System and method of spatiotemporal-accelerated time-resolved MRI

Abstract

The present disclosure provides a method for imaging a subject with a magnetic resonance imaging (MRI) system that achieves millisecond-scale temporal resolution through spatiotemporal-accelerated acquisition. The method acquires accelerated magnetic resonance data by performing a specialized acquisition sequence that applies phase-encoding gradient blips contemporaneously with a bipolar readout during a single excitation. This approach forms a spatiotemporal encoding pattern in which spatial and temporal information are simultaneously encoded, enabling data acquisition at temporal resolutions of one millisecond or less within the single excitation. The method then reconstructs fully sampled spatiotemporal data by estimating missing k-space data points based on spatiotemporal correlations between spatial locations and temporal dynamics in the accelerated data. The method also generates a time series of images from the fully sampled spatiotemporal data, where the time series depicts dynamic processes occurring within the single excitation.

Description

MGH 2024-175-02125141.04928SYSTEM AND METHOD OF SPATIOTEMPORAL-ACCELERATED TIME- RESOLVED MRICROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of U.S. Provisional Patent Application SerialNo. 63 / 723,205, filed on November 21, 2024, and entitled “SYSTEM AND METHOD OF SPATIOTEMPORAL-ACCELERATED TIME-RESOLVED MRI,” which is herein incorporated by reference in its entirety.STATEMENT OF FEDERALLY SPONSORED RESEARCH

[0002] This invention was made with government support under AG083056 and NS 129893 awarded by the National Institutes of Health. The government has certain rights in the invention.BACKGROUND

[0003] Magnetic resonance imaging (MRI) is a technology for non-invasive medical imaging and research applications. Traditional MRI techniques typically operate with temporal resolutions on the order of seconds to minutes, which has been adequate for many clinical and research applications. However, there is growing interest in achieving much higher temporal resolutions to investigate rapid physiological and functional processes, particularly in the brain where fast dynamic changes occur on timescales of milliseconds.

[0004] Achieving ultra-high temporal resolution MRI presents substantial technical challenges. Conventional MRI acquisition methods require multiple excitations to acquire different spatial encodings, such as phase-encoding lines, to generate complete two- dimensional or three-dimensional images. This approach results in extended scan times and makes the acquisition process susceptible to physiological noise and motion artifacts. Additionally, when attempting to capture rapid temporal dynamics, the need for repeated sampling across different spatial encodings can compromise detection sensitivity, particularly for subtle signal changes. These limitations have constrained the ability to study fast physiological processes and have motivated the development of more efficient acquisition strategies that can achieve ultra-high temporal resolution while preserving image quality and spatial coverage.1QB\125141.04928X99564205.3MGH 2024-175-02125141.04928SUMMARY OF THE DISCLOSURE

[0005] According to an aspect of the present disclosure, a method for imaging a subj ect with a magnetic resonance imaging (MRI) system is provided. The method includes acquiring accelerated magnetic resonance data from the subject by performing, with the MRI system, a spatiotemporal-accelerated acquisition sequence that applies phase-encoding gradient blips contemporaneously with a bipolar readout during a single excitation. This forms a spatiotemporal encoding pattern in which spatial information is distributed across the temporal dimension such that spatial information and temporal information are simultaneously encoded, enabling the accelerated magnetic resonance data to be acquired at a temporal resolution of one millisecond or less during the single excitation. The method then reconstructs fully sampled spatiotemporal data from the accelerated magnetic resonance data by estimating missing k- space data points in the accelerated magnetic resonance data based on spatiotemporal correlations between spatial locations and temporal dynamics in the accelerated magnetic resonance data. A time series of images is generated from the fully sampled spatiotemporal data, where the time series of images depicts dynamic processes occurring within the single excitation.BRIEF DESCRIPTION OF THE DRAWINGS

[0006] FIG. 1 illustrates a spatiotemporal-accelerated acquisition sequence with phaseencoding gradient blips applied during a bipolar readout and the generation of magnitude and phase images at millisecond-scale and / or sub-millisecond-scale temporal resolution within each excitation, according to aspects of the present disclosure.

[0007] FIG. 2 depicts a spatiotemporal encoding pattern formed by the spatiotemporal- accelerated acquisition sequence of FIG. 1, according to aspects of the present disclosure.

[0008] FIG. 3 illustrates a flowchart of a method for performing spatiotemporal- accelerated time-resolved MRI, according to aspects of the present disclosure.

[0009] FIG. 4 depicts a block diagram of an MRI system configured to implement the spatiotemporal-accelerated time-resolved imaging methods, according to aspects of the present disclosure.

[0010] FIG. 5 shows phantom results from an example implementation of the disclosed methods. Left) Selected dynamics of the derived phase maps showing the rapid magnetic field changes in the solenoid (center of the phantom) during event 4. middle & right) Both the derived phase and the magnitude were able to capture the rapid milliseconds-scale changes2QB\125141.04928X99564205.3MGH 2024-175-02125141.04928 with consistent timing with the applied electric current during four different tasks. The developed method was able to provide good temporal dynamics while achieving 48x acceleration and allowing for the acquisition of a complete time-series trial at a millisecond / sub-millisecond temporal resolution in a single excitation.

[0011] FIG. 6 show s in-vivo results from an example implementation of the disclosed methods. The original magnitude and phase maps acquired in a single average of 100 ms show good image quality and SNR. The method was able to resolve stable phase and R2* changes across the readout of 75 ms as shown in the standard deviation maps at a millisecond temporal resolution in a single excitation. The spatial resolution of the acquired images is 3x3.5x3mm3.DETAILED DESCRIPTION

[0012] Described here are systems and methods for spatiotemporal-accelerated time- resolved magnetic resonance imaging that may achieve millisecond-scale and / or sub- millisecond-scale temporal resolution through highly accelerated spatiotemporal encoding within single excitations or reduced numbers of excitations. The systems and methods may utilize specialized acquisition schemes that apply phase-encoding gradient blips during bipolar readouts to simultaneously capture spatial and temporal information.

