changes in contrast agent concentration over time in at least two target slices. In this case, a series of dealiased images (e.g., time frames) can be generated, where each image indicates the concentration of contrast agent in a target slice at a certain time point. A target acquisition period may correspond to a time point on the contrast agent's time-concentration curve.

[0089] In some embodiments, the transport phase applied to a target slice may vary over at least two target acquisition periods. For example, the applied transport phase may be different for different target dealiasing images corresponding to the same target slice. In some embodiments, the transport phase applied to a target slice may vary along the phase coding direction. In some embodiments, at least one target slice may not undergo phase modulation, for example, no transport phase may be applied. In some embodiments, the transport phase applied to different target slices may be different.

[0090] In some embodiments, the transmission phase applied to a target slice may include at least two phase elements, which are applied to the target slice during at least two target acquisitions. For example, a phase element It can be applied to the target sheet s such that the magnetic resonance data (e.g., echoes) generated from the target sheet s and filled into the phase encoding line l during target acquisition p has a phase element corresponding to the phase element. The phase value. In some embodiments, the phase elements of two phase encoding lines with adjacent acquisition sequences corresponding to the same target slice and the same target acquisition period may be different. For example, It can be different In some embodiments, phase elements corresponding to the same target slice and the same target acquisition period can be dynamically cycled according to the phase encoding lines, for example... Where Δl can be any integer, for example, equal to the number of at least two target slices. In some embodiments, the phase elements corresponding to the same target slice, the same phase coding line, and two adjacent target acquisition periods can be different. For example, It can be different In some embodiments, phase elements corresponding to the same target slice and the same phase coding line can be dynamically cycled according to the acquisition periods of the at least two targets, for example... Where Δp can be any integer, for example, equal to the number of the at least two target layers. In some embodiments, Where Δ can be any integer.

[0091] In some embodiments, the transmission phase applied to at least two target layers may vary differently in the phase encoding direction and / or during at least two target acquisitions.

[0092] In some embodiments, since the transmission phase varies with the phase encoding direction and the at least two target acquisition periods, the same target layer will exhibit different degrees of field of view (FOV) shift in the image domain after the transmission phase is applied to the target layer during different target acquisition periods. In some embodiments, the FOV shifts of target layers corresponding to two adjacent target acquisition periods may be different. In some embodiments, the FOV shifts of target layers corresponding to the at least two target acquisition periods can be dynamically cycled according to the at least two target acquisition periods. For example, the FOV shift of the target layer corresponding to target acquisition period P1 is FOV / 3, the FOV shift of the target layer corresponding to target acquisition period P2 is 2FOV / 3, the FOV shift of the target layer corresponding to target acquisition period P3 is FOV / 3, and so on. In some embodiments, different phase modulations are applied to different target layers corresponding to the same target acquisition period, resulting in different FOV shifts. For example, the FOV shift of target layer S1 corresponding to target acquisition period P1 is FOV / 3, and the FOV shift of target layer S2 corresponding to target acquisition period P1 is 2FOV / 3.

[0093] In some embodiments, phase modulation of at least one target slice can be achieved by equipping one or more excitation radio frequency pulses with phase elements that transmit phase. For example, for an excitation band used to excite a target slice, phase elements corresponding to the phase encoding line, the target slice, and the target acquisition period can be provided.

[0094] Detailed information about the transmission phase can be found in the reference “Ferrazzi G, et al., Autocalibrated multiband CAIPIRINHA with through-time encoding: Proof of principle and application to cardiac tissue phase mapping, Magn Reson Med. 2019 Feb; 81(2):1016-1030”, which is incorporated herein by reference.

[0095] In 630, processing device 140 (e.g., acquisition module 520) can acquire target k-space data (also known as autocalibrated acquisition data) from at least two excitation target sheets of the region of interest during at least two target acquisitions.

[0096] In some embodiments, following nuclear spin excitation in the at least two target sheets, a series of echo signals can be generated from the at least two excited target sheets. Target k-space data can be obtained by filling the k-space with the echo signals under the action of at least two readout gradients (also known as frequency encoding gradients) and at least two phase encoding gradients. The readout gradients can be used to fill the k-space with the echo signals along the frequency encoding direction. Spatial encoding of the echo signals along the phase encoding direction can be performed using the phase encoding gradient. The phase encoding gradient can be used to determine the filling position (e.g., phase encoding line) of the echo signals in k-space along the phase encoding direction.

[0097] In some embodiments, the k-space may be undersampled, fully sampled, or oversampled to obtain target k-space data. In some embodiments, the acquisition trajectory of the target k-space data may include a Cartesian trajectory or a non-Cartesian trajectory, such as a radial line, a spiral, etc.

[0098] In some embodiments, the pulse sequence may include an excitation pulse, a layer selection gradient, one or more readout gradients, and one or more phase-encoded gradients. The pulse sequence may be repeated once or more to acquire target k-space data. In some embodiments, the pulse sequence may include gradient echo (GRE), equilibrium steady-state free precession (bSSFP), fast spin echo (FSE), echo plane imaging (EPI), etc.

[0099] In some embodiments, the target k-space data may include at least two first k-space datasets, each first k-space dataset corresponding to the at least two target slices and a target acquisition period. The first k-space datasets may be aliased k-space datasets, comprising k-space data acquired from the at least two target slices during a target acquisition period. Aliased images corresponding to the at least two target slices during the target acquisition period can be obtained by performing an inverse Fourier transform on the first k-space datasets.

[0100] In some embodiments, after simultaneously exciting the at least two target slices, at least two sets of echo signals can be sequentially generated from the at least two excited target slices. Each set of echo signals may include at least two echo signals generated simultaneously from the at least two excited target slices. Under the influence of the readout gradient and the phase encoding gradient, the echo signals generated simultaneously from the at least two target slices can be filled into the same position in k-space (e.g., the same phase encoding line). Thus, when k-space filling is complete, the acquired k-space data can be referred to as the first k-space dataset of the target k-space data. The period for acquiring the first k-space dataset can be referred to as the target acquisition period.

