System and method for single automatic ti prediction for cardiac mri

Abstract

A method includes utilizing a first trained neural network (226) to locate a cardiac region in a TI positioning image MIP image (224) to generate a ROI segmentation mask (228) for the cardiac region. The method also includes utilizing the ROI segmentation mask (228) on the TI positioning image MIP image (224) and a series of MR positioning image time frames (222) to generate a ROI-cropped TI positioning image MIP image (230) and a ROI-cropped MR positioning image time frame (232), respectively. The method includes generating a distance fraction map (234) by calculating the distance between each ROI-cropped MR positioning image time frame (232) and the ROI-cropped TI positioning image MIP image (230). The method includes analyzing the distance fraction map (234) to estimate the span of the MR positioning image time frames for optimal TI frame estimation and utilizing a second trained neural network (248) to determine the optimal TI frame from the span of these MR positioning image time frames.

Description

Background Technology

[0001] The subject matter disclosed herein relates to medical imaging, and more specifically, to a system and method for single-shot automatic time-inversion (TI) prediction for cardiac magnetic resonance imaging (MRI).

[0002] Non-invasive imaging techniques allow for the acquisition of images of the internal structures or features of a patient / subject without the need for invasive procedures. Specifically, such non-invasive imaging techniques rely on various physical principles (such as differential transmission of X-rays through a target volume, sound wave reflection within the volume, paramagnetism of different tissues and materials within the volume, and the disintegration of the target radionuclide within the body) to acquire data and construct images or otherwise represent the observed internal features of a patient / subject.

[0003] During MRI, when material such as human tissue is subjected to a uniform magnetic field (polarization field B0), the individual magnetic moments of the spins within the tissue attempt to align with this polarization field, but precess around it in a random order at their characteristic Larmor frequencies. If the material or tissue is subjected to a magnetic field (excitation field B1) that lies in the xy-plane and is close to the Larmor frequency, the net alignment magnetic moment, or "longitudinal magnetization," is M. z It can be rotated or "tilted" into the xy plane to produce a net transverse magnetic moment M. t After the excitation signal B1 is terminated, a signal is emitted by the excitation spin, and this signal can be received and processed to form an image.

[0004] When these signals are used to generate images, the magnetic field gradient (G) is employed. x G y and G z Typically, the area to be imaged is scanned sequentially according to a measurement cycle, during which these gradient fields vary depending on the specific localization method used. The resulting set of received nuclear magnetic resonance (NMR) signals is digitized and processed to reconstruct an image using one of many well-known reconstruction techniques.

[0005] Contrast-enhanced myocardial imaging with delayed contrast is used to assess various myocardial histopathologies. This imaging requires determining the optimal inversion time (TI). This is determined using TI-scout images (TI-scouts) acquired over multiple frames with different inversion times, and then visual inspection or automation is used to determine the optimal TI. A challenge in practical applications is that the number of frames acquired can vary depending on institutional preferences. As the number of frames changes, the contrast presented by the frames also changes. A common automated approach is to use a sliding window over multiple time frames, which utilizes deep learning-based methods to determine the optimal T1 frame. However, the location of the optimal TI frame can also vary based on patient condition and contrast agent injection parameters. This would require multiple traversals of the sliding window to determine the optimal TI frame, which is not an efficient method. Summary of the Invention

[0006] The following provides an overview of some of the embodiments disclosed herein. It should be understood that these aspects are provided merely to give the reader a brief overview of these specific embodiments, and are not intended to limit the scope of this disclosure. In fact, this disclosure may cover various aspects that may not be set forth below.

[0007] In one embodiment, a computer-implemented method for determining the optimal reversal time (TI) is provided. The computer-implemented method includes obtaining a series of MR positioning image timeframes of a subject from MR positioning image imaging data acquired using a magnetic resonance (MR) scanner utilizing TI positioning image sequences via a processing system including one or more processors. The computer-implemented method also includes obtaining TI positioning image maximum intensity projection (MIP) images from the MR positioning image imaging data. The computer-implemented method further includes using a first trained neural network via the processing system to locate a cardiac region in the TI positioning image MIP images to generate a region of interest (ROI) segmentation mask for that cardiac region. The computer-implemented method further includes using the ROI segmentation mask on the TI positioning image MIP images and the series of MR positioning image timeframes via the processing system to generate ROI-cropped TI positioning image MIP images and ROI-cropped MR positioning image timeframes, respectively. The computer-implemented method further includes generating a distance fraction map via the processing system by calculating the distance between each ROI-cropped MR positioning image timeframe and the ROI-cropped TI positioning image MIP image. The computer-implemented method further includes automatically analyzing the distance fractional map via a processing system to estimate the span of the MR positioning image time frame for optimal TI frame estimation, wherein the span of the MR positioning image time frame encompasses a transition region including phase changes in the cardiac region, such phase changes including blood pool zeroing, distal myocardial zeroing, and initial recovery. The computer-implemented method even further includes determining the optimal TI frame from the span of the MR positioning image time frame via a second trained neural network using the processing system, wherein the determination of the optimal TI frame occurs in a single traversal.

[0008] In another embodiment, a system for determining the optimal reversal time (TI) is provided. The system includes a memory encoding processor-executable routines. The system also includes a processing system comprising one or more processors configured to access the memory and execute the processor-executable routines, wherein the processor-executable routines cause the processing system to perform actions when executed by the processing system. These actions include obtaining a series of MR positioning image timeframes of a subject from MR positioning image imaging data acquired using a magnetic resonance (MR) scanner utilizing TI positioning image sequences. These actions also include obtaining TI positioning image maximum intensity projection (MIP) images from the MR positioning image imaging data. These actions further include utilizing a first trained neural network to locate a cardiac region in the TI positioning image MIP images to generate a region of interest (ROI) segmentation mask for that cardiac region. These actions even further include utilizing the ROI segmentation mask on the TI positioning image MIP images and the series of MR positioning image timeframes to generate ROI-cropped TI positioning image MIP images and ROI-cropped MR positioning image timeframes, respectively. These actions further include generating a distance fraction map by calculating the distance between the MR localization image timeframes cropped for each ROI and the TI localization image MIP images cropped for the ROI. These actions further include automatically analyzing the distance fraction map to estimate the span of the MR localization image timeframes for optimal TI frame estimation, wherein the span of the MR localization image timeframes encompasses a transition region including phase changes in the cardiac region, such phase changes include blood pool zeroing, distal myocardial zeroing, and initial recovery. These actions further include utilizing a second trained neural network to determine the optimal TI frame from the span of the MR localization image timeframes, wherein the determination of the optimal TI frame occurs in a single traversal.

