Systems and methods for performing quantitative susceptibility mapping

Abstract

A method of performing MR imaging includes receiving readout data in response to a pulse sequence being applied as a region of interest moves through a plurality of motion states. The pulse sequence includes first and second mGRE sequences with first and second pluralities of interleaved echo times. The method further includes generating phase maps each corresponding to one of the echo times and one of the motion states. The method further includes combining, for each respective motion state, the phase maps for the respective motion state to generate a final phase map across all of the first and second plurality of echo times for the respective motion state. The method further includes standardizing an orientation of each final phase maps across all of the motion states. The method further includes generating a quantitative susceptibility map based on the orientation-standardized final phase map for each of the motion states.

Description

SYSTEMS AND METHODS FOR PERFORMING QUANTITATIVE SUSCEPTIBILITY MAPPINGCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of, and priority to, U.S. Provisional Patent Application No. 63 / 641,801 filed on May 2, 2024, which is hereby incorporated by reference herein in its entirety.TECHNICAL FIELD

[0002] The present disclosure relates generally to systems and methods for performing magnetic resonance imaging, and more particularly, to systems and methods for performing quantitative susceptibility mapping using mGRE pulse sequences with interleaved echo times and multi-orientation QSM utilizing movement during the respiratory cycle.BACKGROUND

[0003] Major technical advances are required for accurate quantification of intramyocardial iron content. Since quantitative susceptibility mapping (QSM) derivation depends on the phase information in MRI, it is extra sensitive to phase errors caused by motion and Bo off-resonance. The effect is extra prominent in the heart, where continuous cardiac and respiratory movement and strong Bo off-resonance at the heart-lung interfaces are presented. Free-breathing strategies for cardiac QSM often come with a compromised imaging efficiency and are susceptible to residual motion, and do not have sufficient resolution for small but important lesions. In addition to motion, mGRE sequences can, particularly when motion compensation gradients are applied, have a restricted echo spacing (ATE). This can limit the dynamic range of the QSM sequence and impair the phase map fidelity at regions with rapidly changing Bo fields (e.g., the heart-lung interfaces and around the focal iron lesion). Furthermore, high susceptibility makes the single-orientation QSM reconstruction prone to streaking artifacts from the ill-posted dipole inversion calculation, which can lead to significant bias in iron quantification. These confounders significantly impaired the applicability of the current QSM techniques in AMI patients. Thus, new systems and methods to facilitate high phase map fidelity and multi-orientation well-posed QSM techniques in the heart are needed.SUMMARY

[0004] A method of magnetic resonance (MR) imaging comprises receiving readout data in response to a pulse sequence being applied to a region of interest of a subject as the region of interest moves through a plurality of motion states, the pulse sequence including a first multigradient echo (mGRE) sequence with a first plurality of echo times and a second mGRE sequence with a second plurality of echo times, the first plurality of echo times being interleaved with the second plurality of echo times. The method further comprises generating, based at least in part on the readout data, a plurality of phase maps of the region of interest each corresponding to (i) one of the echo times of the first plurality of echo times or the second plurality of echo times and (ii) one of the plurality of motion states. The method further comprises combining, for each respective motion state of the plurality of motion states, the plurality of phase maps for the respective motion state to generate a final phase map of the region of interest across all of the first and second plurality of echo times for the respective motion state. The method further comprises standardizing an orientation of each of the plurality of final phase maps across all of the plurality of motion states. The method further comprises generating a quantitative susceptibility map (QSM) of the region of interest based on the orientation-standardized final phase map for each of the plurality of motion states.

[0005] A method of magnetic resonance (MR) imaging comprises applying a pulse sequence to a region of interest of a subject, the pulse sequence including a first multi -gradient echo (mGRE) sequence with a first plurality of echo times, a second mGRE sequence with a second plurality of echo times, and a central k-space navigator signal following both the first mGRE sequence and the second mGRE sequence, the first plurality of echo times being interleaved with the second plurality of echo times. The method further comprises collecting readout data in response to the pulse sequence being applied as the region of interest moves through a plurality of motion states. The method further comprises generating, based at least in part on the readout data, a plurality of phase maps of the region of interest and a plurality of T2*-weighted images of the region of interest, each of the plurality of phase maps and each of the plurality of T2* -weighted images corresponding to (i) one of the echo times of the first plurality of echo times or the second plurality of echo times and (ii) one of the plurality of motion states. The method further comprises combining, for each respective motion state of the plurality of motion states, the plurality of phase maps for the respective motion state and the plurality of T2*-weighted images for the respective motion state to generate a final phase map of the region of interest across all of the first and second plurality of echo times for the respective motion state. The method further comprises standardizing an orientation of theplurality of final phase maps across all of the plurality of motion states. The method further comprises generating a quantitative susceptibility map (QSM) of the region of interest based on the orientation-standardized phase map for each of the plurality of motion states.

[0006] The above summary is not intended to represent each implementation or every aspect of the present disclosure. Additional features and benefits of the present disclosure are apparent from the detailed description and figures set forth below.BRIEF DESCRIPTION OF THE DRAWINGS

[0007] The disclosure, and its advantages and drawings, will be better understood from the following description of representative embodiments together with reference to the accompanying drawings. These drawings depict only representative embodiments and are therefore not to be considered as limitations on the scope of the various embodiments or claims.

[0008] FIG. 1 is a block diagram of a system for performing quantitative susceptibility mapping, according to aspects of the present disclosure.

[0009] FIG. 2 is a flowchart of a method for performing quantitative susceptibility mapping, according to aspects of the present disclosure.

[0010] FIG. 3 A shows images of a heart moving through different motion states as a user breathes, according to aspects of the present disclosure.

[0011] FIG. 3B shows the maximum difference in the motion state of the heart as the user breathes, according to aspects of the present disclosure.

[0012] FIG. 4A is a conventional T2*-weighted image of a region of interest obtained using breath-holding, according to aspects of the present disclosure.

[0013] FIG. 4B is a T2*-weighted image of a region of interest obtained using a free- breathing retrospectively binned 3D mGRE sequence, according to aspects of the present disclosure.

[0014] FIG. 4C is an ex-vivo ground-truth T2*-weighed image of a region of interest, according to aspects of the present disclosure.

[0015] FIG. 5A shows a continuous interleaved mGRE sequence with a temporal navigator, according to aspects of the present disclosure.

[0016] FIG. 5B a golden angle stack of stars gaussian random trajectory for a MR pulse sequence, according to aspects of the present disclosure.

[0017] FIG. 5C shows the RF pulses and magnetic gradients applied during two mGRE sequences and two temporal navigators, according to aspects of the present disclosure.

[0018] FIG. 5D shows the ATE in the two mGRE sequences, according to aspects of thepresent disclosure.

[0019] FIG. 6A shows a first step in an image data reconstruction pipeline , according to aspects of the present disclosure.

[0020] FIG. 6B shows a second step in the image data reconstruction pipeline, according to aspects of the present disclosure.

[0021] FIG. 6C shows a third step in the image data reconstruction pipeline, according to aspects of the present disclosure.

[0022] FIG. 7A shows a quantitative susceptibility map (QSM) of a region of interest obtained using conventional means, according to aspects of the present disclosure.

[0023] FIG. 7B shows an R2* image of the region of interest, according to aspects of the present disclosure.

[0024] FIG. 7C shows a QSM of the region of interest obtained using a free-running HDR- QSM technique, according to aspects of the present disclosure.

[0025] FIG. 8A shows streaking artifacts in a single-orientation QSM obtained using a MEDI algorithm, according to aspects of the present disclosure.

[0026] FIG. 8B shows streaking artifacts in a multi-orientation QSM obtained using a COSMOS algorithm, according to aspects of the present disclosure.

[0027] FIG. 9 is a flowchart for a MOR-HDR-QSM algorithm, according to aspects of the present disclosure.

[0028] While the present disclosure is susceptible to various modifications and alternative forms, specific implementations and embodiments have been shown by way of example in the drawings and will be described in detail herein. It should be understood, however, that the present disclosure is not intended to be limited to the particular forms disclosed. Rather, the present disclosure is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the present disclosure as defined by the appended claims.DETAILED DESCRIPTION

[0029] Magnetic resonance-based imaging (MR imaging) is a technique most often used for imaging the human body that takes into account principles of nuclear magnetic resonance. For example, doctors and other medical professionals often use MR imaging to view tissue within the human body. Nuclear magnetic resonance is a phenomenon in which nuclei (such as protons in body tissue) localized in a magnetic field emit energy that can be detected. This energy that is detected can be used to create an image. MR imaging generally involves two principal steps. First, the magnetic moment of the nuclei (a vector property of a nucleus causedby the intrinsic spin property of elementary particles) are aligned (or polarized) by the presence of an external magnetic field. While in the presence of this external magnetic field, the magnetic moment of each nuclei will generally precess about an axis parallel to the magnetic field. The rate of this precession co is generally proportional to yB0, where Bois the magnitude of the external magnetic field, and y is the gyromagnetic ratio of the nuclei, which is the ratio the nuclei’s magnetic moment to its angular momentum. The rate of the precession a> is considered the nuclei’s resonant frequency.

[0030] The second principal step in MR imaging is to apply an electromagnetic pulse sequence (usually a radiofrequency (RF) pulse) to the nuclei. When the frequency of the RF pulses sequence is generally equal to the resonant frequency of the nuclei, the nuclei absorb the energy of the RF pulse and the magnetic moments are rotated out of alignment with the magnetic field. The magnetic moments of the excited nuclei eventually re-align within the presence of the external magnetic field in a process known as relaxation, which has two components, T1 and T2. T1 relaxation describes how the component of the magnetic moment parallel to the external magnetic field returns to its initial value. T2 relaxation describes how the components of the magnetic moment perpendicular to the external magnetic field return to their initial value. Because the magnetic moments of nuclei in the external magnetic field without the RF pulse sequence applied are generally parallel to the external magnetic field, T1 relaxation generally describes how parallel component of the magnetic moment returns to its maximum value, while T2 relaxation generally describes how the perpendicular components of the magnetic moment decay. The nuclei of different material relax at different rates and thus emit differing signals, which can be detected and used to form an image identifying the different materials.

[0031] Dynamic MR imaging can produce a spatiotemporal image sequence I(x, t), which is a function of (i) spatial location within the subject and (ii) one or more time-varying parameters related to the dynamic processes. The spatial location is denoted by vector x = [X1, X2 / 3]T, which contains up to three spatially-varying parameters xt. The time-varying parameters are denoted by vector t = [t1(t2> — , tR]Tcontaining R time-varying independent variables tt. The imaging data obtained from the MR imaging is generally from a specific region of interest of the subject. In an example, the region of interest could be the subject’s abdomen or chest. In other examples, the region of interest of the subject is more specific. For example, the region of interest could be an organ, such as the subject’s liver, lungs, heart, pancreas, brain, prostate, breast, or any other organ.

[0032] The imaging data is dependent on or related to the spatially-varying and timevarying parameters of the region of interest of the subject referred to above. The spatially- varying parameters include a voxel location, a contrast agent kinetic parameter, or a diffusion parameter (which includes changing strength, changing direction, or both). The spatially- varying parameters can additionally or alternatively be related to physical motion of the region of interest of the subject. The time-varying parameters can include, but is not limited to: the phase of the subject’s heart within a cardiac cycle; the position of the subject’s lungs, chest wall, or other organs within a respiratory cycle; a position of a dome of a liver during respiration; relaxation parameters such as Tl, Tip (also known as Tl-rho), T2, T2* (also known as T2-star), an inversion time (or other time since magnetization preparation); a diffusion weighting strength; a diffusion weighting direction; an echo time; a dynamic contrast enhancement phase; a flip angle; an elapsed time since the start of scanning; elastographic wave propagation, a phase offset of elastographic excitation waves; a frequency offset and duration of saturation preparation pulses (e.g., for chemical exchange saturation transfer); a duration of magnetization transfer preparation pulses; a chemical exchange saturation transfer, a spectral position (e.g., for spectroscopy); a flow encoding strength; a flow encoding direction; free induction decay, or the general passage of time.