[0013] By applying phase-encoding gradient blips during the bipolar readout gradient, simultaneous spatial and temporal encoding can be achieved wdthin a single excitation. In conventional MRI acquisitions to achieve millisecond or sub-millisecond-scale temporal resolution, different phase-encoding steps (e.g., kylines) are acquired in separate excitations or repetitions. For example, to generate a complete 2D image time series, multiple excitations are needed (i.e. , one for each phase-encoding line). This sequential approach means that, as an example, a 150-ms trial of 2D images may require approximately 7 seconds of total acquisition time when using fully-sampled conventional methods.

[0014] The disclosed spatiotemporal-accelerated approach applies phase-encoding gradient blips at regular intervals during the bipolar readout within a single excitation. The gradient blips may be timed to coincide with echo spacing intervals of approximately 0.5 ms, enabling sub-millisecond temporal resolution. Each echo spacing provides atemporal sampling point while simultaneously acquiring spatial encoding information. Each gradient blip corresponds to a different phase-encoding step, allowing multiple kylines to be acquired during the same readout period. This creates a spatiotemporal encoding pattern where spatial information (e.g., different phase-encoding steps) is distributed across the temporal dimension3QB\125141.04928X99564205.3MGH 2024-175-02125141.04928(e.g., different echo times within the readout). By acquiring multiple phase-encoding lines within a single excitation rather than requiring separate excitations, the method achieves substantial acceleration factors (e.g., upwards of 48x compared to conventional fully-sampled approaches) for achieving millisecond-scale temporal resolution to capture ultra-fast dynamics within the readout. The same 150-ms trial that conventionally requires 7 seconds can be completed in just 150 ms.

[0015] The systems and methods may incorporate sliding window signal processing techniques configured to extract rapid magnitude and phase changes while removing baseline signal variations such as slow Bo-phase evolution and T2-decay effects. Reconstruction approaches may recover fully sampled spatiotemporal data from undersampled acquisitions using kernel-based, neural network-based, or low-rank methods that exploit spatiotemporal correlations.

[0016] Advantageously, the disclosed systems and methods provide capabilities for detecting both periodic and non-periodic dynamic processes, including resting-state brain activities, without requiring cyclic paradigm constraints. Additionally, the disclosed methods may be implemented across various pulse sequence types including EPI-based, gradient-echo, and spin-echo configurations to enable investigation of ultrafast physiological and functional processes previously limited by conventional MRI temporal constraints.

[0017] The disclosed MRI systems and methods may be configured to achieve ultra- high temporal resolution for investigating fast functional and physiological processes. Conventional MRI techniques may face limitations when attempting to capture rapid dynamic processes due to constraints associated with RF pulses, readout sequences, and spatial encoding requirements. These limitations may become particularly pronounced when temporal resolution approaches the millisecond scale.

[0018] Spatiotemporal-accelerated time-resolved MRI systems and methods may address these challenges by providing capabilities for millisecond-scale temporal resolution imaging. Such systems may be configured to measure both magnitude and phase signals associated with ultrafast processes while maintaining acquisition efficiency and robustness to physiological noise and motion artifacts. The spatiotemporal-accelerated approach may utilize specialized acquisition schemes that enable the capture of time-series data within single excitations or reduced numbers of excitations. This approach may differ from conventional methods that t pically require multiple excitations for different spatial encodings, which can4QB\125141.04928X99564205.3MGH 2024-175-02125141.04928 result in extended scan times and increased susceptibility to physiological variations between acquisitions.

[0019] In some cases, spatiotemporal-accelerated time-resolved MRI may incorporate processing methods configured to extract fast temporal dynamics occurring within readout periods. These processing methods may be designed to obtain magnitude and phase changes with adjustable sensitivity parameters, allowing for customization based on specific imaging requirements and target processes.

[0020] The spatiotemporal-accelerated approach may provide advantages for detecting both periodic and non-periodic events, potentially enabling more flexible paradigm designs compared to conventional time-resolved imaging methods. Such flexibility may be particularly beneficial for applications where cyclic or periodic tasks are not feasible or desired. Applications of spatiotemporal-accelerated time-resolved MRI may include investigation of rapid neuronal activities, detection of fast physiological changes, and monitoring of other ultrafast biological processes. The enhanced temporal resolution capabilities may enable research opportunities and clinical applications that were previously limited by conventional MRI temporal constraints.

[0021] In some aspects, spatiotemporal-accelerated time-resolved MRI may be configured to enable detection of various types of dynamic processes. The system configuration may provide flexibility in detecting both periodic events and non-periodic activities without requiring specific paradigm constraints. In some cases, the MRI system may be configured to detect periodic events such as periodic tasks. These periodic events may include repetitive physiological processes or controlled experimental paradigms that occur at regular intervals. The spatiotemporal-accelerated acquisition approach may capture these periodic events with millisecond-scale temporal resolution while maintaining acquisition efficiency. The MRI system configuration may also be capable of detecting non-periodic activities and processes. Non-periodic activities may include spontaneous physiological events that do not follow predictable timing patterns. In some cases, non-periodic processes may include resting-state brain activities, which may occur without external stimulation or controlled timing sequences.