[0009] In another embodiment, a non-transitory computer-readable medium is provided, comprising processor-executable code that, when executed by a processing system comprising one or more processors, causes the processing system to perform actions. These actions include obtaining a series of MR positioning image timeframes of a subject from MR positioning image imaging data acquired using a magnetic resonance (MR) scanner utilizing TI positioning image sequences. These actions also include obtaining TI positioning image maximum intensity projection (MIP) images from the MR positioning image imaging data. These actions further include utilizing a first trained neural network to locate a cardiac region in the TI positioning image MIP images to generate a region of interest (ROI) segmentation mask for that cardiac region. These actions even further include utilizing the ROI segmentation mask on the TI positioning image MIP images and the series of MR positioning image timeframes to generate ROI-cropped TI positioning image MIP images and ROI-cropped MR positioning image timeframes, respectively. These actions also further include generating a distance fraction map by calculating the distance between each ROI-cropped MR positioning image timeframe and the ROI-cropped TI positioning image MIP image. These actions further include automatically analyzing the distance fractional map to estimate the span of the MR localization image time frame for optimal TI frame estimation, wherein the span of the MR localization image time frame encompasses a transition region including phase changes in the cardiac region, such phase changes including blood pool zeroing, distal myocardial zeroing, and initial recovery. These actions further include utilizing a second trained neural network to determine the optimal TI frame from the span of the MR localization image time frame, wherein the determination of the optimal TI frame occurs in a single traversal. Attached Figure Description

[0010] These and other features, aspects, and advantages of the invention will be better understood when reading the following detailed description with reference to the accompanying drawings, in which the same reference numerals denote the same parts throughout the drawings, wherein:

[0011] Figure 1 Embodiments of magnetic resonance imaging (MRI) systems suitable for use with the disclosed techniques according to various aspects of this disclosure are illustrated;

[0012] Figure 2 A flowchart illustrating a method for determining the optimal inversion time (TI) according to various aspects of this disclosure is shown;

[0013] Figure 3 A flowchart illustrating a method for determining a local transition region according to various aspects of this disclosure is shown;

[0014] Figure 4 A schematic diagram illustrating the process for determining the optimal reversal time (TI) according to various aspects of this disclosure is shown;

[0015] Figure 5 The structure and function of the best TI frame estimator model or network according to various aspects of this disclosure are illustrated;

[0016] Figure 6 Examples of various aspects of this disclosure are illustrated. Figure 5 Various tables related to the best TI frame estimator model or network 248;

[0017] Figure 7 Examples of various uses according to this disclosure are illustrated. Figure 5 The first example of the best TI frame estimator model or network;

[0018] Figure 8 Examples of various uses according to this disclosure are illustrated. Figure 5 A second example of the best TI frame estimator model or network;

[0019] Figure 9 Examples of various uses according to this disclosure are illustrated. Figure 5 The third example of the best TI frame estimator model or network;

[0020] Figure 10 Examples of the use of various aspects of this disclosure on pediatric subjects being scanned are illustrated. Figure 5 Examples of the best TI frame estimator models or networks;

[0021] Figure 11 Examples of various uses according to this disclosure are illustrated. Figure 5 Examples of the best TI frame estimator models or networks (e.g., where the best TI is not found and feedback is provided); and

[0022] Figure 12 Examples are illustrated of using the disclosed techniques to estimate the optimal TI frame when it is difficult to determine the optimal TI, according to various aspects of this disclosure. Detailed Implementation

[0023] One or more specific implementations will be described below. To provide a concise description of these implementations, not all characteristics of the actual implementations are described in the specification. It should be understood that in the development of any such actual implementation, as in any engineering or design project, many implementation-specific decisions must be made to achieve the developer's specific objectives, such as complying with system-related and business-related constraints that may differ from implementation to implementation. Furthermore, it should be understood that such development work may be complex and time-consuming, but remains a routine task of design, fabrication, and manufacturing for those skilled in the art who benefit from this disclosure.

[0024] When describing elements of various embodiments of the subject matter of this invention, the articles “a,” “an,” “the,” and “the” are intended to indicate the presence of one or more elements among the elements. The terms “comprising,” “including,” and “having” are intended to be inclusive and mean that additional elements may be present in addition to the listed elements. Furthermore, any numerical examples in the following discussion are intended to be non-limiting, and therefore the additional numerical values, ranges, and percentages are within the scope of the disclosed embodiments.

[0025] While the various aspects discussed below are provided within the context of medical imaging, it should be understood that the disclosed techniques are not limited to such medical settings. In fact, the examples and explanations provided in such medical settings are merely for illustrative purposes by offering real-world examples of implementation and application. However, the disclosed techniques can also be used in other settings, such as image reconstruction for non-destructive inspection of manufactured parts or goods (i.e., quality control or quality inspection applications) and / or non-invasive inspection of packages, boxes, luggage, etc. (i.e., security screening or screening applications). Generally, the disclosed techniques can be used in any imaging or screening setting or in the field of image processing or photography, where a set or class of acquired data undergoes a reconstruction process to generate an image or volume.

[0026] The deep learning (DL) methods discussed in this paper can be based on artificial neural networks and therefore may encompass one or more of the following: deep neural networks, fully connected networks, convolutional neural networks (CNNs), unfolded neural networks, perceptrons, encoders / decoders, recurrent networks, wavelet filter banks, u-nets, generative adversarial networks (GANs), dense neural networks, or other neural network architectures. Neural networks may include shortcuts, activation functions, batch normalization layers, and / or other features. These techniques are referred to as DL techniques in this paper, although the term may also be used specifically with reference to the use of deep neural networks, which are neural networks with multiple layers.

[0027] As discussed in this paper, deep learning (DL) techniques (also known as deep machine learning, hierarchical learning, or deep structured learning) are a branch of machine learning techniques that employ mathematical representations of data and artificial neural networks used to learn and process such representations. For example, DL methods can be characterized as using one or more algorithms to extract or model a high-level abstraction of a class of data of interest. This can be accomplished using one or more processing layers, where each layer typically corresponds to a different level of abstraction and thus may take or utilize different aspects of the initial data or the output of the previous layer (i.e., the hierarchical or cascaded structure of the layers) as the target of the process or algorithm for a given layer. In the context of image processing or reconstruction, this can be characterized as different layers corresponding to different levels of features or resolutions in the data. Generally, the processing from one representation space to the next level of representation space can be viewed as a “stage” of a process. Each stage of the process can be performed by a single neural network or by different parts of a larger neural network.

[0028] These techniques are discussed in relation to MRI in the following disclosure. These techniques can also be used for dynamic data characterization in other imaging modalities.