[0033] Some of the spatially-varying parameters can also be time-varying, and some of the time-varying parameters can also be spatially-varying. For example, cardiac motion is generally a time-varying parameter, while the relaxation parameters, the contrast agent kinetic parameter, and the diffusion parameter are generally time-varying. Generally, the imaging data is indicative of the value or magnitude of the spatially-varying parameters and / or the timevarying parameters. In another example, the region of interest is the subject’s abdomen containing their liver, and the spatially-varying parameter that is being measured is the Tl relaxation parameter. The Tl relaxation parameter can be spatially-varying, meaning that the value of the Tl relaxation parameter at a first physical location within the subject’s liver can be different than the value of the Tl relaxation parameter at a second physical location within the subject’s liver. In a resulting image showing the value measured Tl relaxation parameter, different locations in the image (corresponding to different physical locations within the subject’s liver) will show different values. In some implementations, the spatially-varying parameters can also be time-varying. In other implementations, the spatially-varying parameters can additionally or alternatively be related to physical motion of the region of interest of the subject. In general, the techniques disclosed herein can be used to perform dynamic imaging that resolves parameters that can vary across space and time.

[0034] The image sequence I(x, t) can be represented as a matrix A:

[0035] Matrix A can be decomposed as:

[0036] In this formulation, / is the total number of voxels in the image, and Ntis the number of time points. UxE CJxLcontains L spatial basis functions with J total voxels and is known as a spatial factor. Uxcontains one or more spatial basis functions that describe the properties of the spatially-varying parameters. Uxis formed as a matrix or tensor, and may be referred to as a spatial factor matrix or spatial factor tensor. E CLxNtcontains temporal weighting functions, and is known as the temporal factor. The weighting functions describe properties of time-varying parameters, such as relaxation, motion, contrast changes, etc. is formed as a matrix where imaging with only a single time dimension is used, and is formed as a tensor where imaging with multiple time dimensions is used. In general however, the term “spatial basis tensor” can be used to refer to the spatial factor, regardless of whether there is one spatially-varying parameter resulting in a matrix or multiple spatially-varying parameters resulting in a tensor. Similarly, the term “temporal basis tensor” can be used to refer to the temporal factor, regardless of whether there is one temporally-varying parameter resulting in a matrix or multiple temporally-varying parameters resulting in a tensor.

[0037] FIG. 1 illustrates a system 100 for performing magnetic resonance imaging on a subject according to aspects of the present disclosure. The system 100 (which may be an MRI system) includes an imaging apparatus 102 (which may be an MRI machine), a processing device 112, and a reconstruction workstation 122. The imaging apparatus 102 can be one used for standard magnetic resonance imaging, and can include a primary magnet 104, gradient coils 106, an RF transmission system 108, and an RF receiving system 110. The primary magnet 104 can be a permanent magnet, an electromagnet (such as a coil), or any other suitable magnet. Primary magnet 104 is used to create the external magnet field that is applied to the sample during imaging. Gradient coils 106 create a secondary magnet field that distorts the external magnetic field and can cause the resonant frequency of the protons in the sample to vary by position. The gradient coils 106 can thus be used to spatially encode the positions of protonsthroughout the sample, e.g. can be used to select which plane intersecting the sample will be used for imaging.

[0038] The RF transmission system 108 is used to apply the RF pulse sequence that provides energy to the protons in the sample to rotate their magnet moments out of alignment with the external magnetic field, and saturates the solute material protons. The RF transmission system 108 generally includes a frequency generator (such as an RF synthesizer), a power amplifier, and a transmitting coil. The RF receiving system 110 receives the signals emitted by the protons in the sample as they relax back to their standard alignment. The RF receiving system 110 can a receiving coil to receive the emitted signals, and a pre-amplifier for boosting the received signals and ensuring the signals are suitable for processing. In some implementations, the RF receiving system 110 can include a signal processing component that processes the received signals to provide data that is usable by the processing device 112. Each of the component of the imaging apparatus can be disposed within one or more housings. In some implementations, the imaging apparatus 102 is a 3.0 Tesla clinical scanner equipped with an 18-channel phase array body coil.

[0039] The processing device 112 can be communicatively coupled to the imaging apparatus 102, and can include a processor 114, processor-executable memory 116, a display 118, and a user input device 120. The processing device 112 is used to manage the operations of the imaging apparatus 102, and can thus be configured to cause the imaging apparatus 102 to perform dynamic imaging according to the principles disclosed herein. The memory 116 can contain instructions that when executed by processor 114, cause the imaging apparatus 102 to operate as desired. The memory 116 can also store the data obtained from the MRI sequence.

[0040] In some implementations, the imaging apparatus 102 and the processing device 112 are embedded within / formed as part of the same housing or group of housings. Thus, the display 118 may be a display mounted on the housing of the imaging apparatus 102, and the user input device 120 may be a keypad or a touchscreen display (which may also be the display 118) mounted on the housing of the imaging apparatus 102. In other implementations, the processing device 112 is separate from the housing of the imaging apparatus 102.

[0041] The reconstruction workstation 122 is generally a separate processing device or system that receives imaging data from the processing device 112. The reconstruction workstation 122 can be configured as necessary to perform any analysis of the data, include any or all of the steps of any methods or processes disclosed herein. In some implementations, the processing device 112 and the reconstruction workstation 122 are combined into a computing device that controls the imaging apparatus 102 and analyzes the data. In otherimplementations, the processing device 112 and the reconstruction workstation 122 are separate computing devices, wherein the processing device 112 controls the operation of the imaging apparatus 102 and the reconstruction workstation 122 is used to analyze the data.

[0042] FIG. 2 illustrates a flowchart of a method 200 for performing MR imaging on a subject. Method 200 can be implemented, for example, using the system 100 and / or using portions of the system 100 (such as any combination of the imaging apparatus 102, the processing device 112, and the reconstruction workstation 122).

[0043] Step 210 of method 200 includes applying a pulse sequence to a region of interest of a subject, where the pulse sequence includes a first multi -gradient echo (mGRE) sequence and a second mGRE sequence. In some implementations, the pulse sequence is applied before injection of a contrast agent (such as ferumoxytol). In some implementations, the pulse sequence is applied after injection of a contrast agent. In some implementations, the contrast agent is injected during the application of the pulse sequence (which could include the pulse sequence being injected simultaneously with the initiation and / or termination of the pulse sequence).

[0044] A GRE sequence includes the application of an RF pulse, followed by the application of a dephasing and a rephasing gradient to produce an echo that can be detected. An mGRE sequence is similar to a GRE sequence, except that the RF pulse is followed by the application of multiple sets of dephasing and rephasing gradients to produce multiple echoes, each occurring at a different time following the RF pulse. The times at which each of these echoes occurred (which can be manipulated by applying the dephasing and rephasing gradients at specific times) are referred to as “echo times.” Thus, a single mGRE sequence will include a plurality of echo times, which are the times at which each specific echo of the single mGRE sequence occurred.

[0045] The first mGRE sequence will thus have a first plurality of echo times, and the second mGRE sequence will have a second plurality of echo times. Each echo time within the first plurality of echo times will necessarily be different than the rest of the echo times within the first plurality of echo times, and similarly, each echo time within the second plurality of echo times will necessarily be different than the rest of the echo times within the second plurality of echo times. Other than this however, the echo times can vary as needed for different applications.

[0046] In some implementations, the first plurality of echo times is different than the second plurality of echo times. In some cases, this difference is due to the first plurality of echo times having the same number of echo times as the second plurality of echo times, but at leastone of the echo times in the first plurality of echo times is different than at least one of the echo times in the second plurality of echo times. For example, each individual echo time across the first and second plurality of echo times can be a distinct echo time that is different than all the other echo times. In other cases, this difference is due to the first plurality of echo times having a different number of echo times than the second plurality of echo times.

[0047] In some implementations, one or more of the echo times of the first plurality of echo times is identical to one of the echo times of the second plurality of echo times. In general, the first and second plurality of echo times could share any number of identical echo times, such as 0 identical echo times, 1 identical echo time, 2 identical echo times, etc.

[0048] In some implementations, each echo time in the first and second plurality of echo times is between about 1 millisecond (ms) and about 5 ms. In some implementations, the first plurality of echo times includes 5 echo times of about 1.5 ms, about 4.5 ms, about 7.5 ms, about 10.5 ms, and about 13.5 ms. In some implementations, the second plurality of echo times includes 5 echo times of about 3 ms, about 6 ms, about 9 ms, about 12 ms, and about 15 ms.

[0049] In some implementations, the first and second plurality of echo times are interleaved with each other. For example, each echo time may be interleaved, such that the first echo time of the second plurality of echo times is between the first and second echo times of the first plurality of echo times, the second echo time of the first plurality of echo times is between the first and second echo times of the second plurality of echo times, the second echo time of the second plurality of echo times is between the second and third echo times of the first plurality of echo times, and so on and so forth. In another example, multiple echo times of the first or second plurality of echo times may be sandwiched between two echo times of the other plurality of echo times.

[0050] In general, interleaving the echo times of the different mGRE sequences allows for a smaller difference between successive echoes ( TE). In most MRI systems, there are hardware limitations that set a minimum ATE of two echoes of a single mGRE sequence. By using two different mGRE sequences with interleaved echo times, the ATE between one echo time of the first mGRE sequence and the next echo time of the second mGRE sequence, an effective ATE that is smaller than the hardware-based minimum ATE is achieved.

[0051] Thus, with reference to the above example, a first mGRE sequence with echo times of 1.5 ms, 4.5 ms, 7.5 ms, 10.5 ms, and 13.5 ms and a second mGRe sequence with echo times of 3 ms, 6 ms, 9 ms, 12 ms, and 15 ms is generally equivalent to a single mGRE sequence with echo times of 1.5 ms, 3 ms, 4.5 ms, 6 ms, 7.5 ms, 9 ms, 10.5 ms, 12 ms, 13.5 ms, and 15 ms.

[0052] In some implementations, in addition to the dephasing and rephasing gradients applied during the mGRE sequence, one or more of the first and second mGRE sequences may include one or more motion compensation gradients and / or one or more flow compensation gradients. In some implementations, the echo times are selected prior to applying the pulse sequence. The echo times may be selected based on a number of different factors, including constraints based on the hardware used to perform the imaging. By utilizing two different mGRE sequences with different pluralities of echo times, the echo times can be adjusted as needed based on hardware (or other) constraints while still being able to obtain the data needed to perform the imaging.

[0053] In some implementations, the pulse sequence further includes two central-space navigator signals that follow each mGRE sequence. The two central k-space navigator signals are applied in the center of k-space and are used to track motion and other physiological changes that can affect image quality. By applying the navigator signal after each individual mGRE sequence, motion can be identified and corrected for.

[0054] In general, the pulse sequence that is applied at step 210 includes a plurality of subsequences that are repeatedly applied. Each sub-sequence includes the first mGRE sequence, the central k-space navigator signal following the first mGRE sequence, the second mGRE sequence, and the central k-space navigator signal following the second mGRE sequence. In some implementations, the sub-sequence can be repeatedly applied using a golden-angle radial trajectory along the x-y plane in k-space, and a randomized Gaussian distribution along the z- axis of k-space.

[0055] Step 220 of method 200 includes collecting readout data in response to the pulse sequence being applied. The readout data can be collected using, for example, the RF receiving system 110. The pulse sequence is applied and the readout data collected as the region of interest of the subject moves through a plurality of motion states. In some implementations, the region of interest is the user’s heart, and the plurality of motion states includes at least a plurality of different orientations of the user’s heart. For example, the user’s heart will generally move through these different orientations during the user’s respiration (i.e., as the user breathes), so if the pulse sequence is applied and the readout data is collected as the user is breathing, the region of interest (e.g., the heart) will be moving through a plurality of motion states (e.g., orientations). Each orientation of the heart will generally correspond to a respiratory motion state. Thus, as used herein, the term “motion state” will generally refer to an orientation of the heart that is associated with a specific respiratory motion state.