[0022] Resting-state brain activities may represent a particular category of nonperiodic processes that the disclosed spatiotemporal-accelerated MRI systems and methods may be configured to monitor. These activities may occur during periods when a subject is not performing specific tasks or responding to external stimuli. The system may capture these spontaneous activities with the same temporal resolution capabilities used for periodic events.5QB\125141.04928X99564205.3MGH 2024-175-02125141.04928

[0023] Advantageously, the acquisition capabilities of the disclosed spatiotemporal- accelerated MRI systems and methods may enable paradigm designs that do not require cyclic or periodic tasks. This flexibility may allow researchers and clinicians to design imaging protocols based on the specific characteristics of the processes being investigated rather than being constrained by the temporal requirements of conventional MRI methods. In some cases, an MRI system may be configured to switch between detection modes for periodic and nonperiodic events within the same imaging session. This capability may provide versatility for comprehensive studies that examine both controlled periodic responses and spontaneous nonperiodic activities. The detection capabilities for both periodic and non-periodic processes may be achieved through the same underlying spatiotemporal encoding and reconstruction methods. The system may apply consistent temporal resolution parameters regardless of whether the target processes follow periodic or non-periodic patterns.

[0024] The spatiotemporal-accelerated acquisition scheme described in the present disclosure may employ highly accelerated spatiotemporal encoding to produce multiple images within a single excitation or reduced number of excitations. This acquisition approach may differ from conventional methods by utilizing specialized encoding patterns that enable rapid data collection while maintaining image quality and temporal resolution capabilities.

[0025] Referring now to FIG. 1, in some cases, the spatiotemporal -accelerated sequence may utilize phase-encoding gradient blips applied within a bipolar readout. The bipolar readout configuration may provide a framework for implementing the phase-encoding gradient blips at specific intervals during the acquisition sequence. The gradient blips may be timed to coincide with readout periods, allowing for simultaneous spatial and temporal encoding within each excitation. Using bipolar readout gradients allows for phase-encoding gradient blips to be applied at regular intervals throughout the readout period. Each gradient blip may correspond to a different phase-encoding step, allowing for the acquisition of multiple spatial encoding lines within a single excitation. The timing of these gradient blips may be synchronized with the echo spacing intervals to maintain consistent temporal sampling.

[0026] An example of the spatiotemporal encoding pattern formed using this spatiotemporal -accelerated acquisition is illustrated in FIG. 2. This encoding pattern may enable the collection of complete k-space data sets within single excitations. In conventional approaches, different phase-encoding lines may be acquired in separate excitations, requiring multiple repetitions to complete a full image data set. The spatiotemporal approach may acquire6QB\125141.04928X99564205.3MGH 2024-175-02125141.04928 all necessary phase-encoding information within a single excitation period, eliminating the need for multiple excitations.

[0027] The temporal resolution achieved through the spatiotemporal-accelerated acquisition scheme may be adjusted to sub-millisecond scale values. In some cases, the temporal resolution may be configured to less than 1 ms echo spacing, such as approximately 0.5 ms echo spacing. Alternative echo spacing configurations may include approximately 0.3 ms, 0.4 ms, 0.6 ms, 0.7 ms, 0.8 ms, or 0.9 ms echo spacing intervals. In some implementations, the echo spacing may be configured to even shorter intervals, such as approximately 0. 1 ms, 0.2 ms, or 0.25 ms echo spacing. The echo spacing may also be configured within ranges, such as between approximately 0.1 ms to 0.5 ms, 0.2 ms to 0.8 ms, or 0.3 ms to 0.9 ms. These submillisecond temporal resolution values may enable detection of rapid dynamic processes that occur within very short time intervals.

[0028] An example method for performing spatiotemporal-accelerated time-resolved MRI may be implemented as shown in FIG. 3. The method may begin at step 302 by initializing the imaging session, such as by positioning a subject within an MRI system and configuring imaging parameters for millisecond-scale temporal resolution acquisition. The imaging parameters may include selection of echo spacing intervals, readout duration, spatial resolution settings, and temporal window characteristics based on the specific dynamic processes to be investigated.

[0029] At step 304, the method may initiate a spatiotemporal-accelerated acquisition sequence by applying an excitation RF pulse to generate magnetic resonance signals from the subject. The excitation RF pulse may be followed by implementation of the specialized spatiotemporal encoding pattern during the readout period at step 306 to acquire magnetic resonance data from the subject. In some cases, the magnetic resonance data may be undersampled magnetic resonance data. As described above, achieving the spatiotemporal encoding pattern may involve applying phase-encoding gradient blips contemporaneously with a bipolar readout gradient during the single excitation. The gradient blips may be timed at regular intervals throughout the bipolar readout, with each gradient blip corresponding to a different spatial encoding step. The timing may coincide with echo spacing intervals of approximately 1 ms or less, such as 0.5 ms to enable sub-millisecond temporal sampling, while simultaneously acquiring multiple phase-encoding lines. The method includes acquiring magnetic resonance data as spatiotemporal k-space data throughout the readout duration, which as noted above may be spatiotemporal k-space data that is undersampled along at least one7QB\125141.04928X99564205.3MGH 2024-175-02125141.04928 dimension (e g., a phase encoding dimension). The data acquisition may continue for the entire readout period. The spatiotemporal encoding may collect data that contains both spatial and temporal information distributed across the acquisition time within the single excitation.

[0030] The spatiotemporal-accelerated time-resolved acquisition may be implemented using various pulse sequence types to accommodate different imaging requirements and applications. The flexibility of the spatiotemporal encoding approach may enable adaptation to multiple sequence architectures while maintaining millisecond-scale temporal resolution capabilities.