[0101] In some embodiments, at least two first k-space datasets can be acquired sequentially. For example, after acquiring one first k-space dataset, the acquisition of another first k-space dataset can begin. In some embodiments, at least two first k-space datasets can be acquired simultaneously. For example, in cardiac cine imaging, during a first cardiac motion cycle, echo signals corresponding to the first cardiac motion phase can fill the first k-space, and echo signals corresponding to the second cardiac motion phase following the first cardiac motion phase can fill the second k-space. During the second cardiac motion cycle following the first cardiac motion cycle, echo signals corresponding to the first cardiac motion cycle can fill the first k-space, and echo signals corresponding to the second cardiac motion phase can fill the second k-space. Thus, first k-space dataset 1 can be obtained by filling the first k-space, and first k-space dataset 2 can be obtained by filling the second k-space. First k-space dataset 1 and first k-space dataset 2 can be acquired simultaneously. First k-space dataset 1 can correspond to the first cardiac time phase, and first k-space dataset 2 can correspond to the second cardiac time phase.

[0102] In some embodiments, the period between two adjacent excitation radio frequency pulses may be referred to as the repetition time (TR). There can be any correspondence between the repetition time and the target acquisition period. For example, a repetition time may include one or more target acquisition periods, meaning that one or more first k-space datasets can be acquired within one repetition time. As another example, different portions of a target acquisition period may be distributed across multiple repetition times, meaning that the first k-space dataset corresponding to the target acquisition period can be acquired across multiple repetition times.

[0103] Figure 6B and Figure 6C This is an exemplary schematic diagram illustrating the acquisition of target k-space data according to some embodiments of this application.

[0104] like Figure 6B As shown, three target slices s1, s2, and s3 are simultaneously excited. Target k-space data can be acquired from s1, s2, and s3 during the first target acquisition period P1 and the second target acquisition period P2. The target k-space data may include a first k-space dataset D1 and a first k-space dataset D2. The first k-space dataset D1 is acquired in P1 by filling the first k-space K1 with echoes generated from s1, s2, and s3, and the first k-space dataset D2 is acquired in P2 by filling the first k-space K2 with echoes generated from s1, s2, and s3. Slice s1 may not be phase-modulated in P1 and P2, for example, no transport phase may be applied. Slice s2 may be phase-modulated by applying a first transport phase, which varies in both the spatial dimension (e.g., phase encoding direction) and the temporal dimension (e.g., P1 and P2). Slice s3 may be phase-modulated by applying a second transport phase, which also varies in both the spatial dimension (e.g., phase encoding direction) and the temporal dimension (e.g., P1 and P2). The first transmission phase can be different from the second transmission phase.

[0105] In some embodiments, the first k-space dataset D1 and the first k-space dataset D2 can be obtained sequentially. For example, as shown... Figure 6B As shown, after s1, s2, and s3 are excited simultaneously, three echo signals can be generated simultaneously from s1, s2, and s3, respectively. and echo signal It can be without any phase element applied. Phase elements can be applied Phase elements can be applied Under the influence of the readout gradient and the phase encoding gradient, it is possible to achieve the desired result without phase modulation. Having phase value of and having phase value of The signal is filled into the same position in the first k-space K1 (e.g., phase encoding line I1). Subsequently, three more echo signals are simultaneously generated from s1, s2, and s3, respectively. and echo signal It can be without any phase element applied. Phase elements can be applied Phase elements can be applied Under the influence of readout gradient and phase encoding gradient, echo signals without phase modulation can be transformed. Having phase value of and having phase value of Fill the same positions in the first k-space K1 (e.g., phase encoding line I2). The first k-space can be sampled in the manner described above to obtain the first k-space dataset D1 during P1.

[0106] After completing the acquisition of the first k-space dataset D1, the acquisition of the first k-space dataset D2 begins. This involves acquiring the three echo signals generated simultaneously from s1, s2, and s3, respectively. and It can be without any phase element applied. Phase elements can be applied Phase elements can be applied Under the influence of the readout gradient and the phase encoding gradient, it is possible to achieve the desired result without phase modulation. Having phase value of and having phase value of The signal is filled into the same position in the second k-space K2 (e.g., phase encoding line I1). Subsequently, three more echo signals are simultaneously generated from s1, s2, and s3, respectively. and echo signal It can be without any phase element applied. Phase elements can be applied Phase elements can be applied Under the influence of the readout gradient and the phase encoding gradient, it is possible to achieve the desired result without phase modulation. Having phase value of It can be applied to phase and having phase value of Fill the same positions in the second k-space K2 (e.g., phase encoding line I2). The second k-space can be filled in the manner described above to obtain the first k-space dataset D2 during P2.

[0107] Alternatively, both the first k-space dataset D1 and the first k-space dataset D2 can be acquired simultaneously. For example, in cardiac cine imaging, such as... Figure 6B As shown, after s1, s2, and s3 are simultaneously excited, during the first cardiac phase corresponding to the first cardiac motion cycle, the three echo signals... and The echo signal can be generated simultaneously from s1, s2, and s3. It can be without any phase element applied. Phase elements can be applied Phase elements can be applied Under the influence of the readout gradient and the phase encoding gradient, it is possible to achieve the desired result without phase modulation. Having phase value of and having phase value of The signals are filled into the same positions in the first k-space K1 (e.g., phase encoding line I1). Subsequently, during the second cardiac phase corresponding to the first cardiac motion cycle, three additional echo signals are simultaneously generated from s1, s2, and s3, respectively. and echo signal It can be without any phase element applied. Phase elements can be applied Phase elements can be applied Under the influence of the readout gradient and the phase encoding gradient, it is possible to achieve the desired result without phase modulation. Having phase value of It can be applied to phase and having phase value of Fill the same position in the second k space K2 (e.g., phase encoding line I1).