[0056] In some cases, each of the plurality of motion states is the orientation of the heart at the same position within each respiratory cycle, such as a beginning-inspiration position, an end-inspiration position, a beginning-expiration position, an end-expiration position, etc. In other cases, each of the plurality of motion states is the orientation of the heart at a position within the respiratory cycle that is not necessarily the same position within the respiratory cycle as the other motion states, and some motion states are different from others.

[0057] In some implementations, the readout data can be reconstrued using a low-rank tensor modeling framework. In these implementations, the readout data is modeled as a partially separable low-rank tensor that is the product of a spatial basis tensor, a plurality of temporal basis tensors, and a core tensor that relates the spatial basis tensor to each of the temporal basis tensor. The readout data can thus be represented by the following equation: c / Z = C X-L Uxx2Ucx3Urx4Up x5Ut.

[0058] Here, c / Z represents the low-rank tensor, C represents the core tensor, Uxrepresents the spatial factor tensor, Ucrepresents the temporal factor tensor for cardiac motion, Urrepresents the temporal factor tensor for respiratory motion, Uprepresents the temporal factor tensor for phase accumulation, and Utrepresents the temporal factor tensor for T2* decay. The low-rank tensor can be constructed by estimating <P = C (Uc© Ur© Up© Ut)T(where © represents the Kronecker product) from the subspace of the readout data from the central k- space navigator signals, and then fitting <f> to the remainder of the readout data to recover Ux= argminUx||d — E(UX) ||2+ / ?(UX), wherein d is the is the remainder of the readout data, £■(■) describes multichannel MRI encoding and sampling, and / ?(■) is a regularization term. Additional details related to the construction and extraction of the readout data are found in US 10,436,871; US 11,022,666; US 11,796,616; WO 2020 / 061152; and WO 2021 / 226095, each of which are incorporated by reference herein in its entirety.

[0059] In other implementations, the readout data is reconstructed in a different manner. For example, the readout data can be formulated as x = argminx— b||2+ sr Rx,SR) +lEph Ex, Ap / J}, where x is the target 5D tensor, and 'P rare the spatiotemporal four-dimensional regularizes combining total variation and discrete wavelet transform (DWT) to apply 3D regularization in the spatial domain while ensuring smooth transitions between motion states and linear phase accumulated between echoes.

[0060] Step 230 of method 200 includes generating a plurality of phase maps of the region of interest based at least in part on the readout data. In general, each phase map will correspondto (i) one of the plurality of echo times of the first or second plurality of echo times and (ii) one of the plurality of motion states.

[0061] In some implementations, the plurality of phase maps are arranged into groups that are generated for each of the plurality of motion states. In some implementations, the plurality of motion states will include a single motion state for each respiratory cycle of the user across the time during which the pulse sequence was applied and the readout data was collected. Thus, the plurality of phase maps that are generated can include a group of phase maps (one for each echo time) for each respiration. In other implementations, there may not be a single motion state for each respiration. Thus, in general, the plurality of phase maps that are generated can include a group of phase maps (one for each echo time) for each motion state.

[0062] For example, a first group of phase maps can be generated for the first motion state (e.g., a first orientation of the heart and / or an orientation of the heart during a first respiration), a second group of phase maps can be generated for the second motion state (e.g., a second orientation of the heart and / or an orientation of the heart during a second respiration), and so on. Within the group for each motion state, each of the phase maps will correspond to one of the first or second plurality of echo times. For a given echo time of the first and second plurality of echo times, multiple phase maps will be generated, one phase map and T2*-weighted image for each motion state.

[0063] The plurality of phase maps will therefore include a plurality of groups of phase maps, where each group of phase maps corresponds to a distinct one of the plurality of motion states, and where each group of phase maps includes a phase map for each of the first and second plurality of echo times. Thus, each phase map will generally correspond to a unique combination of (i) one echo time of the first or second plurality of echo times and (ii) one motion state of the plurality of motion states.

[0064] In some cases, each of the plurality of motion states is the orientation of the heart at the same position within each respiratory cycle, such as a beginning-inspiration position, an end-inspiration position, a beginning-expiration position, an end-expiration position, etc. In other cases, each of the plurality of motion states is the orientation of the heart at a position within the respiratory cycle that is not necessarily the same position within the respiratory cycle as the other motion states, and some motion states are different from others.

[0065] In some implementation, step 230 further includes generating a plurality of T2*- weighted images of the region of interest based at least in part on the readout data, in addition to the phase maps. Similar to the phase maps, each T2*-weighted image will correspond to one of the plurality of echo times of the first and second plurality of echo times.

[0066] In some implementations, the plurality of T2*-weighted images are arranged into groups that are generated for each of the plurality of motion states. In some implementations, the plurality of motion states will include a single motion state for each respiratory cycle of the user across the time during which the pulse sequence was applied and the readout data was collected. Thus, the plurality of T2*-weighted images that are generated can include a group of T2*-weighted images (one for each echo time) for each respiration. In other implementations, there may not be a single motion state for each respiration. Thus, in general, the plurality of T2* -weighted images that are generated can include a group of T2* -weighted images (one for each echo time) for motion state.

[0067] For example, a first group of T2*-weighted images can be generated for the first motion state (e.g., a first orientation of the heart and / or an orientation of the heart during a first respiration), a second group of T2* -weighted images can be generated for the second motion state (e.g., a second orientation of the heart and / or an orientation of the heart during a second respiration), and so on. Within the group for each motion state, each of T2* -weighted images will correspond to one of the first or second plurality of echo times. For a given echo time of the first and second plurality of echo times, multiple T2*-weighted images will be generated, one T2*-weighted image for each motion state.

[0068] The plurality of T2*-weighted images will include a plurality of groups of T2*- weighted images, where each group of T2* -weighted images corresponds to a distinct one of the plurality of motion states, and where each group of T2* -weighted images includes a T2*- weighted image for each of the first and second plurality of echo times. Thus, each T2*- weighted image will generally correspond to a unique combination of (i) one echo time of the first or second plurality of echo times and (ii) one motion state of the plurality of motion states.

[0069] Step 240 includes combining the plurality of phase maps to generate a final phase map of the region of interest across all of the first and second plurality of echo times, for each of the motion states. Thus, after step 240, there will be a plurality of final phase maps, one for each motion state. In some implementations, combining the phase maps includes, for each respective motion state, combining the corresponding group of phase maps to generate a single phase map for the respective motion state across the first and second plurality of echo times.

[0070] In implementations where T2*-weighted images are generated from the readout data, generating the final phase maps at step 240 involves the T2*-weighted images. In these implementations, combining the phase maps and the T2*-weighted images for each respective motion state is a multi-step process. First, the phase map for each of the first and second plurality of echo times is unwrapped to expand the scale on which the phase of the tissue in theregion of interest is measured. For example, in some cases the initial phase maps measure the phase on a scale between — TT and +TT, and unwrapping the phase maps expands this scale to between — 2TT and +2TT, or to between — 4TT and +4TT. Second, a mask is generated from each of the T2* -weighted images across the first and second plurality of echo times. Third, for each echo time, the mask from the T2*-weighted image is applied to the unwrapped phase maps. Fourth, the masked unwrapped phase maps of each of the first and second plurality of echo times are combined to generate a final high-fidelity phase map for the respective motion state across all of the first and second plurality of echo times. This process is repeated for all of the plurality of motion states, until a final high-fidelity phase map is generated for each of the plurality of motion states.

[0071] In some implementations, unwrapping the phase maps includes applying an iterative graph-cut-based algorithm to the phase maps. In some implementations, combining the masked unwrapped phase maps includes applying an iterative chemical shift correction algorithm to the masked unwrapped phase maps.

[0072] Step 250 includes standardizing the orientation of the final phase maps across the plurality of motion states. In some implementations, this includes performing affine motion registration and orientation extraction in a multi-step process. First, a plurality of Bofield maps can be generated from the readout data for each of the plurality of motion states. In some implementations, the Bofield maps will all be generated from the same echo time of the first and second plurality of echo times. For example, if there are 10 echo times across the first and second mGRE sequences, then all of the Bofield maps could be generated from the third echo time. Second, the plurality of Bofield maps are co-registered to generate an affine transformation matrix. Third, the affine transformation matrix is applied to the final phase maps of each of the plurality of motion states to standardize the orientation of the final phase maps. In some implementations, the Bofield maps are co-registered to the same motion state, which may be the orientation of the heart at the same position within the respiratory cycle, which in some cases is the end-expiration position.

[0073] Step 260 includes generating a quantitative susceptibility map of the region of interest based on the orientation-standardized final phase map for each of the plurality of motion states. In some implementations, generating the quantitative susceptibility map includes feeding the registered Bofield maps into a single-step calculation of susceptibility through a multiple-orientation sampling algorithm, such as COSMOS or MOR-QSM. Additional details about these algorithms are found in (i) Liu T et al., Calculation of susceptibility throughmultiple orientation sampling (COSMOS): a method for conditioning the inverse problem from measured magnetic field map to susceptibility source image in MRI, Magn Reson Med. 2009;61 : 196-204 and (ii) Chen L. et al., Single-step calculation of susceptibility through multiple orientation sampling. NMR Biomed. 2021;34:e4517; each of which is hereby incorporated by reference herein in its entirety. Multiple spherical mean value (SMV, 15mm max radium) kernels will be used for background removal, and the final artifact-free QSM map / can be derived by solving a multidimensional deconvolution system. The multi -orientation QSM equation will be formulated as an optimization problem to minimize the difference of the measured field and the convoluted field from / maps, which can be solved using an iterative solver, such as LSMR. Additional details about the LSMR iterative solver are found in Fong DC-L and Saunders M. LSMR: An iterative algorithm for sparse least-squares problems, SIAM Journal on Scientific Computing. 2011;33:2950-2971, which is hereby incorporated by reference herein in its entirety.

[0074] In some implementations, method 200 may be modified in any number of ways. For example, in some implementations, the pulse sequence includes interleaved echo times from different mGRE sequences, but does not include generating phase maps (and in some implementations T2*-weighted images) at different motion states. In these implementations, a final phase map can be generated by combining the plurality of phase maps across all of the echo times, and using the single phase map to generate a single-motion state quantitative susceptibility map of the region of interest.

[0075] In some implementations, the pulse sequence does not include interleaved echo times from different mGRE sequences, but does include generate phase maps (and in some implementations T2*-weighted images) at different motion states. In these implementations, one or more phase maps are generated for each different motion state, and the plurality of phase maps (one or more phase maps for each motion state) are used to generate a multi-motion state quantitative susceptibility map of the region of interest.

[0076] Disclosed herein is an example of the features discussed herein.

[0077] Hypothesis - A fast, free-running, motion-compensated, Multi -ORientati on cardiac QSM map (MOR-HDR-QSM) can mitigate confounders and accurately quantify myocardial iron in intramyocardial hemorrhage (IMH) lesions comparable to ex-vivo mass spectrometry and can differentiate myocardial iron changes during novel iron depletion therapies. The following Aims will be pursued to test the above hypothesis:

[0078] Aim 1 : To develop a fast, free-running, high-resolution, confounder-minimized cardiac QSM technique to accurately measure magnetic susceptibility ( / ) in phantoms and healthy hearts.

[0079] Hypothesis: MOR-HDR-QSM can significantly increase the image quality compared to standard R2* maps. A highly efficient, respiratory -motion-resolved mGRE sequence will be developed and combined with a well-posed multi-orientation QSM algorithm for robust cardiac QSM. It will be tested in phantom and healthy pigs.

[0080] Aim 2: To test and validate that confounder-minimized QSM can accurately quantify IMH myocardial iron across all myocardial territories relative to ex-vivo mass spectrometry imaging (MSI).

[0081] Hypothesis: The MOR-HDR-QSM enables higher accuracy of IMH detection and myocardial iron quantification compared to standard R2* maps. It will be tested in pigs with hMI in all coronary territories and validated against the gold standard (MSI).

[0082] Aim 3: To show that confounder-minimized QSM can quantify temporal changes in myocardial iron within hMI in response to iron depletion therapy.