[0031] In some cases, echo planar imaging (EPI)-based sequences may sen e as a foundation for implementing the spatiotemporal-accelerated acquisition scheme. The EPI readout structure may provide a framework for incorporating phase-encoding gradient blips wi thin the echo train. The bipolar readout characteristics of EPI sequences may facilitate the timing of gradient blips to achieve the desired spatiotemporal encoding pattern. The echo spacing intervals in EPI-based implementations may be configured to support sub-millisecond temporal resolution while maintaining spatial encoding efficiency.

[0032] Non-EPI gradient-echo sequences may offer alternative implementations for the spatiotemporal-accelerated method. Gradient-recalled echo (GRE) sequences may be configured with modified readout patterns to accommodate the spatiotemporal encoding requirements. The GRE implementation may utilize gradient blips applied during readout periods to achieve spatial encoding while maintaining rapid temporal sampling. The flexibility of GRE sequence timing may enable customization of echo spacing intervals to optimize temporal resolution for specific applications.

[0033] Spin-echo sequences may provide additional implementation options for applications requiring different contrast characteristics. Spin-echo EPI implementations may combine the refocusing capabilities of spin-echo sequences with the rapid readout advantages of EPI. The spatiotemporal encoding approach may be adapted to work within the timing constraints of the spin-echo preparation and EPI readout combination. The refocusing pulse timing may be coordinated with the spatiotemporal encoding pattern to maintain phase coherence throughout the acquisition.

[0034] Turbo spin-echo (TSE) sequences may represent another spin-echo implementation option for the spatiotemporal-accelerated method. TSE sequences may utilize multiple refocusing pulses to acquire multiple echo trains within each excitation. The spatiotemporal encoding may be applied across the multiple echo trains to achieve enhanced8QB\125141.04928X99564205.3MGH 2024-175-02125141.04928 spatial coverage while maintaining temporal resolution capabilities. The timing of refocusing pulses in TSE implementations may be coordinated with the gradient blip pattern to ensure proper spatial encoding.

[0035] In some cases, multi-shot acquisitions may be implemented to achieve enhanced spatial resolution while maintaining millisecond-scale temporal capabilities. The multi-shot approach may utilize a reduced number of excitations compared to conventional methods through the spatiotemporal acceleration principles. Each excitation in the multi -shot mode may acquire a subset of the total spatial encoding requirements, with the spatiotemporal encoding distributing the remaining encoding steps across the temporal dimension. The reduced number of excitations in multi-shot implementations may still provide substantial acceleration compared to conventional fully-sampled approaches. The spatiotemporal acceleration may enable the acquisition of high spatial resolution images with fewer excitations than would be required using conventional phase-encoding strategies. The multi-shot approach may balance between single-shot efficiency and spatial resolution requirements based on specific application needs. High spatial resolution imaging applications may benefit from the multishot spatiotemporal-accelerated approach. The additional excitations in multi-shot mode may enable finer spatial sampling while maintaining the temporal resolution advantages of the spatiotemporal encoding. The trade-off between temporal efficiency and spatial resolution maybe adjusted through the selection of the number of shots and the distribution of spatial encoding across shots and temporal dimensions.

[0036] Three-dimensional imaging capabilities may be achieved by extending the spatiotemporal encoding to a second k-space dimension that is orthogonal to the first k-space dimension. For example, phase encoding may be applied along both kyand kzspatial dimensions along with the time dimension. The 3D implementation may apply gradient blips in both phase-encoding directions during the readout period. The timing of kyand kz gradient blips may be coordinated to achieve efficient sampling of the three-dimensional k-space while maintaining temporal resolution. The kyand kzencoding patterns in 3D implementations may be designed to optimize the spatiotemporal sampling efficiency. The gradient blip timing may alternate between kyand kzdirections or may apply simultaneous encoding in both directions depending on the specific sequence requirements. The 3D spatiotemporal encoding may enable volumetric imaging at millisecond-scale temporal resolution within single excitations or reduced numbers of excitations.9QB\125141.04928X99564205.3MGH 2024-175-02125141.04928

[0037] The time dimension encoding in 3D implementations may maintain the same sub-millisecond temporal resolution capabilities as 2D implementations. The addition of kzencoding may not compromise the temporal sampling density, as the spatiotemporal approach may distribute the additional spatial encoding requirements across the available temporal sampling points. The 3D capability may enable comprehensive volumetric monitoring of dynamic processes with the same temporal resolution advantages.

[0038] The method may implement signal processing to capture fast temporal dynamics occurring within short readout periods. This processing approach may enable the extraction of both magnitude and phase changes with adjustable sensitivity parameters. The processing method may be designed to address challenges associated with baseline signal variations that can obscure rapid dynamic changes of interest. The method may proceed to step 308 by applying reconstruction algorithms to recover fully sampled spatiotemporal k-space data from the undersampled acquisition. The reconstruction process may utilize spatiotemporal correlations present in the data and may apply kernel-based, neural network-based, or low-rank approaches to estimate missing k-space data points. The selection of reconstruction method may depend on factors such as computational resources, desired image quality, and specific application requirements. The reconstruction may incorporate the processed magnitude and phase change information to guide the recovery of missing data points.