[0108] During the first cardiac phase corresponding to the second cardiac cycle, three echo signals and Generated simultaneously from s1, s2 and s3 respectively. It can be without any phase element applied. Phase elements can be applied Phase elements can be applied Under the influence of the readout gradient and the phase encoding gradient, it is possible to achieve the desired result without phase modulation. Having phase value of and having phase value of The signal is filled into the same position in the first k-space K1 (e.g., phase encoding line I2). Subsequently, during the second cardiac phase corresponding to the second cardiac motion cycle, three additional echo signals are simultaneously generated from s1, s2, and s3, respectively. and echo signal It can be without any phase element applied. Phase elements can be applied Phase elements can be applied Under the influence of the readout gradient and the phase encoding gradient, it is possible to achieve the desired result without phase modulation. Having phase value of and having phase value of The first k-space K1 and the second k-space K2 are filled into the same position (e.g., phase encoding line I2). The first k-space K1 and the second k-space K2 can be sampled in parallel as described above to acquire the first k-space dataset D1 (corresponding to the first heart phase) during P1 and the first k-space dataset D2 (corresponding to the second heart phase) during P2.

[0109] In some embodiments, phase modulation can be performed by applying a first transmission phase to s2 that varies in both the spatial dimension (e.g., phase encoding direction) and the temporal dimension (e.g., P1 and P2). For example, the first transmission phase may include phase elements. and It can be different It can be different In some embodiments, phase modulation can be performed by applying a second transmission phase to s3 that varies in both the spatial dimension (e.g., phase encoding direction) and the temporal dimension (e.g., P1 and P2). For example, the second transmission phase may include phase elements. and It can be different It can be different In some embodiments, the first transmission phase may be different from the second transmission phase. For example, and The changes between them can be different and The changes between them.

[0110] In some embodiments, the acquisition of target k-space data can be achieved using a single-channel receiving coil or a multi-channel receiving coil.

[0111] In 640, processing device 140 (e.g., reconstruction module 530) can use a trained reconstruction model to generate at least two target dealiasing images based on target k-space data, where each image corresponds to a target slice and a target acquisition period. For example, the process 600 for simultaneous multi-slice imaging can be used to generate a physiological motion movie of a region of interest (e.g., a movie of cardiac motion or lung respiratory motion). In this case, a series of target dealiasing images (e.g., time frames) can be generated, where each image corresponds to a target slice and depicts a motion phase (e.g., a cardiac phase, such as diastole or systole, or a respiratory phase, such as the end of expiration or the end of inspiration). A target acquisition period can correspond to a time frame and a corresponding motion phase. As another example, the process 600 for simultaneous multi-slice imaging can be used for perfusion analysis reflecting changes in contrast agent concentration over time in at least two target slices. In this case, a series of dealiasing images (e.g., time frames) can be generated, where each image reflects the concentration of contrast agent in a target slice at a certain time point. A target acquisition period can correspond to a time point on a contrast agent time-concentration curve.

[0112] In some embodiments, simultaneous multilayer imaging (SMI) excites multiple layers simultaneously and acquires magnetic resonance (MR) signals generated from the multiple excited layers at the same time, compared to imaging that excites only a single layer at a time. The MR signals are then filled into k-space to generate k-space data. Since the received MR signals include contributions from multiple excited layers, directly reconstructing the k-space data using inverse Fourier transform may result in an aliased image of multiple layers. As used herein, the term "target dealiased image," also referred to as "target image," means an image corresponding to a single layer among multiple simultaneously excited layers that is dealiased or has relatively few aliasing artifacts relative to an aliased image of multiple layers.

[0113] In some embodiments, the trained reconstruction model may include a machine learning model or other forms of computer intelligence. In some embodiments, the trained reconstruction model may include a deep learning model. In some embodiments, the trained reconstruction model may include a neural network model. In some embodiments, the trained reconstruction model may include convolutional neural networks (CNNs), U-nets, V-nets, and recurrent neural networks (RNNs), or any combination thereof.

[0114] In some embodiments, target k-space data can be input into a trained reconstruction model, which can output at least two target de-aliasing images based on the target k-space data.

[0115] In some embodiments, the processing device 140 can generate at least two target aliasing images by performing an inverse Fourier transform on the target k-space data. Each target aliasing image corresponds to the at least two target slices and a target acquisition period. For example, the target k-space data may include at least two first k-space datasets, each corresponding to the at least two target slices and a target acquisition period. The processing device 140 can generate the at least two target aliasing images by performing an inverse Fourier transform on each of the at least two first k-space datasets. The processing device 140 can input the at least two aliasing images into a trained reconstruction model. The trained reconstruction model can output at least two target dealiasing images based on the at least two target aliasing images.

[0116] In some embodiments, the processing device 140 may determine at least two reference datasets based on target k-space data. Each reference dataset corresponds to a target slice and a target acquisition period. The at least two reference datasets may provide dealiasing information for dealiasing the target k-space data. For example, the target k-space data may be aliased data of the at least two target slices, and the dealiasing information may be used to dealias the aliased data of the at least two target slices.

[0117] In some embodiments, for a target acquisition period A, the processing device 140 can merge two or more first k-space datasets to generate an independent dataset. Each independent dataset corresponds to a target slice and a target acquisition period A. The two or more merged first k-space datasets each correspond to two or more adjacent target acquisition periods, one of which includes target acquisition period A. Subsequently, the processing device 140 can generate a reference dataset by applying phase modulation similar to that shown in operation 620, by multiplying the corresponding independent dataset with at least one transport phase to match the phase modulation of the target k-space data.

[0118] In some embodiments, the processing device 140 may use another method to generate a reference dataset. For example, the processing device 140 may generate an average dataset by averaging independent datasets corresponding to at least two target acquisition periods. The processing device 140 may then generate a reference dataset from the average dataset.

[0119] Detailed information on determining the reference dataset based on the target k-space data can be found in the reference “Ferrazzi G, et al., Autocalibrated multiband CAIPIRINHA with through-time encoding: Proofof principle and application to cardiac tissue phase mapping, Magn Reson Med. 2019 Feb; 81(2):1016-1030,” which is incorporated herein by reference.