[0083] Hypothesis: The proposed technique can quantify intramyocardial iron changes that are not inferior to the ex-vivo references during iron depletion therapy. The QSM-based myocardial iron will be compared to ex-vivo reference in IMH pigs with and without iron chelation therapy at key time points.

[0084] Impact: The proposed technique can enable accurate intramyocardial iron quantification in hemorrhagic MI patients. It can facilitate a clinically viable tool for longitudinal monitoring of iron within MI territories, which is instrumental in developing novel iron-targeted therapies and monitoring treatment efficacy in the most vulnerable MI patients at the highest risk for MACE. Furthermore, a reliable cardiac QSM can open the doors for studying various susceptibility-sensitive biomarkers with and without contrast agents in the heart.

[0085] SIGNIFICANCE AND PREMISE

[0086] The standard cardiac iron imaging methods (R2* CMR) are inadequate to quantify intramyocardial iron content for IMH assessment. Given that the iron-rich IMH lesion is paramagnetic with high magnetic susceptibility ( / ), it introduces local magnetic field variations, which can be detected by transverse magnetization in MRI. Currently, R2* (— ) cardiac MRI (CMR) is the clinical standard for quantifying myocardial iron content, where mid-ventricular septal R2* is used to identify global myocardial iron overload and has greatlyimproved the management of thalassemia patients. While R2* imaging has been adopted to identify hMI patients, it can fail in more than 20% of the acute MI scans. Particularly in AMI patients with unstable breath holds, ghosting artifacts corrupt long TE images and lead to unreadable R2* maps. To allow a more reliable R2* imaging, free-breathing sequences have been proposed in the past decade. A fast, free-breathing, multi -gradient echo (mGRE) sequence with low-rank-tensor (LRT) formulation has been developed that can acquire whole heart R2* images within 5 minutes with boosted motion resilience. However, mapping myocardial iron in IMH subjects with R2* is still limited by the following challenges.

[0087] 1. BO inhomogeneity induced signal dropout at the heart-lung interfaces in long TE images and R2* maps: Unlike the septal iron measurement in thalassemia patients, the assessment of IMH requires high iron sensitivity with whole heart coverage to characterize local lesions in different coronary territories. Although extended TEs and high-field scanners (e.g., 3T) can boost iron sensitivity, they also lead to strong BO off-resonance artifacts and signal dropout at the lateral wall, making it often nondiagnosable for left circumflex (LCx) infarctions

[0088] 2. The potential bias between R2* and iron content: R2*= R2 +R2’, where R2=^, which makes R2* sensitive to tissue water content ( similar to myocardial T2). Preliminary data shows that myocardial edema in AMI lesions can lead to a 60% underestimation of R2* due to elevated water content. This underestimation can reduce iron sensitivity and bias iron estimation during longitudinal monitoring. Additionally, R2* varies with scanner field strength, leading to further inconsistency in iron assessments

[0089] 3. Sub-optimal imaging efficiency and image resolution can miss small but important lesions: Even small amounts of IMH iron can prolong inflammation and trigger fatty remodeling. Missing subtle lesions can alter management plans and negatively impact patient outcomes.

[0090] These drawbacks present significant challenges for accurate AMI iron quantification during iron-depletion therapy in hMIs, especially for clinical studies where various lesion locations and edema levels can be evident, and scanners with different field strengths can be adopted.

[0091] Quantitative Susceptibility Mapping (QSM) can quantify iron without R2* confounders.

[0092] However, major technical advances are required for its application in hMI patients. A more direct approach to quantifying iron using MRI is Quantitative Susceptibility Mapping(QSM), which utilizes phase information to map the tissue magnetic susceptibility ( / ) and is insensitive to R2* confounders. Although QSM is thoroughly validated in static organs like the brain, applying QSM to the heart remains challenging. QSM estimates / by solving a dipole inversion problem from a measured Bofield map through mGRE phase images. However, the solution to this problem is ill-posed as the field map only measures Bochanges in the main field direction, and it demands the inversion of the linear field to a / map with 3D dipole moments. Most of today’s QSM methods use energy functions that incorporate mathematical and anatomical priors to ensure smooth susceptibility distributions or shared edges with structural images to stabilize the single-orientation inversion, which has achieved reasonable X estimates in algorithms such as MEDI67 and HEIDI. However, the ill-posed inversion process relies heavily on phase map fidelity, smooth susceptibility changes, and requires careful model tuning of hyper-parameters that can lead to disruptive artifacts, such as blurring, shadowing, and streaking. Several attempts have been made to apply single-orientation QSM in the hearts. Notably, a recent study shows that a high-dynamic-range (HDR) phase mapping algorithm can reduce off-resonance artifacts at the heart-lung interface and significantly enhance QSM quality compared to conventional iron images. However, the current singleorientation QSM techniques, including vastly improved HDR-QSM, are still prone to dipole inversion errors. Especially when the irregular flow, cardiac, and respiratory motion can introduce phase errors, and the focal IMH lesion breaks the smoothness of / distribution in hemorrhagic hearts. This significantly compromises its reliability in patient studies.

[0093] Multi -orientation phase sampling offers a solution to the dipole inversion problem. By changing the angle between the target organ and the Bofield, the inversion process is converted into an over-determined linear system. The multi-orientation reconstruction, termed Calculation Of Susceptibility through Multi-Orientation Sampling (COSMOS), has shown improved susceptibility estimates without hyper-parameter tuning and serves as the in-vivo gold standard for brain QSM measurements. However, the requirement for repeated acquisition with 3+ head orientations leads to prolonged study time, making it impractical for clinical adoption. The above challenges make today’s CMR methods unreliable for assessing focal IMH lesions and lead to bias for longitudinal iron quantification during therapy and disease evolution. Therefore, a free-breathing, high-resolution CMR technique that enables motion- resolved, high-fidelity cardiac phase maps for confounder-resistant multi-orientation QSM derivation is needed to reliably quantify myocardial iron content in hMI patients

[0094] Building on our team's expertise, we aim to overcome the key limitations in post- MI intramyocardial iron imaging by systematically developing and validating a clinically viable cardiac QSM sequence for iron quantification in hMI patients. We hypothesize that by achieving the following:

[0095] Accelerating isotropic whole LV imaging: Using Low-Rank-Tensor (LRT)-enabled motion correction and an optimized multi-dimensional Compressed Sensing (CS) algorithm.

[0096] Mitigating QSM confounders: Developing a novel high-phase fidelity, respiratory motion-resolved sequence combined with a Multi-ORientation High-Dynamic-Range QSM pipeline.

[0097] Validating cardiac QSM’s ability in mapping the IMH iron distribution across different myocardial territories: Validating the confounder-minimized cardiac QSM in hMI animals with different infarction territories and compared against ex-vivo gold standard high- resolution Mass Spectrometry Imaging.

[0098] Evaluating imaging efficacy in monitoring therapy: Testing the QSM method in IMH animals undergoing iron-depletion treatment and compared to ex-vivo measurements.

[0099] Multiple innovative strategies are introduced to address the challenges with QSM that are outlined below.

[0100] A first unmet need is the spatial resolution and imaging efficiency needed to detect small but important IMH lesions in acute MI patients. Existing methods include thick slice imaging with interior imaging efficiency from single motion state gating and a short data acceptance window. The proposed innovation is isotropic whole-heart imaging with 36x image acceleration. Acceptance efficiency will be boosted with low-rank tensor (LRT)-facilitated intra-bin motion correction. Image redundancy from respiratory motion and inter-echo similarities will be utilized.

[0101] A second unmet need is to boost phase map fidelity and QSM dynamic range. Existing methods include non-motion-compensated gradients with residual motion error from prospective gating. The proposed innovation is using motion-compensated gradients and boosted ATE coverage with high freedom interleaved echo acquisition, and retrospective selfgating.

[0102] A third unmet need is to reduce QSM dipole inversion errors from single-orientation samplings. Existing methods cause imaging artifacts such as streaking, shadowing, and blurring, that are hyper-parameter dependent. The proposed innovation is artifact-resistant QSM from a respiratory-motion enabled multi-orientation high dynamic range QSM reconstruction.

[0103] A fourth unmet need is image-based validation of QSM images with iron content using mass spectrometry. Existing methods validated with segmental mass spectrometry, which results in a large signal variation from coarse sampling and suboptimal spatial correspondence. The proposed innovation is validation with high-resolution LA-ICP-MS imaging.

[0104] Overall Impact

[0105] Post-MI intramyocardial hemorrhage (IMH) has been identified as the strongest predictor of adverse long-term outcomes. Studies have also shown that decreasing the iron with hemorrhagic MI zone can move the heart toward favorable remodeling. A clinically viable tool to accurately quantify post-MI intramyocardial iron content can serve as the foundation to advance the understanding of iron-induced adverse myocardial remodeling and guide the development of novel iron-targeted treatments. The proposed free-running, confounder- minimized cardiac QSM can overcome key limitations of current methods and offer a fast, noninvasive means to measure accurate myocardial iron content throughout the left ventricle. It can hence facilitate the ongoing development of novel iron-targeted therapies, enable individualized AMI treatment plans, and ultimately improve the long-term outcome of MI patients. Although originally intended for hMI, an accurate QSM map of the myocardium has the potential to uncover novel susceptibility-sensitive biomarkers with and without contrast agents and open opportunities that can aid in the diagnosis and prognosis of various cardiovascular diseases, such as iron overload cardiomyopathy and myocardial oxygenation impairment in heart failure patients.

[0106] APPROACH

[0107] General Research Design

[0108] The key objective of the proposal is to develop a clinically viable technique, named Multi -ORientati on High-Dynamic-Range QSM (MOR-HDR-QSM), to accurately quantify intramyocardial iron content in hMI patients using cardiac QSM. We will develop a fast, high- phase fidelity, respiratory-motion-resolved mGRE sequence and combine it with an overdetermined multi -orientation QSM algorithm to derive high-resolution, confounder- resistant QSM maps of the whole heart. We will validate the QSM measurements against the myocardial iron content with the ex-vivo gold standard (inductively coupled plasma mass spectrometry imaging; ICP-MSI) in hearts with hMI and test if it has sufficient sensitivity to accurately monitor myocardial iron changes during iron-depletion therapy. To achieve the goal, we will first develop the proposed technique and perform systematic validation against standard (navigator-gated, low-resolution) QSM in healthy animals. Subsequently, we willvalidate the precision of the proposed technique in quantifying regional myocardial iron in IMH lesions. We will use pigs with hMI in different coronary supply territories and validate the QSM-based iron content against ex-vivo ground truth measured by high-resolution Laser- Ablation (LA) ICP-MSI. Finally, we will test the proposed QSM approach at different time points during iron-depletion therapy (using an FDA-cleared iron chelator, deferiprone (DFP)) to demonstrate it has sufficient iron sensitivity to resolve myocardial iron changes through the treatment.

[0109] Animal models: Yucatan minipigs (male / female, weight 20-40 kg) will be used throughout the project because of their similarity to human heart anatomy. In total, accounting for 75% of IMH in acute Mis, we expect to use 70 pigs. Details are provided in the Vertebrate Animals section.

[0110] General Image Acquisition: A 3.0T clinical MR system will be used. Standard CMR sequences(Cine, LGE, Maps 1.6 x 1.6 x 6mm3) with ECG triggered and under breath-hold will be prescribed, covering the whole LV.

[0111] Image Reconstruction: Developed acquisitions will be reconstructed offline using a workstation equipped with MATLAB (Mathworks, Natick, MA) and in-house developed MATLAB scripts.

[0112] Scientific Rigor, Transparency of Data, and Biological Variables: Proposed studies will be conducted with rigorous experimental design as described in the following sections. Data will be made available in online supplements to manuscripts for verification by others. An equal number of male and female animals will be used, and the gender differences in measured parameters will be assessed and reported.

[0113] Aim 1 - To develop a fast, free-running, high-resolution, confounder- minimized cardiac QSM technique to accurately measure magnetic susceptibility in phantoms and healthy heart

[0114] The goal of Aim 1 is to develop a fast, free-running cardiac QSM sequence that can reliably acquire high-resolution, confounder-mitigated QSM maps of the whole heart. We will perform the following task with healthy pigs and ex-vivo phantom.