[0039] As one example, a kernel-based reconstruction approach may be implemented using Bo-informed GRAPPA kernels, or other parallel imaging kernels, applied in the spatiotemporal domain. The Bo-informed GRAPPA kernels may incorporate magnetic field inhomogeneity information to improve reconstruction accuracy compared to conventional GRAPPA methods. The spatiotemporal domain application may enable the kernels to exploit correlations across both spatial and temporal dimensions simultaneously. The GRAPPA-based reconstruction method may utilize calibration data to determine kernel weights that characterize the relationships between acquired and missing k-space data points. In some cases, the calibration data may be obtained from a central region of k-space that is more densely sampled during the acquisition. The kernel weights may be calculated based on the calibration data and then applied to reconstruct the missing k-space locations throughout the spatiotemporal data set. The Bo-informed aspect of the GRAPPA kernels may incorporate magnetic field map information to account for off-resonance effects during reconstruction. Magnetic field inhomogeneities may cause phase variations across the imaging volume that can affect the accuracy of conventional reconstruction methods. The Bo-informed kernels may10QB\125141.04928X99564205.3MGH 2024-175-02125141.04928 compensate for these phase variations by incorporating the field map information into the kernel weight calculations.

[0040] Neural network-based reconstruction methods may provide alternative approaches for recovering fully sampled spatiotemporal (e.g., k-t space) data from undersampled acquisitions. In such methods, a pretrained neural network may be trained using pairs of undersampled and fully sampled data sets to leam optimal reconstruction mappings. The training process may enable the neural networks to capture complex relationships between undersampled and fully sampled data that may be difficult to model using conventional analytical approaches. The neural network architecture may be configured with layers designed to process spatiotemporal data efficiently. In some cases, convolutional neural network architectures may be employed to exploit spatial correlations in the image data. Recurrent neural network components may be incorporated to capture temporal dependencies across the time series. The combination of convolutional and recurrent elements may enable the neural network to process both spatial and temporal correlations effectively. Training of neural network reconstruction methods may involve optimization of network parameters using loss functions that measure reconstruction accuracy. The loss functions may compare reconstructed images to reference fully sampled images and calculate error metrics such as mean squared error or structural similarity measures. The training process may iterate through multiple data sets to optimize the network parameters for accurate reconstruction performance.

[0041] Low-rank reconstruction methods may exploit the inherent low-rank structure of spatiotemporal MRI data to recover missing information from undersampled acquisitions. The low-rank approach may be based on the observation that spatiotemporal image data often exhibits redundancy that can be captured through matrix decomposition techniques. This redundancy may arise from correlations between spatial locations and temporal dynamics in the imaging data. Matrix decomposition techniques such as singular value decomposition may be applied to the spatiotemporal data to identify the dominant low-rank components. The decomposition may separate the data into components with different levels of importance based on their singular values. The reconstruction process may retain the most significant components while discarding noise and less important variations. The low-rank reconstruction approach may formulate the reconstruction problem as an optimization that minimizes the rank of the reconstructed data while maintaining consistency with the acquired measurements. Regularization terms may be incorporated to balance between data fidelity and rank11QB\125141.04928X99564205.3MGH 2024-175-02125141.04928 minimization objectives. The optimization may be solved using iterative algorithms that alternately update the data estimates and enforce the low-rank constraints.

[0042] Hybrid reconstruction approaches may combine elements from multiple reconstruction methods to leverage the advantages of different techniques. In some cases, kernel-based methods may be combined with low-rank constraints to improve reconstruction accuracy. Neural network methods may be integrated with conventional reconstruction approaches to enhance performance while maintaining computational efficiency. The selection of reconstruction method may be influenced by the specific characteristics of the acquired data and the computational resources available for processing. Kernel-based methods may provide good reconstruction quality with moderate computational requirements. Neural network approaches may offer superior reconstruction performance but may require more extensive computational resources and training data. Low-rank methods may be particularly effective for data with strong spatiotemporal correlations but may require careful parameter tuning for optimal performance.

[0043] At step 310, a time series of images is reconstructed from the fully sampled spatiotemporal k-space data. As a non-limiting example, the images may be reconstructed using Fourier transform-based reconstruction techniques to convert the k-space data into spatial domain images. The resulting images may represent dynamic processes at each temporal sampling point throughout the readout period.

[0044] At step 312, the method may perform sliding window signal processing on the acquired data and / or reconstructed images to extract rapid magnitude and phase changes. The sliding window may be positioned at the beginning of the readout period and may encompass a selected time interval. The time interval may be configured within a range of approximately 0.5 ms to 20 ms. with specific examples including 1 ms, 2 ms. 4 ms, 8 ms, 10 ms, 15 ms, or other suitable durations depending on desired sensitivity and temporal resolution characteristics. The window size selection may determine the trade-off between sensitivity for detecting small changes and temporal resolution for capturing rapid variations. Shorter window sizes, such as 0.5 ms to 2 ms, may prioritize temporal resolution for detecting very rapid transient events. Intermediate window sizes, such as 3 ms to 6 ms, may provide balanced performance between sensitivity7and temporal resolution. Longer window sizes, such as 8 ms to 20 ms, may enhance sensitivity for measuring subtle signal changes by incorporating more data points in the analysis.12QB\125141.04928X99564205.3MGH 2024-175-02125141.04928

[0045] Step 314 may involve calculating phase changes within the sliding window by determining derivatives or phase differences between adjacent echoes. These calculations may effectively remove slow baseline Bo-phase evolution that could otherwise obscure rapid phase changes associated with dynamic processes of interest while preserving the rapid phase changes associated with the target dynamic processes. The phase change calculations may be performed using mathematical operations that compare phase values across adjacent temporal sampling points. Phase change calculations may be performed by taking the derivative of phase values across adjacent echo measurements. Alternatively, phase changes may be calculated by determining phase differences between adjacent echoes within the sliding window.