[0120] In some embodiments, the processing device 140 can generate at least two target dealiasing images based on target k-space data and the at least two reference datasets using a trained reconstruction model. For example, the processing device 140 can generate at least two aliasing images by performing an inverse Fourier transform on the target k-space data, each aliasing image corresponding to at least two target slices and a target acquisition period. The processing device 140 can also generate at least two reference images with low image resolution by performing an inverse Fourier transform on the reference datasets. The processing device 140 can input the at least two aliasing images and the at least two reference images into the trained reconstruction model. The trained reconstruction model can output at least two dealiasing images based on the at least two aliasing images and the at least two reference images.

[0121] As another example, the processing device 140 can input target k-space data and the at least two reference datasets into a trained reconstruction model. The trained reconstruction model can output at least two dealiased images based on the target k-space data and the at least two reference datasets.

[0122] Figure 6D This is an exemplary schematic diagram of a trained reconstruction model shown according to some embodiments of this application.

[0123] like Figure 6DAs shown, the trained reconstruction model 650 may include an input layer 651, an output layer 652, and at least two hidden layers 653. The layers of the trained reconstruction model 650 may be connected in a feedforward manner, and the output of the i-th layer may be used as the input of the (i+1)-th layer. In some embodiments, in the trained reconstruction model 650, the input layer 651 may be configured to receive input data (e.g., target k-space data, at least two aliased images, at least two reference datasets, or at least two reference images) from the trained reconstruction model 650. Each hidden layer (e.g., 653-1, 653-2, ..., 653-N, etc.) may perform a specific function, such as convolution, pooling, normalization, matrix multiplication, nonlinear activation, etc. The input to the output layer 652 may come from the previous layer, and one or more transformations are performed on the received input to generate output data (e.g., at least two target dealiased images) from the trained reconstruction model 650. In some embodiments, the trained reconstruction model 650 may transform input data into output data by managing the coefficients of a weighting function. The weighting function can be represented as at least two hidden layers 653 between the input layer 651 and the output layer 652 of the trained reconstruction model 650. In some embodiments, each node of the at least two hidden layers 653 (e.g., Figure 6D An AN (in this context) can read input data, multiply it by various weights or coefficients, and produce output data.

[0124] For illustrative purposes, a convolutional neural network (CNN) model can be used as an example. Exemplary hidden layers may include convolutional layers, pooling layers, and fully connected layers. In some embodiments, k-space data (e.g., target k-space data and / or at least two reference datasets) or one or more images (e.g., at least two aliased images and / or at least two reference images) may be fed as input data into the trained reconstruction model. The k-space data or one or more images may be represented as a 2D or 3D matrix comprising at least two elements (e.g., pixels or voxels). Each of the at least two elements in the matrix may have a value representing a feature or feature of the element.

[0125] A convolutional layer may include one or more convolutional kernels, which can be used to extract features from the input data. In some embodiments, each of the one or more convolutional kernels may have a specific size and stride. In some embodiments, each of the one or more convolutional kernels may filter a portion of the input data to generate specific features corresponding to that portion. Specific features may be determined based on one or more convolutional kernels. Exemplary features may include low-level features (e.g., edge features, texture features, pixel value features), high-level features, or complex features.

[0126] Pooling layers can take the output of a convolutional layer as input. A pooling layer may include at least two pooling nodes, which can be used to sample the output of the convolutional layer to reduce the computational load of data processing and speed up data processing. In some embodiments, the size of the matrix representing the input data can be reduced in the pooling layer.

[0127] A fully connected layer may include at least two neurons. Neurons may be connected to pooling nodes in a pooling layer. In a fully connected layer, at least two vectors corresponding to at least two pooling nodes may be determined based on one or more features of the input data, and at least two weighting coefficients may be assigned to the at least two vectors.

[0128] The output layer can determine its output based on vectors and weighting coefficients obtained from the fully connected layer. In some embodiments, the output of the output layer may include at least two target dealiasing images.

[0129] In some embodiments, the trained reconstruction model can be implemented on one or more processing devices (e.g., processing device 140, processor 210, terminal 130, CPU 340, GPU 330, etc.). For example, one or more layers can be implemented separately on the processing devices. As another example, one or more components of a layer can be implemented on the same processing device. In some embodiments, at least two processing devices can be allocated, for example, for operations on different nodes (e.g., convolutional kernels, pooling nodes, neurons) in the trained reconstruction model, to enable parallel processing operations to be performed in some layers of the trained reconstruction model. For example, a first GPU can perform operations corresponding to convolutional kernel A and convolutional kernel B, and a second GPU can perform operations corresponding to convolutional kernel C and convolutional kernel D. Similarly, at least two GPUs can also perform operations on other nodes (e.g., convolutional kernels, pooling nodes, neurons) in the trained reconstruction model.

[0130] Furthermore, in some embodiments, a storage device (e.g., storage device 150, memory 320, storage device 490, memory 460, etc.) may be provided for storing data related to the trained reconstruction model, such as activation functions, learning weights for each node, and / or network topology (e.g., the number (or count) of hidden layers, the type of each hidden layer, etc.). Optionally, the storage device may further store the training dataset.

[0131] Meanwhile, traditional reconstruction methods for multi-layer imaging can be performed using least squares methods (e.g., Sensitivity Coding (SENSE)) in the image domain or linear methods (e.g., Generalized Automatic Calibration Partial Parallel Acquisition (GRAPPA)) in the k-space. In contrast, simultaneous multi-layer image reconstruction based on machine learning models is a non-linear method that can improve both image quality and reconstruction speed.

[0132] In some embodiments, the reconstruction method using machine learning models in this application can be applied to phase modulation (as shown in operation 620) and multilayer imaging to generate a series of dynamic images.