[0115] Technical advancement: (I) Boost imaging efficiency by 36 folds with optimized respiratory motion regularization, inter-echo regularization, and low-rank enabled intra-bin motion correction. (II) Apply motion-compensation gradients and a novel interleaved variable DTE sequence to achieve robust, high-fidelity phase maps. (Ill) Mitigate QSM artifacts and hyper-parameters dependency via MOR-HDR-QSM reconstruction.

[0116] The endpoint of this Aim is to demonstrate that the high-resolution, free-breathing, MOR-HDR-QSM can alleviate QSM artifacts in healthy hearts and accurately measure c in calibrated iron phantom without tuning hyper-parameters. MOR-HDR-QSM will be tested in healthy animals and show significantly increased image quality and signal homogeneity compared to conventional approaches (single orientation, low-resolution, low efficiency). Following, we will test it in a calibrated iron phantom under the influence of off-resonance and motion and show MOR-HDR-QSM can measure comparable / to the calibrated iron concentration.

[0117] Task 1 is to develop a fast, whole-heart, respiratory motion-resolved, high-phase fidelity mGRE sequence and compare its image quality with a conventional navigator-gated low-efficiency sequence. Because the heart rotates against the Bodirection during breathing cycles (FIGS. 9A and 9B), this creates multiple heart orientations that can be used for multiorientation QSM. Preliminary data suggests that a free-running, motion-resolved mGRE imaging framework can achieve isotropic high-resolution (1.5mm3) imaging within 15 minutes.

[0118] Background: High-resolution T2* imaging can boost iron sensitivity and accuracy. However, conventional sequences require multiple breath-holds, making high-resolution imaging unfeasible for AMI patients. Here, a free-breathing retrospectively binned 3D mGRE sequence to enable high-resolution T2*-weighted images was used.

[0119] Methods: A free-running, 3D mGRE sequence was developed using an interleaved Gaussian Z-Y trajectory and low-rank-tensor (LRT) modeling at a 3T clinical scanner. The 3D mGRE images (6 echoes, TE / AT£7TR=1.4 / 2.0 / 14ms, voxel size: 1.5x1.5x1.5mm3, 4 respiratory motion states, Matrix= 182x182x60x6x4, 15mins acq. time) were acquired in IMH pigs(N=3) and compared to the conventional breath-hold (BH) 2D T2* sequence with thick slice (voxel size: 1.5xl.5x6mm3). T2*-w images with an iron-sensitive TE time (TE=9.6ms) were evaluated.

[0120] Results: The developed technique (FIG. 4B) shows boosted image quality compared to the standard BH T2* images (FIG. 4A) and allows accurate high-resolution isotropic T2*-w images compared to ex-vivo ground truth (FIG. 4C).

[0121] Conclusions: Isotropic high-resolution T2* images of the whole heart can be acquired with a respiratory -motion-resolved sequence in 15 mins (6x acceleration compared to fully sampled k-space with HR=60pbm, 84ms cardiac acceptance window). In Task 1, we willbuild a new sequence to further shorten the sequence to a total of 36x acceleration and enable high resolution with high phase fidelity.

[0122] Study Rationale: The standard mGRE sequence without motion-compensated gradients is prone to phase errors from flow and cardiac motion. Conventional motion compensation gradients are limited by hardware performance, causing prolonged ATE and reduced imaging efficiency, which can impair phase processing and introduce unwrapping and misfitting artifacts. To enhance phase map fidelity and correct motion-induced errors, we propose a motion-compensated, interleaved variable ATE sequence (best shown in FIGS. 5 A, 5C, and 5D) that maintains short ATE to guide accurate phase unwrapping. Additionally, to enable multi-orientation QSM without prolonging scan time for unstable IMH patients, we will implement a Low-rank-tensor (LRT)-based motion-correction and multi-dimensional compressed sensing (CS) framework (FIGS. 6A-6C). This approach will allow respiratory motion-resolved mGRE acquisition with 36x image acceleration of the 5D tensor (Spatiai3DxInbrief, an LRT-formulated real-time motion correction algorithm will be adopted to extend the cardiac acceptance window by more than 2 folds and combined with a 5D CS regularization pipeline utilizing the image similarity between respiratory motion states and linear phase accumulation between echoes. Together, high- resolution (1.53mm3), high-phase-fidelity, respiratory motion-resolved mGRE phase maps will be acquired in 6 minutes. In this task, the developed sequence will be acquired in healthy pigs and compared to a conventional (gated, low resolution, low efficiency) mGRE sequence to assess the image quality and phase map SNR for QSM processing.

[0123] The pulse sequence (shown in FIG. 5C) is comprised of two parts: i) Repeat acquisition of central k-space lines every other TR to serve as a highly sampled temporal navigator for LRT realtime-motion estimation and facilitate motion correction of each TR , and ii) two sets of interleaved mGRE readout lines with variable TEs. The images will be acquired using a hybrid cartesian-radial Gaussian stack-of- stars sampling trajectory (FIG. 5D). To push the limit of the minimum TE, after every excitation, mGRE readout with interleaved two sets of five echoes (In-phase TEsi = 2.2, 5.5, 8.8, 12.1, 15.4ms and out-of-phase TEs? = 3.3, 6.6, 9.9, 13.2, 16.5ms) will be applied to break the hardware limit from motion compensation gradients. The echoes are set to be in-phase and out-of-phase to mitigate potential chemical shifts from fat and are formulated for the following compressed sensing-based reconstruction pipeline, which is illustrated in FIGS. 6A-6C.

[0124] Realtime 3D series reconstruction for motion estimation: As shown in FIG. 6A, the acquired data will first be reconstructed into a real-time series using an explicit-subspace low- rank matrix formulation. All images from the first echo will be extracted to derive a low spatial (2x2x8mm3) and high temporal resolution image series(temporal resolution= 20.5ms), which will be used for motion estimation.

[0125] Motion state binning and intra-bin motion correction: As shown in FIGS. 6 A and 6B, real-time images will be used to identify respiratory and cardiac motion states using an image-based ICA algorithm. The lines in irregular motion states will be rejected to avoid motion contamination. The diastolic images will be identified and separated into 4 respiratory bins. A Deep-Learning-Based fast group registration algorithm developed by our team will be adopted to estimate a non-rigid motion field for each diastolic k-space line to correct for intra- bin motion. The motion-corrected diastolic k-space lines will be used to construct a multidimensional CS formulation.

[0126] Reconstruct high-resolution mGRE images with optimized respiratory and echo sparsity regulator: As shown in FIG. 6C. the final high-resolution image will be reconstructed with an optimized CS formulation. The motion-corrected data will be formulated as: x =the target 5D tensor'P are spatiotemporal four-dimensional regularizers combined total variation and discrete wavelet transform (DWT) to apply 3D regularization in spatial domain while ensuring smooth transitions between respiratory motion states (T’sr) and linear phase accumulation between echoes('Pph) respectively. The optimization problem will be solved iteratively using ADMM.

[0127] Data Collection: Images will be collected in healthy pigs. Following scouting and whole-heart shimming, short-axis view images of the LV will be acquired. Data: total scan time = 6 mins; GRE readout (1stTE / minATE = 2.2 / l.lms,TRi / TR2=17 / 3.5ms; number of echoes: 10; resolution: 1.53mm3). Conventional navigator-gated, ECG-gated monopolar mGRE sequences will be prescribed with matching resolution and 20 mins acquisition time with six TEs. The R2* maps and Phase maps will be derived and compared between the sequences.

[0128] Data analyses: The image quality of R2* maps and Phase derived Bo field maps (BoMyo) from the at each echo will be compared. For qualitative comparison, imaging artifacts will be evaluated with 5 scale image quality scores (IQs) by two CMR experts based onblooming, ghosting, and unwrapping artifacts. The mean and coefficient of variance (COV) of the parameters will be measured in all respiratory positions.

[0129] Expected results and interpretations: R2* and BoMyo IQs and COV will be compared between the methods using paired t-tests with a level of significance of 5%. Based on preliminary data, we anticipate the mean septal R2* (both R2*=30±5s-1) and BoMyo (both BoMyo =0±40Hz) are comparable between the proposed and conventional methods. In addition, we expect a lower COV and higher IQ in the proposed method from to the elimination of artifacts.

[0130] Sample-size Estimation: Based on a preliminary study, a sample size of 16 animals can achieve 80% power, and a level of significance of 5% for declaring the method is superior at 0.30 margin of COV using paired t-test, assuming the true difference of 0.51 and a pooled SD of 0.19.

[0131] Task 2 is to develop an overdetermined Multi-Orientation-High-Dynamic-Range QSM(MOR-HDR-QSM) reconstruction pipeline in a calibrated iron phantom and demonstrate that artifact-mitigated QSM maps with accuracy / can be derived in the phantom and healthy pigs without tuning QSM hyper-parameters.

[0132] Preliminary Data: Single-orientation cardiac high-dynamic-range QSM (HDR- QSM) mapping can alleviate off-resonance artifacts at the heart-lung interfaces and boost IMH detection accuracy in animal models.

[0133] Background: QSM is sensitive to phase fidelity and is often compromised at heartlung interfaces. Here, we applied a free-running HDR-QSM technique to improve phase map fidelity and validated ex vivo to assess IMH.

[0134] Methods: Dogs with AMI (N=10) are studied with a free-breathing mGRE sequence (8 echoes, TE / ATE=1.4 / 1.8ms, voxel size: 1.8xl.8x6mm3). In-vivo QSM maps were derived using an HDR-QSM algorithm and compared to conventional QSM and R2* maps.

[0135] Results: HDR-QSM maps (FIG. 7C) showed superior image quality compared to conventional QSM (FIG. 7A) and R2* (FIG. 7B), mitigating off-resonance artifacts at the heart-lung interfaces. When validated against the ex-vivo ground truth, HDR-QSM significantly boosted iron detection accuracy (AUC: HDR-QSM=0.96, Conv.QSM=0.69, R2*=0.75).

[0136] Conclusion: HDR-QSM with fine-tuned hyper-parameters can mitigate off- resonance artifacts present in conventional iron imaging. We will build on these findings in this Task by increasing the number of image orientations to overcome its dependency on hyperparameter tuning.

[0137] Preliminary Data: Multi -orientation QSM using “calculation of susceptibility through multiple orientation sampling” (COSMOS) can eliminate QSM artifacts in ex-vivo IMH hearts without tuning hyper-parameters.

[0138] Background: Single orientation QSM accuracy can be impaired by confounders, such as streaking and shadowing artifacts from the ill-posted dipole inversion. A multiorientation QSM (COSMOS) reconstruction can eliminate these artifacts. However, its effect in the IMH is unknown.

[0139] Methods: COSMOS was tested in ex-vivo hearts with hMI. COSMOS QSM maps were compared to the single orientation QSM approach (MEDI).

[0140] Results: Streaking artifacts in the MEDI images (FIG. 8A) around the hemorrhagic lesion were corrected with the COSMOS (FIG. 8B).

[0141] Conclusion: Multi -orientation QSM effectively overcomes streaking artifacts without tuning hyper-parameters. In Task 2, we will adopt COSMOS to in-vivo cardiac QSM.

[0142] Study Rationale: To mitigate the streaking and blurring artifacts associated with dipole inversion error from the ill-posed conventional QSM algorithms and eliminate hyperparameter tuning, we will develop a novel well-posed cardiac QSM pipeline. This will utilize the respiratory motion-resolved multi -orientation phase maps, which we refer hereinafter as Multi -Orientation-High-Dynamic-Range QSM (MOR-HDR-QSM) (summarized in FIG. 9). In brief, the mGRE images from different respiratory positions with ATE and extended echo number will be processed by a self-guided high-dynamic-range(Auto-HDR) algorithm to derive respiratory motion-resolved high-fidelity myocardial field maps. Following, magnitude images will be used to register the images from each motion state and extract the heart orientations. Finally, the registered field maps will be fed into a multi-orientation algorithm for artifact-resistant QSM. In this Task, we will test MOR-HDR-QSM’s image quality in the same pigs from Task 1 and its accuracy in a calibrated iron phantom with known iron concentrations.