[0029] This disclosure provides systems and methods for determining the optimal inversion time. Specifically, the disclosed systems and methods combine the localization of a region (e.g., time span) containing the optimal TI frame with the precise TI value within that region, which is subsequently determined. Specifically, only a coarse localization of the cardiac region is required, followed by the utilization of temporal information from the entire image (which provides adaptability to local variations), and then the optimal TI is predicted. The determination of the optimal inversion time occurs in a single pass or single shot, independent of the subjectivity of the TI localization imaging scheme and / or changes in patient condition. Specifically, the disclosed systems and methods rapidly localize transition zones (including phase changes in the cardiac region, including pool zeroing, distal myocardium, and initial recovery) based on frame-by-frame feature distance metrics. The disclosed systems and methods do not involve organ segmentation or preprocessing, thus making the solution very lightweight. The disclosed systems and methods utilize advanced intelligent blocks such as frame attention and spatial attention to accurately predict inversion time, while supporting multiple views (including 4-chamber view, 3-chamber view, 2-chamber view, axial view, long axis view (4-chamber), and short axis view).

[0030] Considering the above, Figure 1 The magnetic resonance imaging (MRI) system 100 is schematically illustrated as including a scanner 102, scanner control circuitry 104, and system control circuitry 106. According to the embodiments described herein, the MRI system 100 is typically configured to perform MR imaging.

[0031] System 100 further includes: a remote access and storage system or device, such as a Picture Archiving and Communication System (PACS) 108; or other devices, such as remote radiology equipment, enabling on-site or remote access to data acquired by system 100. In this way, MR data can be acquired and then processed and evaluated on-site or remotely. While MRI system 100 may include any suitable scanner or detector, in the illustrated embodiment, system 100 includes a whole-body scanner 102 with a housing 120 through which an aperture 122 is formed. An examination table 124 is movable into the aperture 122 to allow a patient 126 (e.g., a subject) to be positioned therein for imaging of selected anatomical structures within the patient's body.

[0032] Scanner 102 includes a series of associated coils for generating a controlled magnetic field used to excite gyromagnetic material within the anatomical structures of the patient being imaged. Specifically, a primary magnetic coil 128 is provided to generate a primary magnetic field B0 generally aligned with an aperture 122. A series of gradient coils 130, 132, and 134 allow the generation of a controlled gradient magnetic field during the examination sequence for positional encoding of certain gyromagnetic nuclei within the patient 126. A radio frequency (RF) coil 136 (e.g., an RF transmit coil) is configured to generate radio frequency pulses for exciting certain gyromagnetic nuclei within the patient. In addition to the coils that may be located locally on scanner 102, system 100 also includes a set of receiving coils or RF receiving coils 138 (e.g., an array of coils) configured to be placed proximal to (e.g., against) the patient 126. For example, receiving coils 138 may include cervical / thoracic / lumbar (CTL) coils, head coils, single-sided spinal coils, etc. Generally, the receiving coil 138 is placed near or above the patient 126 in order to receive weak RF signals generated by certain magnetic nuclei in the patient's body when the patient 126 returns to its relaxed state (weak in relation to the transmission pulses generated by the scanner coil).

[0033] The various coils of system 100 are controlled by external circuitry to generate desired fields and pulses and to read out emissions from the gyromagnetic material in a controlled manner. In an illustrated embodiment, a main power supply 140 powers the primary field coil 128 to generate a primary magnetic field Bo. Power inputs (e.g., power from a utility or grid), a power distribution unit (PDU), a power supply (PS), and drive circuitry 150 may together provide power to cause gradient field coils 130, 132, and 134 to generate pulses. Drive circuitry 150 may include amplification and control circuitry for supplying current to the coils according to a sequence of digitized pulses output by scanner control circuitry 104.

[0034] Another control circuit 152 is provided for regulating the operation of the RF coil 136. Circuit 152 includes a switching device for alternating between an active operating mode and a passive operating mode, wherein the RF coil 136 transmits a signal and does not transmit a signal, respectively. Circuit 152 also includes an amplifier configured to generate RF pulses. Similarly, a receiving coil 138 is connected to a switch 154 capable of switching the receiving coil 138 between a receiving mode and a non-receiving mode. Thus, in receiving mode, the receiving coils 138 resonate with the RF signal generated by the release of the vortex nucleus within the patient 126, and in non-receiving mode, they do not resonate with the RF energy from the transmitting coil (i.e., coil 136) to prevent undesirable operation. Additionally, the receiving circuit 156 is configured to receive data detected by the receiving coil 138 and may include one or more multiplexing and / or amplification circuits.

[0035] It should be noted that although the scanner 102 and the control / amplification circuit described above are illustrated as being coupled by a single wire, in practice, many such wires may exist. For example, separate wires may be used for control, data communication, power transmission, etc. Furthermore, appropriate hardware may be provided along each type of wire for proper handling of data and current / voltage. In practice, various filters, digitizers, and processors may be provided between the scanner and either or both of the scanner control circuit 104 and system control circuit 106.

[0036] As shown in the figure, the scanner control circuit 104 includes an interface circuit 158 ​​that outputs signals for driving the gradient field coil and the RF coil, and for receiving data representing magnetic resonance signals generated in the examination sequence. The interface circuit 158 ​​is coupled to a control and analysis circuit 160. Based on a defined scheme selected via the system control circuit 106, the control and analysis circuit 160 executes commands for driving circuits 150 and 152.

[0037] The control and analysis circuit 160 is also used to receive magnetic resonance signals and perform subsequent processing before sending the data to the system control circuit 106. The scanner control circuit 104 also includes one or more memory circuits 162 that store configuration parameters, pulse sequence descriptions, inspection results, etc. during operation.

[0038] Interface circuitry 164 is coupled to control and analysis circuitry 160 for exchanging data between scanner control circuitry 104 and system control circuitry 106. In some embodiments, control and analysis circuitry 160, while exemplified as a single unit, may include one or more hardware devices. System control circuitry 106 includes interface circuitry 166 that receives data from scanner control circuitry 104 and transmits data and commands back to scanner control circuitry 104. Control and analysis circuitry 168 may include a CPU in a general-purpose or special-purpose computer or workstation. Control and analysis circuitry 168 is coupled to memory circuitry 170 to store programming code for operating the MRI system 100, and to store processed image data for subsequent reconstruction, display, and transmission. The programming code may execute one or more algorithms configured to perform reconstruction of the acquired data as described below when executed by a processor. In some embodiments, memory circuitry 170 may store one or more neural networks. For example, the neural network may include a first trained neural network (e.g., a region of interest (ROI) detection model or network) used to locate a heart region in a TI-localized MIP image to generate a region of interest (ROI) segmentation mask for that heart region. The neural network may also include a second trained neural network (e.g., an optimal TI frame estimator model or network) for determining or estimating the optimal TI frame. The neural network forms a deep learning-based neural framework. In some embodiments, the techniques disclosed herein can occur on a separate computing device having processing and memory circuitry.