[0133] For example, the reconstruction method using machine learning models in this application can be applied to the reconstruction of a physiological motion (e.g., cardiac motion or respiratory motion) movie of a region of interest (e.g., at least a portion of the heart or lungs). In this case, each of the acquisition periods of at least two targets may correspond to a physiological motion phase (e.g., cardiac phase or respiratory phase) of the region of interest, and the at least two de-aliased images of the targets can form a physiological motion movie of the region of interest.

[0134] As another example, the reconstruction method using machine learning models in this application can be applied to perfusion imaging. In this case, the at least two target dealiasing images can be used for perfusion analysis, reflecting the change in contrast agent concentration over time in at least two target slices, with one target acquisition period corresponding to a time point in the time-concentration curves of the contrast agent in at least two target slices.

[0135] It should be noted that the above description is provided for illustrative purposes only and is not intended to limit the scope of this application. Various changes and modifications can be made by those skilled in the art based on the description herein. However, these changes and modifications do not depart from the scope of this application.

[0136] Figure 7 This is a flowchart illustrating an exemplary process for obtaining a trained reconstruction model according to some embodiments of this application. In some embodiments, process 700 may be performed in... Figure 1 This is implemented in the magnetic resonance imaging system 100 shown. For example, process 700 may be stored as instructions in a storage device (e.g., storage device 150 or memory 320 of processing device 140), and may be executed by processing device 140 (e.g., processor 310 of processing device 140), or... Figure 5 (One or more modules of the processing device 140 shown are invoked and / or executed). The operation of the illustrated process 700 presented below is intended to illustrate the purpose. In some embodiments, process 700 may be accomplished using one or more additional operations not described and / or one or more operations not discussed. Additionally, Figure 7 The order of operations in process 700 shown and described below is not intended to be restrictive.

[0137] In 710, the reconstruction module 530 can acquire at least two training datasets. Each of the at least two training datasets may include sample k-space data and at least two sample images.

[0138] In some embodiments, each of the at least two sample images may correspond to a sample sheet and a sample acquisition period. As used herein, the term "sample image," also known as a "sample dealiasing image," refers to an image corresponding to a single sheet of at least two sample sheets of a sample that is dealiased or has relatively few aliasing artifacts relative to an aliased image of at least two sample sheets.

[0139] In some embodiments, phase modulation can be applied to the sample k-space data corresponding to at least one sample slice. In some embodiments, for at least one sample slice, the phase of the corresponding sample k-space data can vary based on at least two sample acquisition periods. In some embodiments, for one sample slice, the phase of the corresponding sample k-space data can vary along the phase encoding direction. In some embodiments, the modulation phases of the sample k-space data corresponding to two or more slices can be different.

[0140] In some embodiments, the sample k-space data may include at least two second k-space datasets, each second k-space dataset corresponding to at least two sample slices and a sample acquisition period. The second k-space datasets may be aliased k-space datasets, comprising k-space data acquired from the at least two sample slices during a sample acquisition period.

[0141] In some embodiments, the sample k-space data may include true k-space data obtained by simultaneously exciting at least two sample patches. Phase modulation of the excited sample patches can be achieved by applying a sample phase that varies with the at least two sample acquisitions to the at least one excited sample patch. In some embodiments, the sample phase may further vary along the phase encoding direction. In some embodiments, the sample phases applied to different sample patches may be different. In some embodiments, sample k-space data may be acquired from at least two sample patches during at least two sample acquisitions. At least two sample images can be generated from the sample k-space data using any reconstruction algorithm (e.g., SENSE, GRAPPA, etc.). In some embodiments, the sample k-space data may be undersampled, oversampled, or fully sampled.

[0142] In some embodiments, the sample k-space data may include synthetic k-space data. In this case, the at least two sample images may be generated based on simultaneous multilayer imaging or single-layer excitation imaging. At least two sample k-space datasets can be obtained by performing a Fourier transform on the at least two sample images. Each of the at least two sample k-space datasets may correspond to a sample image. Phase modulation may be applied to the sample k-space datasets, wherein the phase-modulated sample k-space datasets correspond to at least one sample slice and at least two sample acquisition periods. For example, for at least one sample slice, a sample phase that varies based on at least two sample acquisition periods may be applied to the corresponding sample k-space dataset. In some embodiments, the sample phase may be further varied along the phase encoding direction. In some embodiments, the sample phases applied to different sample slices may be different. In some embodiments, after phase modulation, sample k-space datasets corresponding to the same sample acquisition period may be combined to obtain a second k-space dataset. Each second k-space dataset corresponds to at least two sample slices and one sample acquisition period, and these second k-space datasets may form sample k-space data.

[0143] In some embodiments, synthetic k-space data can be further obtained by applying an undersampling strategy to at least two second k-space datasets by replacing a portion of the data in those at least two sample k-space datasets with zeros. In some embodiments, synthetic k-space data can also be obtained by applying an undersampling strategy to at least two sample k-space datasets by replacing a portion of the data in those at least two sample k-space datasets with zeros. For example, for sample k-space datasets corresponding to the same sample collection period, the same undersampling strategy can be applied to the sample k-space datasets by replacing a portion of the data at the same k-space location with zeros.

[0144] In 720, the reconstruction module 530 can obtain a trained reconstruction model by training an initial model based on at least two training datasets.

[0145] In some embodiments, the reconstruction module 530 can obtain a trained reconstruction model by training an initial model using sample k-space data and sample images from at least two training datasets.

[0146] In some embodiments, for a training dataset, the reconstruction module 530 can generate at least two sample aliasing images by performing inverse Fourier transforms on at least two second k-space datasets of the sample k-space data in the training dataset. Each sample aliasing image corresponds to at least two sample slices and one sample acquisition period.