[0143] Multi-Orientation-High-Dynamic-Range QSM (MOR-HDR-QSM) Pipeline:

[0144] Obtain high-fidelity phase maps by combining the Task I images with an HDR algorithm: First a self-guided phase unwrapping approach, named Auto-HDR, will be applied to the complex mGRE data in each respiratory motion state. It will combine the temporal continuity and spatial smoothness of the phase maps from all echoes using an iterative graphcut-based algorithm. Thereafter, the field maps will be processed with an iterative chemical shift correction algorithm (IDEAL) to derive high-fidelity field maps, as shown in part I of FIG. 9.

[0145] Respiratory motion registration and extraction of spatial orientation of the hearts: Field maps obtained at all motion states will be co-registered to that of the end-of expiratory motion state by rigid affine registration using the magnitude images from the third echo (TE=4.5ms) and the Advanced Normalization Tools (ANTs). The rigid affine matrix will be applied to the field maps, and the Bo orientation of all motion states will be determined for multi -orientation QSM derivation, as shown in part II of FIG. 9.

[0146] Well-posed multi -orientation QSM reconstruction: The registered field maps will then be fed into a single-step calculation of susceptibility through a multiple-orientation sampling algorithm, as shown in part III of FIG. 9. Multiple spherical mean values (SMV, 15mm max radius) kernels will be used for background removal and derive the final artifact- free QSM maps ( / ) by solving a multidimensional deconvolution system. The MultiOrientation QSM equation will be formulated as an optimization problem to minimize the difference of the measured field and convoluted field from / maps. It will be solved with an iterative solver (LSMR).

[0147] Ex- vivo phantom study: An iron phantom with known iron concentration and an off-resonance source (air balloon) at the center will be used to test MOR-HDR-QSM. The phantom will be attached to an MR-compatible motion stage (MRI4DMotion Phantom with a rotation stage, QUASAR, Ontario CA) to emulate respiratory motion. The MOR-HDR-QSM sequence will be acquired to derive ( / MOR-HDR) (DTE / lstTE=1.1 / 2.2ms, MaxTE=17ms, Echo number=10; image resol ution= 1.5mm isotropic) images will be acquired for 6 minutes. A reference / map ( / GT) will be acquired using a standard fully sampled mGRE sequence without motion and the air balloon. Four rotation orientations (0°, 5°, 10°, and 15°) will be applied and processed with standard COSMOS. The QSM ground truth ( / GT) will be acquired with a high- resolution (1.5-mm isotropic) fully sampled cartesian image with 10 echoes (DTE / max=l.lms / 17ms). / MOR-HDR will be compared to the / GT.

[0148] Animal study: Healthy pig data from Task 1 will be used to test the MOR-HDR- QSM. In addition, the conventional single-orientation QSM maps will be derived with MEDI( / MEDI) for comparison.

[0149] Data analyses: For ex-vivo phantom, Peak SNR (PSNR = log / MSE)W’"calculated between / MOR-HDR and GT to assess the accuracy of / MOR-HDR. For in-vivo images, data will be analyzed using an AHA 16-segment. Mean, and COV will be calculated and compared between proposed / MOR-HDR and conventional / MEDI.

[0150] Expected results and interpretations: From our preliminary data, we expect ex-vivo / MOR-HDRto closely assemble / GT, with a PSNR>60dB. For the in-vivo study, preliminary data suggests that / MOR-HDR will exhibit higher image quality with higher homogeneity in the myocardium compared to / MEDI. Specifically, we expect a higher IQ score (IQMEDI =3.2±0.3 vs. IQ MOR-HDR.= 3.8±0.3), lower COV, and a mean comparable to the normal values. ( Normal c =-0.01±0.02ppm)

[0151] Aspects of the present disclosure can be implemented on a variety of types of processing devices, such as general purpose computer systems, microprocessors, digital signal processors, micro-controllers, application specific integrated circuits (ASICs), programmable logic devices (PLDs) field programmable logic devices (FPLDs), programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), mobile devices such as mobile telephones, personal digital assistants (PDAs), or tablet computers, local servers, remote servers, wearable computers, or the like.

[0152] Memory storage devices of the one or more processing devices can include a machine-readable medium on which is stored one or more sets of instructions (e.g., software) embodying any one or more of the methodologies or functions described herein. The instructions can further be transmitted or received over a network via a network transmitter receiver. While the machine-readable medium can be a single medium, the term “machine- readable medium” should be taken to include a single medium or multiple media (e.g., a centralized or distributed database, and / or associated caches and servers) that store the one or more sets of instructions. The term “machine-readable medium” can also be taken to include any medium that is capable of storing, encoding, or carrying a set of instructions for execution by the machine and that cause the machine to perform any one or more of the methodologies of the various embodiments, or that is capable of storing, encoding, or carrying data structures utilized by or associated with such a set of instructions. The term “machine-readable medium” can accordingly be taken to include, but not be limited to, solid-state memories, optical media, and magnetic media. A variety of different types of memory storage devices, such as a random access memory (RAM) or a read only memory (ROM) in the system or a floppy disk, hard disk, CD ROM, DVD ROM, flash, or other computer readable medium that is read from and / or written to by a magnetic, optical, or other reading and / or writing system that is coupled to the processing device, can be used for the memory or memories.

[0153] In some implementations, a system to implement steps of methods described herein can include a control system, which may include one or more processors. In someimplementations, a computer program product comprises instructions, which when executed by a computer (e.g., a control system, one or more processing devices and / or processors, etc.), carries out steps of methods described herein. The computer program product may be a non- transitory computer readable medium. In some implementations, a system to implement steps of methods described herein includes a memory device and a control system. The memory device has stored thereon machine-readable instructions. The control system includes one or more processors that are configured to execute the machine-readable instructions to carry out the steps of methods described herein.

[0154] ALTERNATIVE IMPLEMENTATIONS

[0155] Alternative Implementation 1. A method of magnetic resonance (MR) imaging, the method comprising: receiving readout data in response to a pulse sequence being applied to a region of interest of a subject as the region of interest moves through a plurality of motion states, the pulse sequence including a first multi -gradient echo (mGRE) sequence with a first plurality of echo times and a second mGRE sequence with a second plurality of echo times, the first plurality of echo times being interleaved with the second plurality of echo times; generating, based at least in part on the readout data, a plurality of phase maps of the region of interest each corresponding to (i) one of the echo times of the first plurality of echo times or the second plurality of echo times and (ii) one of the plurality of motion states; combining, for each respective motion state of the plurality of motion states, the plurality of phase maps for the respective motion state to generate a final phase map of the region of interest across all of the first and second plurality of echo times for the respective motion state; standardizing an orientation of each of the plurality of final phase maps across all of the plurality of motion states; and generating a quantitative susceptibility map (QSM) of the region of interest based on the orientation-standardized final phase map for each of the plurality of motion states.

[0156] Alternative Implementation 2. The method of Alternative Implementation 1, wherein the first plurality of echo times is different than the second plurality of echo times.

[0157] Alternative Implementation 3. The method of Alternative Implementation 2, wherein each of the first plurality of echo times is different than the second plurality of echo times.

[0158] Alternative Implementation 4. The method of Alternative Implementation 1, wherein the first plurality of echo times includes an identical number echo times as the second plurality of echo times.

[0159] Alternative Implementation 5. The method of Alternative Implementation 4, wherein the first plurality of echo times and the second plurality of echo times each include five echo times.

[0160] Alternative Implementation 6. The method of any one of Alternative Implementations 1 to 5, wherein the first plurality of echo times includes echo times of 1.5 ms, 4.5 ms, 7.5 ms, 10.5 ms, and 13.5 ms.

[0161] Alternative Implementation 7. The method of any one of Alternative Implementations 1 to 6, wherein the second plurality of echo times includes echo times of 3 ms, 6 ms, 9 ms, 12 ms, and 15 ms.

[0162] Alternative Implementation 8. The method of any one of Alternative Implementations 1 to 7, wherein each echo time of the first plurality of echo times and the second plurality of echo times is between about 1 ms and about 15 ms.

[0163] Alternative Implementation 9. The method of any one of Alternative Implementations 1 to 8, wherein the first plurality echo times and the second plurality of echo times are all interleaved with each other.

[0164] Alternative Implementation 10. The method of Alternative Implementation 9, wherein a first echo time of the second plurality of echo times is between a first echo time of the first plurality of echo times and a second echo time of the first plurality of echo times.

[0165] Alternative Implementation 11. The method of Alternative Implementation 10, wherein the second echo time of the first plurality of echo times is between the first echo time of the second plurality of echo times and a second echo time of the second plurality of echo times.

[0166] Alternative Implementation 12. The method of any one of Alternative Implementations 9 to 11, wherein each echo time of the first plurality of echo times is less than at least one echo time of the second plurality of echo times.

[0167] Alternative Implementation 13. The method of any one of Alternative Implementations 9 to 12, wherein each echo time of the second plurality of echo times is greater than at least one echo time of the first plurality of echo times.

[0168] Alternative Implementation 14. The method of any one of Alternative Implementations 9 to 13, wherein a difference between a first echo time of the plurality of echo times and a first echo time of the second plurality of echo times is less than both (i) a difference between the first echo time and a second echo time of the first plurality of echo times and (ii) a difference between the first echo time and a second echo time of the second plurality of echo times.Alternative Implementation 15. The method of any one of Alternative Implementations 1 to 14, wherein each of the first plurality of echo times and the second plurality of echo times are selected by an operator prior to the pulse sequence being applied.

[0169] Alternative Implementation 16. The method of Alternative Implementation 15, wherein the selection of the first and second plurality of echo times is based at least in part on MR hardware used to applying the pulse sequence.

[0170] Alternative Implementation 17. The method of any one of Alternative Implementations 1 to 16, wherein at least one of the first plurality of echo times is identical to at least one of the second plurality of echo times.

[0171] Alternative Implementation 18. The method of any one of Alternative Implementations 1 to 16, wherein the first plurality of echo times includes a different number echo times as the second plurality of echo times.

[0172] Alternative Implementation 19. The method of any one of Alternative Implementations 1 to 18, wherein the region of interest moving through the plurality of motion states is caused by respiration of the user.

[0173] Alternative Implementation 20. The method of Alternative Implementation 19, wherein the region of interest includes a heart of the user, and wherein the heart of the user moves through the plurality of motion states during respiration of the user.

[0174] Alternative Implementation 21. The method of Alternative Implementation 19 or Alternative Implementation 20, wherein the plurality of motion states includes a plurality of orientations of the heart of the user.

[0175] Alternative Implementation 22. The method of any one of Alternative Implementations 1 to 21, wherein both the first mGRE sequence and the second mGRE sequence include one or more motion compensation gradients.

[0176] Alternative Implementation 23. The method of any one of Alternative Implementations 1 to 22, wherein both the first mGRE sequence and the second mGRE sequence include one or more flow compensation gradients.

[0177] Alternative Implementation 24. The method of any one of Alternative Implementations 1 to 23, wherein each of the plurality of phase maps further corresponds to one of the plurality of motion states.

[0178] Alternative Implementation 25. The method of any one of Alternative Implementations 1 to 24, wherein standardizing the orientation of the final phase maps includes: generating a plurality of B0 field maps each corresponding to a distinct one of the plurality of motion states; co-registering the plurality of B0 field maps to generate an affinetransformation matrix; and applying the affine transformation matrix to the final phase map of each of the plurality of motion states.

[0179] Alternative Implementation 26. The method of Alternative Implementation 25, wherein the pulse sequence includes one or more mGRE sequences with a plurality of echo times, and wherein each of the plurality of BO field maps corresponds to an identical one of the plurality of echo times.

[0180] Alternative Implementation 27. The method of Alternative Implementation 25 or Alternative Implementation 26, wherein the plurality of BO field maps are co-registered to an end-expiratory position within a respiratory cycle of the user.