[0039] The processing unit (e.g., a microprocessor or processing circuitry) and memory (such as those residing in scanner control circuitry 104 and / or system control circuitry 106) of the magnetic resonance imaging system 100 can be used to execute stored software code, instructions, or routines for acquiring and processing MR data. As used herein, the term "code" or "software code" refers to any instruction or set of instructions that controls the magnetic resonance imaging system 100. The code or software code can exist in the following forms: a computer-executable form, such as machine code, which is a set of instructions and data directly executed by the processing unit of scanner control circuitry 104 and / or system control circuitry 106; a human-understandable form, such as source code, which can be compiled for execution by the processing unit of scanner control circuitry 104 and / or system control circuitry 106; or an intermediate form, such as object code, which is generated by a compiler. In some embodiments, the magnetic resonance imaging system 100 may include multiple controllers.

[0040] For example, the memory may store processor-executable software code or instructions (e.g., firmware or software) tangibly stored on a non-transitory computer-readable medium. Additionally or alternatively, the memory may store data. As an example, the memory may include volatile memory (such as random access memory (RAM)) and / or non-volatile memory (such as read-only memory (ROM), flash memory, hard disk drive, or any other suitable optical, magnetic, or solid-state storage medium or combinations thereof). Furthermore, the processing unit may include multiple microprocessors, one or more "general-purpose" microprocessors, one or more application-specific microprocessors, and / or one or more application-specific integrated circuits (ASICs) or some combination thereof. For example, the processing unit may include one or more Reduced Instruction Set Computing (RISC) or Complex Instruction Set Computing (CISC) processors. The processing unit may include multiple processors and / or the memory may include multiple memory devices.

[0041] In some implementations (e.g., for determining the optimal reversal time (TI) in a single pass or traversal), the processing unit is configured to obtain a series of MR positioning image timeframes of a subject from MR positioning image imaging data acquired using a magnetic resonance (MR) scanner with a TI positioning image sequence. In some implementations, the series of MR positioning image timeframes includes a 4-cavity view, a 3-cavity view, a 2-cavity view, an axial view, a major axis view, or a minor axis view. The processing unit is configured to obtain TI positioning image maximum intensity projection (MIP) images from the MR positioning image imaging data. The processing unit is configured to utilize a first trained neural network to locate a cardiac region in the TI positioning image MIP image to generate a region of interest (ROI) segmentation mask for that cardiac region. The processing unit is configured to use the ROI segmentation mask on the TI positioning image MIP image and the series of MR positioning image timeframes to generate ROI-cropped TI positioning image MIP images and ROI-cropped MR positioning image timeframes, respectively. The processing unit is configured to generate a distance fraction map by calculating the distance between the MR localization image time frame cropped for each ROI and the TI localization image MIP image cropped for the ROI. The processing unit is configured to automatically analyze the distance fraction map to estimate the span of the MR localization image time frame for optimal TI frame estimation, wherein the span of the MR localization image time frame encompasses a transition region including phase changes in the cardiac region, such phase changes including blood pool zeroing, distal myocardial zeroing, and initial recovery. The processing unit is configured to determine the optimal TI frame from the span of the MR localization image time frame using a second trained neural network, wherein the determination of the optimal TI frame occurs in a single traversal.

[0042] In some embodiments, the processing unit may be configured to normalize the ROI-cropped TI positioning image MIP image and the ROI-cropped MR positioning image time frame before generating the distance fraction map. In some embodiments, the processing unit may be configured to, when automatically analyzing the distance fraction map to estimate the span of the MR positioning image time frame for optimal TI frame estimation, determine the point with the maximum distance on the distance fraction map, where this point represents a transition region, and select from a series of MR positioning image time frames multiple temporally consecutive frames located on both sides of the transition region and including the transition region as the span of the MR positioning image time frame.

[0043] In some embodiments, the processing unit can be configured to train a neural network to generate a second trained neural network using ROI-cropped MR localization image timeframes with a sufficiently large time window size to provide an accurate estimate of the optimal TI frame for the blood pool and / or myocardium. In some embodiments, the span of the MR localization image timeframes is within the time window size.

[0044] In some embodiments, the processing unit may be configured to input the span of the MR positioning image timeframes and the ROI segmentation mask into the second trained neural network when determining the optimal TI frame using the second trained neural network, and output the optimal TI frame from the span of the MR positioning image timeframes. In some embodiments, the processing unit may be configured to determine the optimal TI frame using the second trained neural network by predicting a score between 0 and 1 for each MR positioning image timeframe within the span, wherein the score for the optimal TI frame is close to or equal to 1. In some embodiments, the processing unit may be configured to output a user-perceptible indication when none of the MR positioning image timeframes within the span have a sufficiently high corresponding score to be considered the optimal TI frame. In some embodiments, the processing unit may be configured to output a user-perceptible indication of the optimal TI based on the optimal TI frame and a pulse sequence map of the TI positioning image sequence.

[0045] Additional interface circuitry 172 may be provided for exchanging image data, configuration parameters, etc., with external system components, such as remote access and storage device 108. Finally, system control and analysis circuitry 168 may be communicatively coupled to various peripheral devices for use in facilitating the operator interface and for generating hard copies of the reconstructed images. In the illustrated embodiment, these peripheral devices include a printer 174, a monitor 176, and a user interface 178, which includes devices such as a keyboard, mouse, and touchscreen (e.g., integrated with monitor 176).

[0046] Figure 2A flowchart illustrating method 180 for determining the optimal inversion time (TI) is provided. One or more steps of method 180 may be performed by... Figure 1 The processing circuitry of the magnetic resonance imaging system 100 or a remote computing device is used to execute the steps. One or more steps of method 180 can be performed simultaneously and / or in conjunction with... Figure 2 The different execution sequences are shown. Method 180 supports multiple views (i.e., 4-cavity view, 3-cavity view, 2-cavity view, axis view, long axis view (4 cavities), and short axis view).