[0147] In some embodiments, the reconstruction module 530 can obtain a trained reconstruction model by training an initial model based on the at least two sample aliased images. For example, if all sample k-space data in the at least two training datasets are converted into sample aliased images, the reconstruction module 530 can obtain a trained reconstruction model by training an initial model based on the sample aliased images and sample images in the at least two training datasets. As another example, if a portion of the sample k-space data in the at least two training datasets is converted into sample aliased images, the reconstruction module 530 can obtain a trained reconstruction model by training an initial model using the sample aliased images, the remaining sample k-space data, and sample images in the at least two training datasets.

[0148] In some embodiments, an initial model can be obtained during the training of a trained reconstruction model. The initial model can be trained based on training input data (e.g., sample k-space data and / or sample aliasing images) and sample images (e.g., known outputs of the training input data (standard data)) from at least two training datasets to obtain the trained reconstruction model. In some embodiments, the initial model may include at least two weight parameters to be determined during training, which may be referred to as the training process. During training, the training input data may be processed by the initial model so that the initial model learns how to provide outputs for new input data by summarizing information learned from the training data during training. The purpose of learning may be to adjust the weight parameters to predict the correct output given an input.

[0149] A detailed description of the training process for the trained reconstruction model can be found elsewhere in this application (e.g., in conjunction with...). Figure 8 (Description).

[0150] In some embodiments, after collecting additional sample images, the training process can be repeated to update the trained reconstruction model with the additional sample images, which may or may not include previous sample images used in the first few rounds of training.

[0151] In some embodiments, the trained reconstruction model may be determined by the magnetic resonance imaging system 100 (e.g., processing device 140, terminal 130, storage device (storage device 150, memory 320, storage device 490)) or a third party (e.g., an external device). In some embodiments, the magnetic resonance imaging system 100 may determine and / or update the trained reconstruction model offline and store the trained reconstruction model in a storage device. In some embodiments, the trained reconstruction model may be determined and / or updated (or maintained) by, for example, the manufacturer or supplier of the scanning device 110. For example, the manufacturer or supplier may load one of the trained reconstruction models into the magnetic resonance imaging system 100 or a portion thereof (e.g., processing device 140 and / or terminal 130) before or during the installation of the scanning device 110, processing device 140, and / or terminal 130, and maintain or update the trained reconstruction model from time to time (regularly or irregularly). The maintenance or updating of the trained reconstruction model can be achieved through an installation program, which can be stored on a storage device (e.g., optical disc, USB drive, etc.) or obtained from an external resource (e.g., a server maintained by the manufacturer or supplier) via network 120. The program may include a new model (e.g., a new trained reconstruction model) or a part thereof to replace or supplement the trained reconstruction model.

[0152] It should be noted that the above description is provided for illustrative purposes only and is not intended to limit the scope of this application. Various changes and modifications can be made by those skilled in the art based on the description herein. However, these changes and modifications do not depart from the scope of this application.

[0153] Figure 8 This is an exemplary flowchart illustrating a training process for obtaining a trained reconstruction model according to some embodiments of this application. In some embodiments, it is possible to... Figure 1 The imaging system 100 shown implements process 800. For example, process 800 may be stored as instructions in a storage medium (e.g., memory 320 of storage device 150 or processing device 140, memory 490 of terminal 130, memory 460 of terminal 130), and may be invoked and / or executed by processing device 140 or terminal 130 (e.g., processor 310 of processing device 140, CPU 440 and / or GPU 430 of terminal 130, or...). Figure 5 (One or more modules of the processing device 140 shown). The operation of the process 800 presented below is intended to illustrate the process. In some embodiments, process 800 may be accomplished using one or more additional operations not described and / or without one or more operations discussed. Additionally, Figure 8The order of operations of process 800 shown and described below is not intended to be limiting. In some embodiments, operation 720 of process 700 may be performed based on process 800.

[0154] In some embodiments, the reconstruction module 530 can obtain a trained reconstruction model by performing an iterative process including one or more iterations. In some embodiments, the reconstruction module 530 can determine the updated values ​​of the weight parameters of the trained reconstruction model by performing an iterative process of a backpropagation neural network training process (e.g., stochastic gradient descent backpropagation training technique) to update the weight parameters of the initial model. For example, the reconstruction module 530 can adjust the parameters of the neural network layers by backpropagating the error determined for the output of the neural network.

[0155] In 810, the reconstruction module 530 can output at least two output images through an intermediate model based on sample k-space data from a training dataset. For example, the reconstruction module 530 can input the sample k-space data into the intermediate model. The intermediate model can output an output image based on the sample k-space data. As another example, the reconstruction module 530 can input a sample aliasing image corresponding to the sample k-space data into the intermediate model. The intermediate model can output an output image based on the sample aliasing image. In some embodiments, each of the at least two output images may correspond to a sample image.

[0156] In some embodiments, for the first iteration, the intermediate model may include the initial model, or the intermediate model may include an updated model generated in a previous iteration of the current iteration.

[0157] In step 820, reconstruction module 530 can determine the differences between the at least two output images and the at least two sample images. In some embodiments, reconstruction module 530 can determine the value of a loss function based on the differences.

[0158] In step 830, reconstruction module 530 may determine whether a termination condition is met. An exemplary termination condition could be that the value of the loss function in the current iteration is less than a threshold. Other exemplary termination conditions may include the maximum number of iterations (or counts) performed, and / or the difference between the loss function values ​​obtained in previous iterations and the current iteration (or the difference between loss function values ​​within a specific number of consecutive iterations) is less than a specific threshold. In response to the termination condition not being met, process 800 may proceed to 850 and initiate a new iteration by further repeating steps 810-830 until the termination condition is met. In response to the termination condition being met, process 800 may proceed to operation 840, whereby, for example, the iterative process may be terminated, and the intermediate model in the current iteration may be identified as the trained reconstruction model and may be stored and / or output.

[0159] In step 850, the reconstruction module 530 can update the intermediate model based on the differences between the at least two output images and the at least two sample images. For example, the reconstruction module 530 can update the weight parameters in the intermediate model based on the differences between the at least two output images and the at least two sample images.