[0181] Alternative Implementation 28. The method of Alternative Implementation 25 or Alternative Implementation 26, wherein the plurality of BO field maps are co-registered to an orientation of a heart of the user associated with an end-expiratory position within a respiratory cycle of the user.

[0182] Alternative Implementation 29. The method of any one of Alternative Implementations 1 to 28, wherein the readout data includes imaging data collected in response to the first mGRE sequence and the second mGRE second being applied, and training data collected in response to the central k-space navigator signal being applied.

[0183] Alternative Implementation 30. The method of any one of Alternative Implementations 1 to 29, wherein the plurality of phase maps includes a plurality of groups of phase maps, each group of phase maps corresponding to a distinct one of the plurality of motion states, each group of phase maps including one phase map for each of the first and second plurality of echo times.

[0184] Alternative Implementation 31. The method of Alternative Implementation 30, wherein the plurality of phase maps includes at least (i) a first group of phase maps each corresponding to different ones of the first and second plurality of echo times and a first one of the plurality of motion states, and (ii) a second group of phase maps each corresponding to different ones of the first and second plurality of echo times and a second one of the plurality of motion states that is different than the first one of the plurality of motion states.

[0185] Alternative Implementation 32. The method of any one of Alternative Implementations 1 to 31, wherein the pulse sequence includes a plurality of sub-sequence that each include the first mGRE sequence, the second mGRE sequence, and the central k-space navigator signal following both the first mGRE sequence and the second mGRE sequence, and wherein applying the pulse sequence includes repeatedly applying the sub-sequence.

[0186] Alternative Implementation 33. The method of Alternative Implementation 32, wherein repeatedly applying the sub-sequence includes applying the sub-sequence with a golden-angle radial trajectory along an x-y plane of k-space and a randomized Gaussian distribution along a z-axis of k-space.

[0187] Alternative Implementation 34. The method of any one of Alternative Implementations 1 to 33, further comprising generating, based at least in part on the readout data, a plurality of T2*-weighted images of the region of interest each corresponding to (i) one of the echo times of the first plurality of echo times or the second plurality of echo times and (ii) one of the plurality of motion states.

[0188] Alternative Implementation 35. The method of Alternative Implementation 34, wherein generating the final phase map of the region of interest across all of the first and second plurality of echo times for each respective motion state includes combining, for each respective motion state of the plurality of motion states, the plurality of phase maps for the respective motion state and the plurality of T2* -weighted images for the respective motion state.

[0189] Alternative Implementation 36. The method of Alternative Implementation 34 or Alternative Implementation 35, wherein the plurality of T2*-weighted images includes at least (i) a first group of T2*-weighted images each corresponding to different ones of the first and second plurality of echo times and a first one of the plurality of motion states, and (ii) a second group of T2*-weighted images each corresponding to different ones of the first and second plurality of echo times and a second one of the plurality of motion states that is different than the first one of the plurality of motion states.

[0190] Alternative Implementation 37. The method of any one of Alternative Implementations 34 to 36, wherein the plurality of T2*-weighted images includes a plurality of groups of T2*-weighted images, each group of T2*-weighted images corresponding to a distinct one of the plurality of motion states, each group of T2* -weighted images including one T2*-weighted image for each of the first and second plurality of echo times.

[0191] Alternative Implementation 38. The method of any one of Alternative Implementations 35 to 37, wherein combining the plurality of phase maps and the plurality of T2*-weighted images includes, for each distinct one of the plurality of motion states, combining the corresponding group of phase maps and the corresponding group of T2*- weighted images to generate the single phase map for the distinct one of the plurality of motion states.

[0192] Alternative Implementation 39. The method of any one of Alternative Implementations 35 to 37, wherein combining the plurality of phase maps and the plurality ofT2*-weighted images includes, for each distinct one of the plurality of motion states: unwrapping the phase map of each of the first and second plurality of echo times to expand a scale on which a phase of tissue in the region of interest is measured; generating a mask from the T2*-weighted image of each of the first and second plurality of echo times; applying the mask of each of the first and second plurality of echo times to a corresponding unwrapped phase map of each of the first and second plurality o echo times; and combine the masked unwrapped phase map of each of the first and second plurality of echo times to generate the final phase map for the region of interest.

[0193] Alternative Implementation 40. The method of Alternative Implementation 39, wherein unwrapping the phase maps includes applying an iterative graph-cut-based algorithm to the phase maps.

[0194] Alternative Implementation 41. The method of Alternative Implementation 39 or Alternative Implementation 40, wherein combining the masked unwrapped phase maps includes applying an iterative chemical shift correction algorithm to the masked unwrapped phase maps.

[0195] Alternative Implementation 42. A method of magnetic resonance (MR) imaging, the method comprising: applying a pulse sequence to a region of interest of a subject, the pulse sequence including a first multi -gradient echo (mGRE) sequence with a first plurality of echo times and a second mGRE sequence with a second plurality of echo times, the first plurality of echo times being interleaved with the second plurality of echo times; collecting readout data in response to the pulse sequence being applied; generating, based at least in part on the readout data, a plurality of phase maps of the region of interest each corresponding to one of the echo times of the first plurality of echo times or the second plurality of echo times; combining the plurality of phase maps to generate a final phase map of the region of interest across all of the first and second plurality of echo times; and generating a quantitative susceptibility map (QSM) of the region of interest based on the final phase map of the region of interest.

[0196] Alternative Implementation 43. A method of magnetic resonance (MR) imaging, the method comprising: receiving readout data in response to a pulse sequence being applied to a region of interest of a subject as the region of interest moves through a plurality of motion states; generating, based at least in part on the readout data, a plurality of phase maps of the region of interest each corresponding to one of the plurality of motion states; standardizing an orientation of each of the plurality of phase maps across all of the plurality of motion states; and generating a quantitative susceptibility map (QSM) of the region of interest based on the orientation-standardized phase map for each of the plurality of motion states.

[0197] Alternative Implementation 44. A method of magnetic resonance (MR) imaging, the method comprising: applying a pulse sequence to a region of interest of a subject, the pulse sequence including a first multi -gradient echo (mGRE) sequence with a first plurality of echo times, a second mGRE sequence with a second plurality of echo times, and a central k-space navigator signal following both the first mGRE sequence and the second mGRE sequence, the first plurality of echo times being interleaved with the second plurality of echo times; collecting readout data in response to the pulse sequence being applied; generating, based at least in part on the readout data, a plurality of phase maps of the region of interest and a plurality of T2*- weighted images of the region of interest, each of the plurality of phase maps and each of the plurality of T2*-weighted images corresponding to one of the echo times of the first plurality of echo times or the second plurality of echo times; combining the plurality of phase maps and the plurality of T2*-weighted images to generate a final phase map of the region of interest across all of the first and second plurality of echo times; and generating a quantitative susceptibility map (QSM) of the region of interest based on the final phase map of the region of interest.

[0198] Alternative Implementation 45. A method of magnetic resonance (MR) imaging, the method comprising: applying a pulse sequence to a region of interest of a subject, the pulse sequence including at least one multi -gradient echo (mGRE) sequence with a first plurality of echo times and a central k-space navigator signal following each of the at least one mGRE sequence; collecting readout data in response to the pulse sequence being applied as the region of interest moves through a plurality of motion states; generating, based at least in part on the readout data, a plurality of phase maps of the region of interest and a plurality of T2*-weighted images of the region of interest, each of the plurality of phase maps and each of the plurality of T2* -weighted images corresponding to one of the plurality of motion states; combining, for each respective motion state of the plurality of motion states, the phase map for the respective motion state and the T2*-weighted image for the respective motion state to generate a final phase map of the region of interest for the respective motion state; standardizing an orientation of each of the plurality of final phase maps across all of the plurality of motion states; and generating a quantitative susceptibility map (QSM) of the region of interest based on the orientation-standardized phase map for each of the plurality of motion states.

[0199] Alternative Implementation 46. A method of magnetic resonance (MR) imaging, the method comprising: applying a pulse sequence to a region of interest of a subject, the pulse sequence including a first multi -gradient echo (mGRE) sequence with a first plurality of echo times, a second mGRE sequence with a second plurality of echo times, and a central k-spacenavigator signal following both the first mGRE sequence and the second mGRE sequence, the first plurality of echo times being interleaved with the second plurality of echo times; collecting readout data in response to the pulse sequence being applied as the region of interest moves through a plurality of motion states; generating, based at least in part on the readout data, a plurality of phase maps of the region of interest and a plurality of T2* -weighted images of the region of interest, each of the plurality of phase maps and each of the plurality of T2*-weighted images corresponding to (i) one of the echo times of the first plurality of echo times or the second plurality of echo times and (ii) one of the plurality of motion states; combining, for each respective motion state of the plurality of motion states, the plurality of phase maps for the respective motion state and the plurality of T2*-weighted images for the respective motion state to generate a final phase map of the region of interest across all of the first and second plurality of echo times for the respective motion state; standardizing an orientation of the plurality of final phase maps across all of the plurality of motion states; and generating a quantitative susceptibility map (QSM) of the region of interest based on the orientation- standardized phase map for each of the plurality of motion states.

[0200] Alternative Implementation 47. A system for performing magnetic resonance (MR) imaging, the system comprising: a magnet operable to provide a magnetic field; a transmitter operable to transmit to a region within the magnetic field; a receiver operable to receive a magnetic resonance signal from the region with the magnetic field; and one or more processors operable to control the transmitter and the receiver, the one or more processors being configured to cause the method of any one of Alternative Implementations 1 to 46 to be performed.

[0201] Alternative Implementation 48. A system for performing magnetic resonance (MR) imaging, the system comprising: a memory storing machine-readable instructions; and one or more processors configured to execute the machine-readable instructions to cause the method of any one of Alternative Implementations 1 to 46 to be performed.

[0202] Alternative Implementation 49. A non-transitory machine-readable medium having stored thereon instructions for performing magnetic resonance (MR) imaging, which when executed by at least one processor, cause the method of any one of Alternative Implementations 1 to 46 to be performed.

[0203] One or more elements or aspects or steps, or any portion(s) thereof, from one or more of any of the Alternative Implementations and / or claims herein can be combined with one or more elements or aspects or steps, or any portion(s) thereof, from one or more of any ofthe other Alternative Implementations and / or claims and / or combinations thereof, to form one or more additional implementations and / or claims of the present disclosure.

[0204] While the present disclosure has been described with reference to one or more particular embodiments or implementations, those skilled in the art will recognize that many changes may be made thereto without departing from the spirit and scope of the present disclosure. Each of these implementations and obvious variations thereof is contemplated as falling within the spirit and scope of the present disclosure. It is also contemplated that additional implementations according to aspects of the present disclosure may combine any number of features from any of the implementations described herein.

Claims

CLAIMSWHAT IS CLAIMED IS:

1. A method of magnetic resonance (MR) imaging, the method comprising: receiving readout data in response to a pulse sequence being applied to a region of interest of a subject as the region of interest moves through a plurality of motion states, the pulse sequence including a first multi -gradient echo (mGRE) sequence with a first plurality of echo times and a second mGRE sequence with a second plurality of echo times, the first plurality of echo times being interleaved with the second plurality of echo times; generating, based at least in part on the readout data, a plurality of phase maps of the region of interest each corresponding to (i) one of the echo times of the first plurality of echo times or the second plurality of echo times and (ii) one of the plurality of motion states; combining, for each respective motion state of the plurality of motion states, the plurality of phase maps for the respective motion state to generate a final phase map of the region of interest across all of the first and second plurality of echo times for the respective motion state; standardizing an orientation of each of the plurality of final phase maps across all of the plurality of motion states; and generating a quantitative susceptibility map (QSM) of the region of interest based on the orientation-standardized final phase map for each of the plurality of motion states.

2. The method of claim 1, wherein the first plurality of echo times is different than the second plurality of echo times.

3. The method of claim 2, wherein each of the first plurality of echo times is different than the second plurality of echo times.

4. The method of claim 1, wherein the first plurality of echo times includes an identical number echo times as the second plurality of echo times.