[0047] Method 180 includes acquiring MR localization imaging data (e.g., three-dimensional (3D) imaging data) of a subject (e.g., the chest region of the subject) using an MR scanner with a TI localization imaging sequence (box 182). Method 180 also includes obtaining a series of MR localization imaging time frames of the subject from the MR localization imaging data acquired using a magnetic resonance (MR) scanner with a TI localization imaging sequence (box 184). Method 180 further includes obtaining TI localization maximum intensity projection (MIP) images from the MR localization imaging data (box 186). Method 180 even further includes utilizing a first trained neural network (e.g., a region of interest (ROI) detection model or network) to localize a heart region in the TI localization imaging MIP images to generate a region of interest (ROI) segmentation mask for that heart region (box 188). In some embodiments, the first trained neural network is configured to utilize an object detection method. In some embodiments, the first trained neural network is configured to utilize a semantic segmentation method. Method 180 further includes using the ROI segmentation mask on the TI positioning image MIP image and a series of MR positioning image time frames to generate ROI-cropped TI positioning image MIP images and ROI-cropped MR positioning image time frames respectively (box 190).

[0048] Method 180 includes training a neural network to generate a second trained neural network (e.g., an optimal TI frame estimator model or network) using a subset of ROI segmentation masks and MR positioning image timeframes. This subset of MR positioning image timeframes has a sufficiently large time window size to provide an accurate estimate of the optimal TI frame for transition regions (e.g., during the positioned transition) that include phase changes in the myocardium (e.g., blood pool zeroing, distal myocardial zeroing, and initial recovery) during the TI positioning image sequence (box 192). The second trained neural network is configured to determine or estimate the optimal TI frame. As described in more detail below, in some embodiments, the second trained neural network is a dual-attention (Datt) three-dimensional (3D) CNN. Method 180 further includes generating a distance fraction map by calculating the distance between each ROI-cropped MR positioning image timeframe and the ROI-cropped TI positioning image MIP image (box 194). In some embodiments, the ROI-cropped TI positioning image MIP image and the ROI-cropped MR positioning image timeframe are normalized before generating the distance fraction map. Method 180 further includes automatically analyzing the distance fraction map to estimate the span of the MR localization image time frame for optimal TI frame estimation, wherein the span of the MR localization image time frame covers a transition region that includes phase changes in the cardiac region, including blood pool zeroing, distal myocardial zeroing, and initial recovery (box 196). The span of the MR localization image time frame is within the expected time window size of the second trained neural network.

[0049] Method 180 further includes determining an optimal TI frame from the span of MR positioning image timeframes using a second trained neural network, wherein the determination of the optimal TI frame occurs in a single pass (or single iteration) (box 198). In some embodiments, determining the optimal TI frame using the second trained neural network includes inputting the span of the MR positioning image timeframes into the second trained neural network. In some embodiments, method 180 includes outputting the optimal TI frame from the span (e.g., for display) from the second trained neural network (box 200). In some embodiments, determining the optimal TI frame using the second trained neural network includes predicting a score (predicted probability score) between 0 and 1 for each MR positioning image timeframe within the span, wherein the score of the TI frame is close to or equal to 1. In some embodiments, the score is compared to a threshold (e.g., 0.5 or another set threshold), and only scores equal to or higher than the threshold can potentially be considered the optimal TI. In some embodiments, method 180 includes outputting the scores of the optimal TI frame (and, in some embodiments, other frames) (box 202). In some embodiments, method 180 includes outputting a user-perceptible indication (and in some embodiments, outputting feedback on why or how to improve the analysis) when none of the localization image time frames within the span have a sufficiently high (e.g., approximately a threshold) corresponding score to be considered the optimal TI frame (e.g., due to disease). (Box 204) In some embodiments, method 180 includes outputting a user-perceptible indication of the optimal TI based on the optimal TI frame and a pulse sequence map of the TI localization image sequence. (Box 206)

[0050] Figure 3 A flowchart illustrating method 208 for determining local transition regions is provided. One or more steps of method 208 may be performed by... Figure 1 The processing circuitry of the magnetic resonance imaging system 100 or a remote computing device is used to execute the operation.

[0051] Method 208 includes (e.g.) Figure 2 The ROI-cropped TI positioning image MIP image and the ROI-cropped MR positioning image time frame obtained as described in method 180 are normalized (box 210). Each data frame (i.e., the ROI-cropped MR positioning image time frame) is normalized between 0 and 1. Method 208 also includes calculating a metric (Jensen-Shannon distance) (e.g., distance) between each normalized ROI-cropped MR positioning image time frame and the normalized ROI-cropped TI positioning image MIP image (box 212). The Jenson-Shannon measure is the dissimilarity between two probability distributions (i.e., the corresponding normalized ROI-cropped MR positioning image time frame and the normalized ROI-cropped TI positioning image MIP image). The metric (distance, d) is used to calculate the distance.JS The equation is as follows:

[0052] (1)

[0053] Where P represents the probability distribution of the normalized ROI-cropped TI localization image MIP image, and Q represents the probability distribution of the corresponding normalized ROI-cropped MR localization image time frame.

[0054] Method 208 further includes generating a distance score map based on a computed metric (box 214). Method 208 further includes identifying the point (and corresponding frame) with the maximum distance (of all computed distances) within the distance score map (box 216). This point is a transition region. Method 208 also further includes selecting a frame (i.e., the span of the frame) for a single prediction as input to a second trained neural network (as described above). Figure 2 (as described in method 180) (Box 218).

[0055] Figure 4 A schematic diagram illustrating process 220 for determining the optimal reversal time (TI) is provided. Process 220 includes obtaining a series of MR positioning image timeframes 222 of the subject from MR positioning image imaging data acquired using a magnetic resonance (MR) scanner utilizing the TI positioning image sequence. Process 220 also includes obtaining TI positioning image maximum intensity projection (MIP) images 224 from the MR positioning image imaging data. Process 220 includes utilizing a series of MR positioning image timeframes 222 and TI positioning image MIP images 224 while performing a single transition point localization. Specifically, process 220 includes utilizing a first trained neural network 226 (e.g., a region of interest (ROI) detection model or network) to locate a cardiac region in the TI positioning image MIP image 224 to generate a region of interest (ROI) segmentation mask 228 for that cardiac region. In some embodiments, the first trained neural network 226 is configured to utilize an object detection method. In some embodiments, the first trained neural network 226 is configured to utilize a semantic segmentation method. Process 220 includes generating ROI-cropped TI positioning image MIP image 230 and ROI-cropped MR positioning image time frame 232 using ROI segmentation mask 228 on TI positioning image MIP image 224 and a series of MR positioning image time frames 222, respectively.