[0160] It should be noted that the above description is provided for illustrative purposes only and is not intended to limit the scope of this application. Various changes and modifications can be made by those skilled in the art based on the description herein. However, these changes and modifications do not depart from the scope of this application.

[0161] Figure 9 This is a flowchart illustrating an exemplary model training phase and an exemplary model application phase according to some embodiments of this application.

[0162] Training data 910 (e.g., including at least two training datasets) can be obtained. Figure 7 (As shown in operation 710 of process 700). Figure 9 As shown, the training dataset 912 may include sample k-space data 914 and ground truth data 916 used as training input data. The sample k-space data 914 may be subjected to phase modulation 918 (e.g., ...). Figure 7 (As shown in operation 710 of process 700). Sample k-space data 914 may include at least two second k-space datasets (e.g., D′1, D′2, ...), each second k-space dataset being an aliased k-space dataset comprising k-space data acquired during a sample acquisition period (e.g., P′1, P′2, ...) from at least two sample slices (e.g., s′1+s′2+...). Ground truth data 916 may include at least two sample images (e.g., I′...). 1,1 、I′ 1,2 、I′ 2,1 、I′ 2,2 …), each sample image corresponds to a sample slice layer (e.g., s′1, s′2…) and a sample acquisition period (e.g., P′1, P′2…).

[0163] Training data 910 can be used in the training process 920 (e.g.) Figure 7 Operations 720 and 700 in process 700 Figure 8 The process shown in 800 is used to obtain a trained reconstruction model 930.

[0164] In the application phase of the trained reconstruction model 930, the target k-space data 940 (such as...) Figure 6B-6C As shown, and Figure 6AThe process described in step 600 (operation 630) can be input into the trained reconstruction model 930. For example... Figure 9 As shown, the target k-space data 940 can be subjected to phase modulation 942 (e.g., Figure 6B-6C As shown in 9, and Figure 6A (As described in operation 620 of process 600). The target k-space data 940 may include at least two first k-space datasets (e.g., D1, D2…), wherein each first k-space dataset is an aliased k-space dataset, including k-space data from at least two target slices (e.g., s1+s2+…) acquired during a target acquisition period (e.g., P1, P2…). Based on the target k-space data 940, the trained reconstruction model 930 can output at least two target dealiased images 950 (e.g., I…). 1,1 I 1,2 I 2,1 I 2,2 …), where each image corresponds to a target slice (e.g., s1, s2…) and a target acquisition period (e.g., P1, P2…).

[0165] It should be noted that the above description is provided for illustrative purposes only and is not intended to limit the scope of this application. Various changes and modifications can be made by those skilled in the art based on the description herein. However, these changes and modifications do not depart from the scope of this application.

[0166] Figure 10 This is a flowchart illustrating an exemplary process for MR imaging according to some embodiments of this application. In some embodiments, process 1000 can be performed in... Figure 1 This is implemented in the magnetic resonance imaging system 100 shown. For example, process 1000 can be stored as instructions in a storage device (e.g., storage device 150 or memory 320 of processing device 140), and can be processed by processing device 140 (e.g., processor 310 of processing device 140, or...). Figure 5 The process 1000 shown may be invoked and / or executed by one or more modules of the processing device 140. The operation of the process 1000 presented below is intended to be illustrative. In some embodiments, the process 1000 may be accomplished by one or more additional operations not described, and / or without the one or more operations discussed. Additionally, Figure 10 The order of operations of process 1000 shown and described below is not intended to be restrictive.

[0167] In 1010, the processing device 140 (e.g., acquisition module 520) can acquire target k-space data related to the region of interest of the object.

[0168] In some embodiments, target k-space data can be acquired by scanning the region of interest of the object using a magnetic resonance imaging device (e.g., scanning device 110). In some embodiments, target k-space data can be acquired by single-layer excitation or multi-layer excitation (e.g., simultaneous multilayer imaging). In some embodiments, the target k-space data can be undersampled or fully sampled.

[0169] In some embodiments, the target k-space data can be automatically calibrated acquisition data including the following characteristics: The target k-space data can be acquired from at least two slices in which the region of interest of the object is simultaneously excited. The target k-space data can be acquired during at least two acquisition periods to reconstruct a series of time frames from the target k-space data, where each time frame corresponds to a slice and an acquisition period. The target k-space data can be phase-modulated by applying a transport phase that varies along the phase encoding direction and / or the time dimension (e.g., at least two acquisition periods). In some embodiments, this can be achieved by performing... Figure 6A The process 600 uses operations 610 and 620 to obtain the target k-space data.

[0170] In 1020, processing device 140 (e.g., reconstruction module 530) can use a trained reconstruction model to generate one or more target images based on target k-space data. The trained reconstruction model can be a machine learning model. In some embodiments, processing device 140 can input target k-space data into the trained reconstruction model. The trained reconstruction model can output one or more target images based on the target k-space data.

[0171] In some embodiments, the processing device 140 may use a trained reconstruction model to generate a target image corresponding to a target slice of the region of interest, based on target k-space data. In some embodiments, the processing device 140 may use a trained reconstruction model to generate at least two target images (time frames) based on target k-space data, the at least two target images (time frames) corresponding to the same target slice of the region of interest and different target acquisition periods. In some embodiments, the processing device 140 may use a trained reconstruction model to generate at least two target images based on target k-space data, each target image corresponding to a different target slice of the region of interest. In some embodiments, the processing device 140 may use a trained reconstruction model to generate at least two target images (time frames) based on target k-space data, each target image (time frame) corresponding to a target slice of the region of interest and a target acquisition period.

[0172] By way of example only, the target k-space data can be the automatically calibrated acquisition data described above. The processing device 140 can use a trained reconstruction model to generate at least two target images (time frames) based on the target k-space data, each target image (time frame) corresponding to a target slice of the region of interest and a target acquisition period. A detailed description of the imaging processing in this embodiment can be found in conjunction with this application. Figure 6A- Figure 9 The description.