[0046] The method may continue to step 316 by calculating magnitude changes within the sliding window to address baseline T2* decay effects, which can mask rapid magnitude variations. The magnitude changes may be calculated through division operations between adjacent echo magnitudes or through T2* decay model fitting. The division approach may normalize magnitude values to remove expected T2* decay behavior, highlighting rapid magnitude variations that deviate from the baseline decay pattern. In some cases, magnitude changes may be calculated by dividing the magnitude of adjacent echoes within the sliding window. Alternative magnitude processing approaches may involve fitting T2* values through a T2* decay signal model. The T2* decay signal model may be applied to the magnitude data within the sliding window to characterize the baseline decay behavior. The fitted T2* parameters may then be used to remove the baseline T2* decay effects, allowing for the detection of rapid magnitude changes that deviate from the expected decay pattern.

[0047] A determination is made at step 318 whether the sliding window has been advanced to each temporal position, if not the method may advance the sliding window to the next temporal position at step 320 after completing the phase and magnitude change calculations for the current window location. The advancement may occur at sub-millisecond intervals, such as 0.5 ms steps, throughout the entire readout period. This advancement pattern may ensure comprehensive coverage of the temporal data while maintaining desired temporal resolution for detecting rapid changes.

[0048] Step 322 may involve generating time series of processed phase and magnitude change values as the sliding window progresses through the readout period. These processed values may represent the dynamic behavior occurring during the acquisition and may capture rapid variations while suppressing slower baseline signal evolution that could interfere with detection of fast dynamic processes.13QB\125141.04928X99564205.3MGH 2024-175-02125141.04928

[0049] At step 324. the reconstructed images may be analyzed to identify and characterize rapid dynamic processes occurring within the subject. The analysis may include detection of both periodic and non-periodic events, quantification of temporal dynamics, and extraction of physiological or functional information from the millisecond-scale temporal resolution data. Periodic events may include cardiac pulsations, respiratory cycles, controlled task-based stimuli, repetitive motor activities, oscillatory neural rhythms, cyclic metabolic processes that occur at regular intervals, and the like. Non-periodic events may encompass spontaneous neuronal firing patterns, irregular vascular fluctuations, transient cognitive responses, unpredictable physiological variations, resting-state brain network activations, stochastic biological processes that do not follow predictable timing patterns, and the like.

[0050] At step 326, the processed results may be output. The processed results may include the reconstructed time series of images, quantitative measurements of dynamic changes, and / or other analysis results. The output may encompass various types of physiological and functional information depending on the specific application. For neuronal activation studies, the output may include maps of rapid neural signal changes occurring at millisecond or sub-millisecond timescales. For brain connectivity analysis, the output may provide dynamic functional connectivity maps, network activation patterns, and inter-regional communication timing. For cardiovascular applications, the output may display rapid cardiac cycle dynamics, blood flow variations, and myocardial contractility changes. For respiratory monitoring, the output may show real-time lung ventilation patterns, airway dynamics, and gas exchange processes. For metabolic studies, the output may include rapid changes in tissue oxygenation, perfusion dynamics, and cellular metabolic activity. For pharmaceutical applications, the output may show drug uptake kinetics, tissue response patterns, and therapeutic effect timing. The output may be formatted for display as color-coded activation maps, temporal waveforms, statistical parametric maps, or quantitative measurement tables, and may be stored in standard medical imaging formats or exported for further analysis depending on the specific application requirements and clinical or research objectives.

[0051] FIG. 5 shows phantom results from an example implementation of the disclosed methods, demonstrating the capability of the spatiotemporal-accelerated time-resolved MRI system to detect rapid electromagnetic field changes in a controlled experimental setup. The phantom experiment utilized a solenoid coil positioned within a 1% NaCl solution doped with CuSO to achieve T2* relaxation properties that mimic human tissue characteristics. The14QB\125141.04928X99564205.3MGH 2024-175-02125141.04928 solenoid was connected to a pulse generator system that could produce controlled electric current pulses of varying durations and timing patterns to simulate rapid dynamic processes.

[0052] Selected dynamics of the derived phase maps are illustrated on the lefthand side of FIG. 5, showing the rapid magnetic field changes in the solenoid (center of the phantom) during event 4, where the phase changes correspond directly to the magnetic field perturbations generated by the electric cunent flowing through the solenoid coil. The phase maps demonstrate the spatial localization of the magnetic field changes, with the strongest signal changes occurring at the center of the phantom where the solenoid is positioned.

[0053] The middle and righthand side of FIG. 5 show both the derived phase and the magnitude were able to capture the rapid milliseconds-scale changes with consistent timing with the applied electric current during four different tasks, including pulses with different durations (20 ms versus 40 ms) and different onset times (30 ms versus 50 ms from the start of the acquisition). The temporal correlation between the applied current pulses and the detected MRI signal changes validates the accuracy of the millisecond-scale temporal resolution capabilities. The developed method was able to provide good temporal dynamics while achieving 48x acceleration compared to conventional fully-sampled acquisition approaches and allowing for the acquisition of a complete time-series trial at a millisecond / sub-millisecond temporal resolution (approximately 0.53 ms) in a single excitation lasting only 150 ms, compared to the 7 seconds that would be required using conventional multi-excitation phaseencoding methods.