[0056] Process 220 further includes generating a distance score map 234 by calculating the distance between the MR localization image time frame 232 of each ROI cropped and the TI localization image MIP image 230 of the ROI cropped, as described above. Figure 3As described in method 208. As shown, the distance fraction map 234 includes an x-axis 236 representing the frame number of the MR positioning image time frame 232 cropped from the ROI and a y-axis 238 representing the calculated distance metric (e.g., Jensen-Shannon distance). Figure 240 shows the distance metric value for each frame in frame 232. Dashed lines 242 represent the determined transition regions (i.e., maximum distances). Dashed lines 242 represent optimal TI points. In some embodiments, the ROI-cropped TI positioning image MIP image 230 and the ROI-cropped MR positioning image time frame 232 are normalized before generating the distance fraction map 234. Process 220 further includes automatically analyzing the distance fraction map 234 to estimate the span (indicated by box 246 and referred to as the number of frames in the located transition frame batch 250) of the MR localization image time frame 222 (which corresponds to the MR localization image time frame 232 cropped for the optimal TI frame estimation), wherein the span of the MR localization image time frame 222 covers a transition region that includes phase changes in the cardiac region, including blood pool zeroing, distal myocardial zeroing, and initial recovery (box 196). The number of frames used for the located transition pool is any number located on either side of the transition point 242 (e.g., 10).

[0057] The span of the MR localization image timeframe 222 (or the localized transition frame batch 250) is within the expected time window size of the second trained neural network 248 (e.g., an optimal TI frame estimator model or network). The second trained neural network 248 is trained using the MR localization image timeframe 232 cropped from the ROI, with a time window size large enough to provide an accurate estimate of the optimal TI frame in the transition region (e.g., during the localized transition) that includes phase changes in the myocardium (e.g., blood pool zeroing, distal myocardial zeroing, and initial recovery) during the TI localization image sequence. The second trained neural network 248 is configured to determine or estimate the optimal TI frame. In some embodiments, the second trained neural network 248 is a dual-attention (Datt) three-dimensional (3D) CNN.

[0058] like Figure 4 As shown, process 220 includes inputting the span of MR positioning image time frames 222 (or the positioned transition frame batch 250) and ROI segmentation mask 228 into a second trained neural network 248. Process 220 includes using the second trained neural network 248 to determine an optimal TI frame from the span of the MR positioning image time frames (or the positioned transition frame batch 250), wherein the determination of the optimal TI frame occurs in a single traversal (or a single pass). As shown, for each MR positioning image time frame 222 in the positioned transition frame batch 250, a score between 0 and 1 (e.g., a predicted probability score) is output, as indicated by reference numeral 252.

[0059] Figure 5 The structure and function of the optimal TI frame estimator model, or network 248 (i.e., the second trained neural network), are illustrated. The structure of network 248 may differ from... Figure 5 The structure is shown. As shown, the optimal TI frame estimator model or network 248 is a dual-attention 3D network. The optimal TI frame estimator model or network 248 includes both a frame attention module (FAM) 255 and a spatial attention module (SAM) 257. The pruning used to train the optimal TI frame estimator model or network 248 is not performed explicitly, but rather by providing an ROI segmentation mask 228 as an additional channel to make it more sensitive to the first trained neural network (i.e., Figure 4 Any failures in the ROI detection model or network 226 are adaptively and implicitly handled. FAM 255 (i.e., the temporal attention module) enables better localization on the correct frames. The optimal TI frame estimator model or network 248 utilizes soft labels to incorporate any user bias in the annotation process. As shown, the localized transition frame batch 250 and the ROI segmentation mask 228 are fed into the optimal TI frame estimator model or network 248. A time window 10 is selected based on available data statistics. This time window is flexible while providing a good bound for permissible temporal localization errors. As shown, the optimal TI frame estimator model or network 248 outputs a score between 0 and 1 (e.g., a predicted probability score) for each MR localization image time frame in the localized transition frame batch 250, as indicated by reference numeral 259 and a graph 260 plotting the score for each frame in the localized transition frame batch 250. Figure 6 Depicting and Figure 5 Tables 262 and 264 relate to the optimal TI frame estimator model or network 248. Table 262 depicts the number of test and training samples used for the optimal TI frame estimator model or network 248. Table 262 also depicts the frame range of the test and training samples. Table 264 depicts the accuracy of the optimal TI frame estimator model or network 248 within ±1 frames and the number of parameters of the optimal TI frame estimator model or network 248.

[0060] Figure 7 Describes the use of Figure 5 The first example of the best TI frame estimator model or network 248. Figure 7 The right side depicts a series of MR localization image time frames 266 (e.g., short axis view). Image 268 represents a TI localization image MIP image. Figure 7The curve 270 in the upper left corner is a distance fraction plot, which includes an x-axis 272 representing the frame number of the MR localization image time frame for ROI cropping and a y-axis 274 representing the calculated distance metric (e.g., Jensen-Shannon distance). Figure 275 shows the distance metric value for each frame in the frame. Dashed line 276 indicates the determined transition region (i.e., maximum distance). Arrow 277 depicts frames with localized transition points. Dashed line 278 indicates the optimal TI point. Figure 7 Table 280 in the lower left corner compares the true value of the best TI frame with that of the [other frames]. Figure 5 The best TI frame estimator model or the best TI frame predicted by Network 248 is compared. As shown in the figure, the predicted best TI frame is the same as the true best TI frame. Table 280 also provides the prediction score of the predicted best TI frame (which is close to 1).

[0061] Figure 8 Describes the use of Figure 5 The second example of the best TI frame estimator model or network 248. Figure 8 The right side depicts a series of MR localization image time frames 282 (e.g., long axis view). Image 284 represents a TI localization image MIP image. Figure 8 The curve 286 in the upper left corner is a distance fraction plot, which includes an x-axis 288 representing the frame number of the MR localization image time frame for ROI cropping and a y-axis 290 representing the calculated distance metric (e.g., Jensen-Shannon distance). Figure 292 shows the distance metric value for each frame in the frame. Dashed line 294 indicates the determined transition region (i.e., maximum distance). Arrow 296 depicts frames with localized transition points. Dashed line 298 indicates the optimal TI point. Figure 8 Table 300 in the lower left corner compares the true value of the best TI frame with that of the [other frames]. Figure 5 The best TI frame estimator model or the best TI frame predicted by Network 248 is compared. As shown in the figure, the predicted best TI frame is within one frame of the true best TI frame. Table 300 also provides the prediction score of the predicted best TI frame (which is close to 1).