[0173] One aspect of this application relates to systems and methods for simultaneous multi-layer imaging, and more particularly to systems and methods for reconstructing autocalibrated acquisition data using machine learning models. The autocalibrated acquisition data may include the following characteristics: The autocalibrated acquisition data can be acquired from at least two slices of a region of interest (ROI) of simultaneously excited objects. The autocalibrated acquisition data can be acquired during at least two acquisition periods to reconstruct a series of dealiased images, wherein each dealiased image corresponds to one of the at least two slices and one of the at least two acquisition periods. The autocalibrated acquisition data can be phase-modulated by applying a transport phase. The transport phase varies along the phase encoding direction and / or the time dimension (e.g., during the at least two acquisition periods). Reconstructing the autocalibrated acquisition data using machine learning models is a non-linear method that can improve image quality and reconstruction speed.

[0174] At least two reference datasets can be determined based on automatically calibrated acquisition data (k-space data), which provide dealiasing information, indicating that the automatically calibrated acquisition data itself includes dealiasing information. Therefore, after the automatically calibrated acquisition data is input into a machine learning model, the model can automatically extract the implicit dealiasing information from the data and output a dealiased image. This eliminates the need for additional operations to acquire reference datasets in simultaneous multi-layer imaging reconstruction (e.g., additional computations or scanning operations required to acquire reference datasets), thus improving the efficiency of simultaneous multi-layer imaging reconstruction.

[0175] It should be noted that the above description is provided for illustrative purposes only and is not intended to limit the scope of this application. Various changes and modifications can be made by those skilled in the art based on the description herein. However, these changes and modifications do not depart from the scope of this application.

[0176] The basic concepts have been described above. Obviously, for those skilled in the art who have read this application, the above disclosure is merely illustrative and does not constitute a limitation of this application. Although not explicitly stated herein, those skilled in the art may make various modifications, improvements, and corrections to this application. Such modifications, improvements, and corrections are suggested in this application, and therefore, such modifications, improvements, and corrections still fall within the spirit and scope of the exemplary embodiments of this application.

Claims

1. A method for simultaneous multilayer imaging (SMS), implemented on a computer including one or more processors and one or more storage devices, the method comprising: Obtain target k-space data related to the region of interest (ROI) of the object; as well as Based on the target k-space data, at least two target images are generated using a trained reconstruction model. Each target image corresponds to one of at least two target slices of the region of interest and one of at least two target acquisition periods. The trained reconstruction model is obtained by training an initial model based on at least two training datasets. Each of the at least two training datasets includes sample k-space data and at least two sample dealiasing images. Each of the at least two sample dealiasing images corresponds to one of at least two sample slices of a sample and one of at least two sample acquisition periods. For at least one sample slice, the phase of the corresponding sample k-space data varies based on the at least two sample acquisition periods and / or the phase encoding direction. The sample k-space data includes synthetic k-space data, which is obtained through the following steps: By performing a Fourier transform on the at least two sample dealiased images, at least two sample k-space datasets are obtained, each of the at least two sample k-space datasets corresponding to one of the at least two sample dealiased images; Phase modulation is applied to the at least two sample k-space datasets such that, for the at least one sample slice, the sample phase of the corresponding sample k-space dataset varies based on the at least two sample acquisition periods and / or the phase encoding direction; and By combining the sample k-space datasets corresponding to the same sample collection period, at least two second k-space datasets are obtained, each of the at least two second k-space datasets corresponding to one of the at least two sample slices and the at least two sample collection periods, and the sample k-space datasets include the at least two second k-space datasets.

2. The method according to claim 1, characterized in that, The target k-space data includes at least two first k-space datasets, each of which is acquired in one of the at least two target slices and during one of the at least two target acquisition periods.

3. The method according to claim 1, characterized in that, The step of generating the at least two target images using a trained reconstruction model based on the target k-space data includes: The target k-space data is input into the trained reconstruction model; and Based on the target k-space data, the trained reconstruction model outputs the at least two target images.

4. The method according to claim 1, characterized in that, The acquisition of target k-space data related to the region of interest of the object includes: The magnetic resonance imaging device applies one or more multi-band excitation radio frequency (RF) pulses to the region of interest to simultaneously excite at least two target slices in the region of interest once or more. By applying a transfer phase to at least one excited target slice using the magnetic resonance imaging device, the magnetic resonance imaging device performs phase modulation on the at least one excited target slice, the transfer phase varying with the acquisition period of the at least two targets and / or the phase encoding direction; and During the at least two target acquisitions, target k-space data are acquired through at least two excited target sheets.

5. The method according to claim 4, characterized in that, The step of generating at least two target images based on the target k-space data and using a trained reconstruction model includes: At least two target aliasing images are generated by performing an inverse Fourier transform on the target k-space data, each of the at least two target aliasing images corresponding to one of the at least two target acquisition periods and the at least two target slices; The at least two aliased images are input into the trained reconstruction model; and Based on the at least two target aliased images, the at least two target images are output through the trained reconstruction model.

6. The method according to claim 4, characterized in that, The step of generating at least two target images based on the target k-space data and using a trained reconstruction model includes: Based on the target k-space data, at least two reference datasets are determined, each of the at least two reference datasets corresponding to one of the at least two target slices and one of the at least two target acquisition periods. The at least two reference datasets provide dealiasing information for dealiasing the target k-space data. Based on the target k-space data and the at least two reference datasets, the at least two target images are generated using the trained reconstruction model.

7. A non-transitory computer-readable medium comprising at least one set of instructions, wherein when executed by one or more processors of a computing device, the at least one set of instructions causes the computing device to perform the method of any one of claims 1-6.

8. A simultaneous multilayer imaging (SMS) system, comprising: Magnetic resonance imaging (MRI) equipment is configured to scan the region of interest (ROI) of an object; At least one storage device, including a set of instructions; as well as At least one processor for communicating with the at least one storage device, wherein, when executing the set of instructions, the at least one processor is used to instruct the system to perform the method of any one of claims 1-6.

Images