[0054] FIG. 6 shows in-vivo results from an example implementation of the disclosed methods, demonstrating the feasibility and performance of the spatiotemporal-accelerated acquisition approach for human brain imaging applications. The in-vivo experiment was conducted during a resting-state condition, where the subject was not performing any specific tasks, thereby testing the method's capability to detect spontaneous physiological variations without requiring controlled periodic stimuli. The original magnitude and phase maps acquired in a single average of 100 ms show good image quality and signal -to-noise ratio (SNR), indicating that the highly accelerated single-excitation approach maintains sufficient data quality for practical imaging applications. The image quality demonstrates that the 48x acceleration factor does not compromise the fundamental imaging performance needed for detecting physiological signals. The method was able to resolve stable phase and R2* changes across the readout of 75 ms as shown in the standard deviation maps at a millisecond temporal resolution (approximately 0.56 ms) in a single excitation, where R2* represents the effective15QB\125141.04928X99564205.3MGH 2024-175-02125141.04928 transverse relaxation rate that is sensitive to local magnetic field variations caused by blood oxygenation changes and other physiological processes. The standard deviation maps provide a measure of the temporal stability and noise characteristics of the phase and magnitude measurements throughout the readout period. The spatial resolution of the acquired images is 3x3.5x3 mm3, which represents a practical compromise between spatial detail and the signal- to-noise requirements for millisecond-scale temporal resolution imaging. This spatial resolution is sufficient for investigating regional brain dynamics while maintaining the temporal resolution necessary for detecting rapid physiological processes that occur on millisecond timescales.

[0055] Referring particularly now to FIG. 4, an example of an MRI system 400 that can implement the methods described here is illustrated. The MRI system 400 includes an operator workstation 402 that may include a display 404, one or more input devices 406 (e.g., a keyboard, a mouse), and a processor 408. The processor 408 may include a commercially available programmable machine running a commercially available operating system. The operator workstation 402 provides an operator interface that facilitates entering scan parameters into the MRI system 400. The operator workstation 402 may be coupled to different servers, including, for example, a pulse sequence server 410, a data acquisition server 412, a data processing server 414, and a data store server 416. The operator workstation 402 and the servers 410, 412, 414, and 416 may be connected via a communication system 440, which may include wired or wireless network connections.

[0056] The pulse sequence server 410 functions in response to instructions provided by the operator workstation 402 to operate a gradient system 418 and a radiofrequency (“RF’’) system 420. Gradient waveforms for performing a prescribed scan are produced and applied to the gradient system 418. which then excites gradient coils in an assembly 422 to produce the magnetic field gradients Gx, G , and Gzthat are used for spatially encoding magnetic resonance signals. The gradient coil assembly 422 forms part of a magnet assembly 424 that includes a polarizing magnet 426 and a whole-body RF coil 428.

[0057] RF waveforms are applied by the RF system 420 to the RF coil 428, or a separate local coil to perform the prescribed magnetic resonance pulse sequence. Responsive magnetic resonance signals detected by the RF coil 428, or a separate local coil, are received by the RF system 420. The responsive magnetic resonance signals may be amplified, demodulated, filtered, and digitized under direction of commands produced by the pulse sequence server 410. The RF system 420 includes an RF transmitter for producing a wide variety of RF pulses used16QB\125141.04928X99564205.3MGH 2024-175-02125141.04928 in MRI pulse sequences. The RF transmiter is responsive to the prescribed scan and direction from the pulse sequence server 410 to produce RF pulses of the desired frequency, phase, and pulse amplitude waveform. The generated RF pulses may be applied to the whole-body RF coil 428 or to one or more local coils or coil arrays.

[0058] The RF system 420 also includes one or more RF receiver channels. An RF receiver channel includes an RF preamplifier that amplifies the magnetic resonance signal received by the coil 428 to which it is connected, and a detector that detects and digitizes the I and Q quadrature components of the received magnetic resonance signal. The magnitude of the received magnetic resonance signal may, therefore, be determined at a sampled point by the square root of the sum of the squares of the I and Q components:

[0059] and the phase of the received magnetic resonance signal may also be determined according to the following relationship:

[0060] The pulse sequence server 410 may receive patient data from a physiological acquisition controller 430. By way of example, the physiological acquisition controller 430 may receive signals from a number of different sensors connected to the patient, including electrocardiograph (“ECG”) signals from electrodes, or respiratory' signals from a respiratory' bellows or other respiratory’ monitoring devices. These signals may be used by the pulse sequence server 410 to synchronize, or “gate,” the performance of the scan with the subject’s heart beat or respiration.

[0061] The pulse sequence server 410 may also connect to a scan room interface circuit 432 that receives signals from various sensors associated with the condition of the patient and the magnet system. Through the scan room interface circuit 432, a patient positioning system 434 can receive commands to move the patient to desired positions during the scan.

[0062] The digitized magnetic resonance signal samples produced by the RF system 420 are received by the data acquisition server 412. The data acquisition server 412 operates in response to instructions downloaded from the operator workstation 402 to receive the realtime magnetic resonance data and provide buffer storage, so that data is not lost by data overrun. In some scans, the data acquisition server 412 passes the acquired magnetic resonance data to the data processor server 414. In scans that require information derived from acquired17QB\125141.04928X99564205.3MGH 2024-175-02125141.04928 magnetic resonance data to control the further performance of the scan, the data acquisition server 412 may be programmed to produce such information and convey it to the pulse sequence server 410. For example, during pre-scans, magnetic resonance data may be acquired and used to calibrate the pulse sequence performed by the pulse sequence server 410. As another example, navigator signals may be acquired and used to adjust the operating parameters of the RF system 420 or the gradient system 418, or to control the view order in which k-space is sampled. In still another example, the data acquisition server 412 may also process magnetic resonance signals used to detect the arrival of a contrast agent in a magnetic resonance angiography (“MRA”) scan. For example, the data acquisition server 412 may acquire magnetic resonance data and processes it in real-time to produce information that is used to control the scan.

[0063] The data processing server 414 receives magnetic resonance data from the data acquisition serv er 412 and processes the magnetic resonance data in accordance with instructions provided by the operator workstation 402. Such processing may include, for example, reconstructing two-dimensional or three-dimensional images by performing a Fourier transformation of raw k-space data, performing other image reconstruction algorithms (e.g., iterative or backproj ection reconstruction algorithms), applying fdters to raw k-space data or to reconstructed images, generating functional magnetic resonance images, or calculating motion or flow images.