[0062] Figure 9 Describes the use of Figure 5 The third example of the best TI frame estimator model or network 248. Figure 9 The right side depicts a series of MR localization image time frames 302 (e.g., short axis view). Image 304 represents a TI localization image MIP image. Figure 9The curve 306 in the upper left corner is a distance fraction plot, which includes an x-axis 308 representing the frame number of the MR localization image time frame for ROI cropping and a y-axis 310 representing the calculated distance metric (e.g., Jensen-Shannon distance). Figure 312 shows the distance metric value for each frame in the frame. Dashed line 314 indicates the determined transition region (i.e., maximum distance). Arrow 316 depicts frames with localized transition points. Dashed line 318 indicates the optimal TI point. Figure 9 Table 320 in the lower left corner compares the true value of the best TI frame with that of the [other frames]. Figure 5 The best TI frame estimator model or the best TI frame predicted by Network 248 is compared. As shown in the figure, the predicted best TI frame is within two frames of the true best TI frame. Table 320 also provides the prediction score of the predicted best TI frame (which is close to 1).

[0063] Figure 10 An example is shown using [the technology] on a pediatric patient being scanned. Figure 5 An example of the best TI frame estimator model or network 248. Figure 10 The left side depicts time frame 322 of an MR localization image acquired from a pediatric patient. (As shown...) Figure 2 As described in method 180, a first trained neural network 226 is used to generate... Figure 2 The MR localization image time frame 324 of the ROI cropped on the right. Localization from a single transition point 326 (e.g.) Figure 2 Method 180 and Figure 3 As described in method 208), a distance fraction plot (i.e., graph 328) is generated. Graph 328 includes an x-axis 330 representing the frame number of the MR localization image time frame for ROI cropping and a y-axis 332 representing the calculated distance metric (e.g., Jensen-Shannon distance). Figure 334 shows the distance metric value for each frame in the graph. Dashed line 336 represents the determined transition region (i.e., maximum distance). Dashed line 338 represents the optimal TI point. As shown, frame 11 represents the transition point. The optimal TI frame estimator model or network 248 outputs the optimal TI frame 340 (i.e., frame 15). Figure 10 The paper describes how cardiac region localization is normalized without considering that the cardiac region is the size of the image matrix frame, and how robust TI estimates are produced.

[0064] Figure 11 Examples of using Figure 5 Examples of the best TI frame estimator models or networks 248 (e.g., where no best TI was found). Figure 11The left side depicts MR localization image timeframe 341, which is cropped from the ROI and input into the optimal TI frame estimator model or network 248. MR localization image data was acquired during free breathing TI localization (from which MR localization image timeframe 341 with ROI cropped was obtained). The TI window was not set correctly, and therefore no myocardial zeroing occurred. This is depicted in Table 342, which shows the signal intensity of the ROI placed in the myocardium without bounce points. Figure 11 A graph 344 depicts the output from the optimal TI frame estimator model or network 248. Graph 344 includes an x-axis 346 representing the frame number and a y-axis 348 representing the predicted probability score. Line 350 represents a set threshold (i.e., 0.5). Point 351 represents the predicted probability score of the frame. In graph 344, each frame has a predicted probability score below 0.5. This means the localization image is negative. In this case, a user-perceptible indication is provided that no optimal TI frame exists. Furthermore, reasonable methods and suggestions (e.g., adjusting the TI used) can be provided.

[0065] Figure 12 An example is given of using the disclosed technique to estimate the optimal TI frame when it is difficult to determine the optimal TI. Figure 12 The top of the image depicts time frame 352 of an MR localization image with ROI cropping, which is utilized in the disclosed technique. Figure 12 The curve 354 in the lower right corner is a distance fraction plot, which includes an x-axis 356 representing the frame number of the MR positioning image time frame for ROI cropping and a y-axis 358 representing the calculated distance metric (e.g., Jensen-Shannon distance). Figure 360 ​​shows the distance metric value for each frame in the frame. The dashed line 362 represents the determined transition region (i.e., the maximum distance). Arrow 364 depicts frames with the located transition point. The dashed line 366 represents the optimal TI point. In this case, it is difficult to determine the optimal TI. However, there is also a deviation from the optimal TI from the true values ​​provided by the agency. Even if the optimal TI marker cannot be determined, the disclosed technique is still useful in providing user guidance on the most suitable area.

[0066] The technical effects of the disclosed subject matter include providing systems and methods for determining the optimal reversal time. Technical effects include determining the optimal reversal time in a faster and more accurate manner. Technical effects of the disclosed subject matter include enabling rapid localization of the region of interest in the temporal dimension without raster scanning the entire MR localization image. Technical effects of the disclosed subject matter include introducing standardization into cardiac region localization, ensuring consistent performance regardless of changes in field of view, patient size, or view. Technical effects of the disclosed subject matter include providing a robotic automated TI solution that performs in a single pass (i.e., a single traversal) instead of relying on multiple traversals, thereby reducing computation time. Technical effects of the disclosed subject matter include providing reliable and consistent delayed-enhancement imaging of the myocardium regardless of variations in the TI localization image scheme. Technical effects of the disclosed subject matter include utilizing deep learning-based automated frame detection to help overcome any risks associated with alternative methods for determining the optimal TI frame, which consist of myocardial segmentation. Technical effects of the disclosed subject matter include providing the user with guidance on approximate temporal location for manual refinement, even when the optimal TI cannot be determined (e.g., due to disease), without having to scroll through all frames. The technical advantages of the disclosed subject matter include independence from the number of frames or temporal scaling, thus enabling adaptation to the subjectivity of TI localization imaging schemes across institutions. The technical advantages of the disclosed subject matter include requiring only a coarse region of interest around the myocardium (labeled via a deep learning-based model) without requiring fine segmentation of cardiac anatomy. The technical advantages of the disclosed subject matter include detecting whether myocardial zeroing has occurred in the localization imaging sequence. The technical advantages of the disclosed subject matter include solutions supporting multiple views (i.e., 4-chamber view, 3-chamber view, 2-chamber view, axial view, major axis view (4-chamber), and minor axis view).

[0067] Referring to the technology presented herein and protected by the claims, and applying it to physical objects and concrete examples of practical nature, which explicitly improves the present art, and is therefore not abstract, intangible, or purely theoretical. Furthermore, if any claim appended to the end of this specification contains one or more elements designated as “component for [performing]…the function” or “step for [performing]…the function,” such elements are intended to be interpreted according to 35 USC 112(f). However, for any claim containing elements designated in any other manner, such elements are not intended to be interpreted according to 35 USC 112(f).

[0068] This written description uses examples to disclose the subject matter of the invention, including best practices, and also enables those skilled in the art to practice the subject matter, including making and using any device or system and performing any included methods. The patent scope of this subject matter is defined by the claims and may include other examples that would occur to those skilled in the art. Such other examples are intended to fall within the scope of the claims if they have structural elements that are not indistinguishable from the literal language of the claims, or if they include equivalent structural elements that have minor differences from the literal language of the claims.