5. The method of claim 4, wherein the first plurality of echo times and the second plurality of echo times each include five echo times.

6. The method of claim 1, wherein the first plurality of echo times includes echo times of 1.5 ms, 4.5 ms, 7.5 ms, 10.5 ms, and 13.5 ms.

7. The method of claim 1, wherein the second plurality of echo times includes echo times of 3 ms, 6 ms, 9 ms, 12 ms, and 15 ms.

8. The method of claim 1, wherein each echo time of the first plurality of echo times and the second plurality of echo times is between about 1 ms and about 15 ms.

9. The method of claim 1, wherein the first plurality echo times and the second plurality of echo times are all interleaved with each other.

10. The method of claim 9, wherein a first echo time of the second plurality of echo times is between a first echo time of the first plurality of echo times and a second echo time of the first plurality of echo times.

11. The method of claim 10, wherein the second echo time of the first plurality of echo times is between the first echo time of the second plurality of echo times and a second echo time of the second plurality of echo times.

12. The method of claim 9, wherein each echo time of the first plurality of echo times is less than at least one echo time of the second plurality of echo times.

13. The method of claim 9, wherein each echo time of the second plurality of echo times is greater than at least one echo time of the first plurality of echo times.

14. The method of claim 9, wherein a difference between a first echo time of the plurality of echo times and a first echo time of the second plurality of echo times is less than both (i) a difference between the first echo time and a second echo time of the first plurality of echo times and (ii) a difference between the first echo time and a second echo time of the second plurality of echo times.

15. The method of claim 1, wherein each of the first plurality of echo times and the second plurality of echo times are selected by an operator prior to the pulse sequence being applied.

16. The method of claim 15, wherein the selection of the first and second plurality of echo times is based at least in part on MR hardware used to applying the pulse sequence.

17. The method of claim 1, wherein at least one of the first plurality of echo times is identical to at least one of the second plurality of echo times.

18. The method of claim 1, wherein the first plurality of echo times includes a different number echo times as the second plurality of echo times.

19. The method of claim 1, wherein the region of interest moving through the plurality of motion states is caused by respiration of the user.

20. The method of claim 19, wherein the region of interest includes a heart of the user, and wherein the heart of the user moves through the plurality of motion states during respiration of the user.

21. The method of claim 19, wherein the plurality of motion states includes a plurality of orientations of the heart of the user.

22. The method of claim 1, wherein both the first mGRE sequence and the second mGRE sequence include one or more motion compensation gradients.

23. The method of claim 1, wherein both the first mGRE sequence and the second mGRE sequence include one or more flow compensation gradients.

24. The method of claim 1, wherein each of the plurality of phase maps further corresponds to one of the plurality of motion states.

25. The method of claim 1, wherein standardizing the orientation of the final phase maps includes:generating a plurality of Bo field maps each corresponding to a distinct one of the plurality of motion states; co-registering the plurality of Bo field maps to generate an affine transformation matrix; and applying the affine transformation matrix to the final phase map of each of the plurality of motion states.

26. The method of claim 25, wherein the pulse sequence includes one or more mGRE sequences with a plurality of echo times, and wherein each of the plurality of Bo field maps corresponds to an identical one of the plurality of echo times.

27. The method of claim 25, wherein the plurality of Bo field maps are co-registered to an end-expiratory position within a respiratory cycle of the user.

28. The method of claim 25, wherein the plurality of Bo field maps are co-registered to an orientation of a heart of the user associated with an end-expiratory position within a respiratory cycle of the user.

29. The method of claim 1, wherein the readout data includes imaging data collected in response to the first mGRE sequence and the second mGRE second being applied, and training data collected in response to the central k-space navigator signal being applied.

30. The method of claim 1, wherein the plurality of phase maps includes a plurality of groups of phase maps, each group of phase maps corresponding to a distinct one of the plurality of motion states, each group of phase maps including one phase map for each of the first and second plurality of echo times.

31. The method of claim 30, wherein the plurality of phase maps includes at least (i) a first group of phase maps each corresponding to different ones of the first and second plurality of echo times and a first one of the plurality of motion states, and (ii) a second group of phase maps each corresponding to different ones of the first and second plurality of echo times and a second one of the plurality of motion states that is different than the first one of the plurality of motion states.

32. The method of claim 1, wherein the pulse sequence includes a plurality of sub-sequence that each include the first mGRE sequence, the second mGRE sequence, and the central k- space navigator signal following both the first mGRE sequence and the second mGRE sequence, and wherein applying the pulse sequence includes repeatedly applying the subsequence.

33. The method of claim 32, wherein repeatedly applying the sub-sequence includes applying the sub-sequence with a golden-angle radial trajectory along an x-y plane of k-space and a randomized Gaussian distribution along a z-axis of k-space.

34. The method of claim 1, further comprising generating, based at least in part on the readout data, a plurality of T2* -weighted images of the region of interest each corresponding to (i) one of the echo times of the first plurality of echo times or the second plurality of echo times and (ii) one of the plurality of motion states.

35. The method of claim 34, wherein generating the final phase map of the region of interest across all of the first and second plurality of echo times for each respective motion state includes combining, for each respective motion state of the plurality of motion states, the plurality of phase maps for the respective motion state and the plurality of T2*-weighted images for the respective motion state.

36. The method of claim 34, wherein the plurality of T2* -weighted images includes at least (i) a first group of T2*-weighted images each corresponding to different ones of the first and second plurality of echo times and a first one of the plurality of motion states, and (ii) a second group of T2*-weighted images each corresponding to different ones of the first and second plurality of echo times and a second one of the plurality of motion states that is different than the first one of the plurality of motion states.

37. The method of claim 34, wherein the plurality of T2*-weighted images includes a plurality of groups of T2*-weighted images, each group of T2*-weighted images corresponding to a distinct one of the plurality of motion states, each group of T2* -weighted images including one T2* -weighted image for each of the first and second plurality of echo times.

38. The method of claim 35, wherein combining the plurality of phase maps and the plurality of T2*-weighted images includes, for each distinct one of the plurality of motion states, combining the corresponding group of phase maps and the corresponding group of T2*- weighted images to generate the single phase map for the distinct one of the plurality of motion states.

39. The method of claim 35, wherein combining the plurality of phase maps and the plurality of T2*-weighted images includes, for each distinct one of the plurality of motion states: unwrapping the phase map of each of the first and second plurality of echo times to expand a scale on which a phase of tissue in the region of interest is measured; generating a mask from the T2* -weighted image of each of the first and second plurality of echo times; applying the mask of each of the first and second plurality of echo times to a corresponding unwrapped phase map of each of the first and second plurality o echo times; and combine the masked unwrapped phase map of each of the first and second plurality of echo times to generate the final phase map for the region of interest.

40. The method of claim 39, wherein unwrapping the phase maps includes applying an iterative graph-cut-based algorithm to the phase maps.

41. The method of claim 39, wherein combining the masked unwrapped phase maps includes applying an iterative chemical shift correction algorithm to the masked unwrapped phase maps.

42. A method of magnetic resonance (MR) imaging, the method comprising: applying a pulse sequence to a region of interest of a subject, the pulse sequence including a first multi -gradient echo (mGRE) sequence with a first plurality of echo times and a second mGRE sequence with a second plurality of echo times, the first plurality of echo times being interleaved with the second plurality of echo times; collecting readout data in response to the pulse sequence being applied;generating, based at least in part on the readout data, a plurality of phase maps of the region of interest each corresponding to one of the echo times of the first plurality of echo times or the second plurality of echo times; combining the plurality of phase maps to generate a final phase map of the region of interest across all of the first and second plurality of echo times; and generating a quantitative susceptibility map (QSM) of the region of interest based on the final phase map of the region of interest.

43. A method of magnetic resonance (MR) imaging, the method comprising: receiving readout data in response to a pulse sequence being applied to a region of interest of a subject as the region of interest moves through a plurality of motion states; generating, based at least in part on the readout data, a plurality of phase maps of the region of interest each corresponding to one of the plurality of motion states; standardizing an orientation of each of the plurality of phase maps across all of the plurality of motion states; and generating a quantitative susceptibility map (QSM) of the region of interest based on the orientation-standardized phase map for each of the plurality of motion states.

44. A method of magnetic resonance (MR) imaging, the method comprising: applying a pulse sequence to a region of interest of a subject, the pulse sequence including a first multi -gradient echo (mGRE) sequence with a first plurality of echo times, a second mGRE sequence with a second plurality of echo times, and a central k-space navigator signal following both the first mGRE sequence and the second mGRE sequence, the first plurality of echo times being interleaved with the second plurality of echo times; collecting readout data in response to the pulse sequence being applied; generating, based at least in part on the readout data, a plurality of phase maps of the region of interest and a plurality of T2* -weighted images of the region of interest, each of the plurality of phase maps and each of the plurality of T2*- weighted images corresponding to one of the echo times of the first plurality of echo times or the second plurality of echo times;combining the plurality of phase maps and the plurality of T2*-weighted images to generate a final phase map of the region of interest across all of the first and second plurality of echo times; and generating a quantitative susceptibility map (QSM) of the region of interest based on the final phase map of the region of interest.

45. A method of magnetic resonance (MR) imaging, the method comprising: applying a pulse sequence to a region of interest of a subject, the pulse sequence including at least one multi-gradient echo (mGRE) sequence with a first plurality of echo times and a central k-space navigator signal following each of the at least one mGRE sequence; collecting readout data in response to the pulse sequence being applied as the region of interest moves through a plurality of motion states; generating, based at least in part on the readout data, a plurality of phase maps of the region of interest and a plurality of T2* -weighted images of the region of interest, each of the plurality of phase maps and each of the plurality of T2*- weighted images corresponding to one of the plurality of motion states; combining, for each respective motion state of the plurality of motion states, the phase map for the respective motion state and the T2*-weighted image for the respective motion state to generate a final phase map of the region of interest for the respective motion state; standardizing an orientation of each of the plurality of final phase maps across all of the plurality of motion states; and generating a quantitative susceptibility map (QSM) of the region of interest based on the orientation-standardized phase map for each of the plurality of motion states.

46. A method of magnetic resonance (MR) imaging, the method comprising: applying a pulse sequence to a region of interest of a subject, the pulse sequence including a first multi -gradient echo (mGRE) sequence with a first plurality of echo times, a second mGRE sequence with a second plurality of echo times, and a central k-space navigator signal following both the first mGRE sequence and the second mGRE sequence, the first plurality of echo times being interleaved with the second plurality of echo times;collecting readout data in response to the pulse sequence being applied as the region of interest moves through a plurality of motion states; generating, based at least in part on the readout data, a plurality of phase maps of the region of interest and a plurality of T2* -weighted images of the region of interest, each of the plurality of phase maps and each of the plurality of T2*- weighted images corresponding to (i) one of the echo times of the first plurality of echo times or the second plurality of echo times and (ii) one of the plurality of motion states; combining, for each respective motion state of the plurality of motion states, the plurality of phase maps for the respective motion state and the plurality of T2*- weighted images for the respective motion state to generate a final phase map of the region of interest across all of the first and second plurality of echo times for the respective motion state; standardizing an orientation of the plurality of final phase maps across all of the plurality of motion states; and generating a quantitative susceptibility map (QSM) of the region of interest based on the orientation-standardized phase map for each of the plurality of motion states.

47. A system for performing magnetic resonance (MR) imaging, the system comprising: a magnet operable to provide a magnetic field; a transmitter operable to transmit to a region within the magnetic field; a receiver operable to receive a magnetic resonance signal from the region with the magnetic field; and one or more processors operable to control the transmitter and the receiver, the one or more processors being configured to cause the method of any one of claims 1 to 46 to be performed.

48. A system for performing magnetic resonance (MR) imaging, the system comprising: a memory storing machine-readable instructions; and one or more processors configured to execute the machine-readable instructions to cause the method of any one of claims 1 to 46 to be performed.

49. A non-transitory machine-readable medium having stored thereon instructions for performing magnetic resonance (MR) imaging, which when executed by at least one processor, cause the method of any one of claims 1 to 46 to be performed.

Images