[0064] Images reconstructed by the data processing server 414 are conveyed back to the operator workstation 402 for storage. Real-time images may be stored in a data base memory cache, from which they may be output to operator display 402 or a display 436. Batch mode images or selected real time images may be stored in a host database on disc storage 438. When such images have been reconstructed and transferred to storage, the data processing server 414 may notify the data store server 416 on the operator workstation 402. The operator workstation 402 may be used by an operator to archive the images, produce fdms, or send the images via a network to other facilities.

[0065] The MRI system 400 may also include one or more networked workstations 442. For example, a networked workstation 442 may include a display 444, one or more input devices 446 (e.g., a keyboard, a mouse), and a processor 448. The networked workstation 442 may be located within the same facility as the operator workstation 402, or in a different facility, such as a different healthcare institution or clinic.18QB\125141.04928X99564205.3MGH 2024-175-02 125141.04928

[0066] The networked workstation 442 may gain remote access to the data processing server 414 or data store server 416 via the communication system 440. Accordingly, multiple networked workstations 442 may have access to the data processing server 414 and the data store server 416. In this manner, magnetic resonance data, reconstructed images, or other data may be exchanged between the data processing server 414 or the data store server 416 and the networked workstations 442, such that the data or images may be remotely processed by a networked workstation 442.

[0067] The present disclosure has described one or more preferred embodiments, and it should be appreciated that many equivalents, alternatives, variations, and modifications, aside from those expressly stated, are possible and within the scope of the disclosure.19QB\125141.04928X99564205.3

Claims

MGH 2024-175-02 125141.04928CLAIMS1. A method for imaging a subject with a magnetic resonance imaging (MRI) system, comprising: acquiring accelerated magnetic resonance data from the subject by performing, with the MRI system, a spatiotemporal-accelerated acquisition sequence that applies phase-encoding gradient blips contemporaneously with a bipolar readout during a single excitation to form a spatiotemporal encoding pattern in which spatial information is distributed across the temporal dimension such that spatial information and temporal information are simultaneously encoded, wherein the accelerated magnetic resonance data are acquired at a temporal resolution of one millisecond or less during the single excitation; reconstructing fully sampled spatiotemporal data from the accelerated magnetic resonance data by estimating missing k-space data points in the accelerated magnetic resonance data based on spatiotemporal correlations between spatial locations and temporal dynamics in the accelerated magnetic resonance data; and generating a time series of images from the fully sampled spatiotemporal data, wherein the time series of images depicts dynamic processes occurring within the single excitation.

2. The method of claim 1. wherein the temporal resolution is 0.5 milliseconds or less.

3. The method of claim 1, wherein the temporal resolution is approximately 0.5 milliseconds.

4. The method of claim 1, further comprising processing the time series of images using a sliding window' approach to extract magnitude and phase changes at submillisecond temporal intervals.

5. The method of claim 4, wherein the sliding window advances at temporal intervals of approximately 0.5 milliseconds.20QB\125141.04928X99564205.3MGH 2024-175-02125141.049286. The method of claim 4. wherein the sliding window has a size configured to prioritize high temporal frequency information.

7. The method of claim 6, wherein the size of the sliding window is approximately 1 millisecond.

8. The method of claim 4, wherein the sliding window has a size configured to increase sensitivity for measuring phase and magnitude changes.

9. The method of claim 8. wherein the size of the sliding window is approximately 8 milliseconds.

10. The method of claim 4, further comprising calculating phase changes by taking one of a derivative or difference of phase values between adjacent echoes within the sliding window.

11. The method of claim 10, further comprising removing slow baseline Bo-phase evolution from the accelerated magnetic resonance data using the phase change calculations.

12. The method of claim 4. further comprising calculating magnitude changes by dividing magnitudes of adjacent echoes within the sliding window.

13. The method of claim 12, further comprising removing baseline T2* decayeffects from the accelerated magnetic resonance data using at least one of the magnitude change calculations or a T2* fitting.

14. The method of claim 1, wherein reconstructing the fully sampled spatiotemporal data comprises applying parallel imaging reconstruction kernels in a spatiotemporal domain.

15. The method of claim 14, wherein the parallel reconstruction kernels comprise Bo-informed GRAPPA kernels.21QB\125141.04928\99564205.3MGH 2024-175-02125141.0492816. The method of claim 1. wherein reconstructing the fully sampled spatiotemporal data comprises inputting the accelerated magnetic resonance data to a pretrained neural network.

17. The method of claim 1, wherein reconstructing the fully sampled spatiotemporal data comprises using a low-rank reconstruction method that exploits the spatiotemporal correlations between the spatial locations and the temporal dynamics in the accelerated magnetic resonance data.

18. The method of claim 1. wherein the spatiotemporal-accelerated acquisition sequence is implemented using a gradient-echo sequence.

19. The method of claim 1, wherein the spatiotemporal-accelerated acquisition sequence is implemented using a spin-echo sequence.

20. The method of claim 1, wherein the spatiotemporal-accelerated acquisition sequence is implemented using an echo planar imaging sequence.

21. The method of claim 1. wherein the spatiotemporal-accelerated acquisition sequence applies the phase-encoding gradient blips in a first k-space dimension and a second k-space dimension that is orthogonal to the first k-space dimension.

22. The method of claim 1. wherein the dynamic processes comprise neuronal activities occurring at millisecond-scale temporal resolution.

23. The method of claim 1, wherein the dynamic processes comprise periodic events.

24. The method of claim 1, wherein the dynamic processes comprise non-periodic activities.22QB\125141.04928X99564205.3

Images