Claims

1. A computer-implemented method for determining the optimal reversal time (TI), the method comprising: A series of MR positioning image time frames (222) of the subject are obtained from MR positioning image imaging data acquired by a magnetic resonance (MR) scanner (102) using TI positioning image sequence via a processing system including one or more processors. The TI positioning image maximum intensity projection (MIP) image (224) is obtained from the MR positioning image imaging data via the processing system. The processing system utilizes a first trained neural network (226) to locate the heart region in the TI localization image MIP image (224) to generate a region of interest (ROI) segmentation mask (228) for the heart region. The processing system utilizes the ROI segmentation mask (228) on the TI positioning image MIP image (224) and the series of MR positioning image time frames (222) to generate ROI-cropped TI positioning image MIP image (230) and ROI-cropped MR positioning image time frames (232), respectively. The processing system generates a distance fraction map (234) by calculating the distance between the time frame (232) of the MR positioning image cropped for each ROI and the MIP image (230) of the TI positioning image cropped for the ROI. The processing system automatically analyzes the distance fraction map (234) to estimate the span of the MR localization image time frame for optimal TI frame estimation, wherein the span of the MR localization image time frame encompasses a transition region including phase changes in the cardiac region, such phase changes including blood pool zeroing, distal myocardial zeroing, and initial recovery; and The optimal TI frame is determined from the span of the MR positioning image time frame by means of a second trained neural network (248) via the processing system, wherein the determination of the optimal TI frame occurs in a single traversal.

2. The computer-implemented method of claim 1, the method further comprising: Before generating the distance fraction map (234), the TI positioning image MIP image (230) cropped from the ROI and the MR positioning image time frame (232) cropped from the ROI are normalized by the processing system.

3. The computer-implemented method of claim 2, wherein automatically analyzing the distance fraction map (234) to estimate the span of the MR positioning image time frame for optimal TI frame estimation comprises: Determine the point with the maximum distance on the distance fraction map (234), wherein the point represents the transition region; as well as From the series of MR positioning image time frames (222), select multiple time-continuous frames located on both sides of the transition region and including the transition region as the span of the MR positioning image time frame.

4. The computer-implemented method of claim 1, the method further comprising: The neural network is trained via the processing system to generate a second trained neural network (248) using the ROI segmentation mask (228) and a subset of the MR localization image time frames, the subset of the MR localization image time frames having a sufficiently large time window size to provide an accurate estimate of the optimal TI frame for the blood pool and / or myocardium.

5. The computer-implemented method of claim 4, wherein the span of the MR positioning image time frame is within the time window size.

6. The computer-implemented method of claim 1, wherein determining the optimal TI frame using the second trained neural network comprises: The span of the MR positioning image time frame and the ROI segmentation mask (228) are input into the second trained neural network (248) via the processing system; as well as The optimal TI frame for the span of the MR positioning image time frame is output from the second trained neural network (248) via the processing system.

7. The computer-implemented method of claim 1, wherein determining the optimal TI frame using the second trained neural network (248) comprises predicting a score between 0 and 1 for each MR positioning image time frame within the span, wherein the score of the optimal TI frame is close to or equal to 1.

8. The computer-implemented method of claim 7, the method further comprising: When none of the MR positioning image time frames within the span have a sufficiently high corresponding score to be considered as the optimal TI frame, the processing system outputs a user-perceptible indication.

9. The computer-implemented method according to claim 1, wherein the series of MR positioning image time frames (222) includes a 4-cavity view, a 3-cavity view, a 2-cavity view, an axial view, a major axis view, or a minor axis view.

10. The computer-implemented method of claim 1, the method further comprising outputting a user-perceptible indication of the optimal TI via the processing system based on a pulse sequence map of the optimal TI frame and the TI positioning image sequence.

11. A system for determining the optimal reversal time (TI), the system comprising: Memory (170) that encodes processor-executable routines; and A processing system comprising one or more processors and configured to access the memory (170) and execute processor-executable routines, wherein the processor-executable routines, when executed by the processing system, cause the processing system to: A series of MR positioning image time frames (222) of the subject were obtained from MR positioning image imaging data acquired by a magnetic resonance (MR) scanner (102) using TI positioning image sequence. The maximum intensity projection (MIP) image of the TI positioning image is obtained from the MR positioning image imaging data (224). The first trained neural network (226) is used to locate the heart region in the TI localization image MIP image (224) to generate a region of interest (ROI) segmentation mask (228) for the heart region. Using the ROI segmentation mask (228) on the TI positioning image MIP image (224) and the series of MR positioning image time frames (222), ROI-cropped TI positioning image MIP image (230) and ROI-cropped MR positioning image time frames (232) are generated respectively. A distance fraction map (234) is generated by calculating the distance between the time frame (232) of the MR localization image of each ROI cropped and the MIP image (230) of the TI localization image of the ROI cropped. Automatic analysis of the distance fraction map (234) estimates the span of the MR localization image time frame for optimal TI frame estimation, wherein the span of the MR localization image time frame encompasses a transition region including phase changes in the cardiac region, the phase changes including blood pool zeroing, distal myocardial zeroing, and initial recovery; and The optimal TI frame is determined from the span of the MR positioning image time frame using a second trained neural network (248), wherein the determination of the optimal TI frame occurs in a single traversal.

12. The system of claim 11, wherein the processor executable routine, when executed by the processing system, further causes the processing system to normalize the ROI-cropped TI positioning image MIP image (230) and the ROI-cropped MR positioning image time frame (232) before generating the distance fraction map (234).

13. The system of claim 12, wherein automatically analyzing the distance fraction map (234) to estimate the span of the MR positioning image time frame for optimal TI frame estimation comprises: Determine the point with the maximum distance on the distance fraction map (234), wherein the point represents the transition region; as well as From the series of MR positioning image time frames (222), select multiple time-continuous frames located on both sides of the transition region and including the transition region as the span of the MR positioning image time frame.

14. The system of claim 11, wherein determining the optimal TI frame using the second trained neural network comprises: The span of the MR positioning image time frame and the ROI segmentation mask (228) are input into the second trained neural network (248) via the processing system; as well as The optimal TI frame for the span of the MR positioning image time frame is output from the second trained neural network (248) via the processing system.

15. The system of claim 11, wherein determining the optimal TI frame using the second trained neural network (248) comprises predicting a score between 0 and 1 for each MR positioning image time frame within the span, wherein the score of the optimal TI frame is close to or equal to 1.

Images