Respiratory status alignment in MRI

Abstract

Apparatus, systems, methods and computer program products are disclosed for respiratory monitoring of a subject's breathing during a magnetic resonance imaging session. The method includes acquiring magnetic resonance imaging data including k-space values ​​along at least one trajectory segment in a k-space matrix, monitoring the subject's breathing by measuring one or more respiratory parameters, determining (103) a time instant at which a center of the k-space matrix and / or a closest point of the k-space trajectory segment to the center of the k-space matrix is ​​sampled in acquiring the magnetic resonance imaging data, and determining (104) the at least one respiratory parameter for the time instant using the monitoring of breathing.

Description

[Technical field]

[0001] The present invention relates to the field of respiratory motion monitoring and image processing in magnetic resonance imaging, more particularly to methods, devices, systems and computer program products for respiratory monitoring of a subject's breathing during a magnetic resonance imaging session and for controlling data acquisition and / or processing acquired data taking into account said respiratory monitoring. [Background technology]

[0002] In diagnostic imaging, the motion of the imaged patient is often a major concern. Moreover, motion can also be considered one of the most important challenges for diagnostic imaging and various treatments such as radiotherapy (RT). Different sources of motion can be considered, some more spontaneous than others. Respiratory motion is one type of motion that needs to be considered especially when scanning, for example, the torso, since it can easily result in motion artifacts that can significantly impair the image quality of the acquired images, for example magnetic resonance (MR) images.

[0003] In magnetic resonance imaging (MRI), such as MRI scans acquired of the thoracic or abdominal regions, it is important to suppress and / or account for respiratory induced motion. Fortunately, there are various approaches known in the art for suppressing this motion for MR images, such as breath-holding techniques and respiratory-gated or respiratory-triggered acquisition techniques.

[0004] With the breath-hold approach, many patients may have difficulty holding their breath long enough (e.g., often more than 20 seconds) to acquire sufficient data, and furthermore, breath holds are difficult to perform consistently and multiple times.

[0005] Gated and triggered acquisitions also have drawbacks: the configured gate time window can be set too small, which leads to excessive scan times, while a window set too wide can lead to residual motion artifacts. On the other hand, the quality achieved by triggering strongly depends on the correct detection of inhalation or exhalation and the trigger delay, which can also result in motion artifacts.

[0006] Thus, it is often not possible, or at least very difficult, for different imaging sequences in a single examination (e.g., over an extended period of time) to be acquired in exactly the same way, i.e., images are still acquired with slightly different respiratory states. For example, respiratory suppression techniques used in MRI result in images that show slightly different representations of the anatomy because they are not all acquired with exactly the same respiratory states, i.e., with different amplitudes and / or phases with respect to the respiratory cycle.

[0007] To reduce this discrepancy, image registration techniques can be used, for example, to project the images into a common reference frame. For example, this allows information from these MR images to be used effectively (and / or efficiently) in radiation therapy, for example to depict the target and / or + to be treated, to plan the overall treatment, and / or to evaluate the treatment results. For example, the images may generally be registered to each other, e.g., one scan is used as a master (reference image) to which the other scans are registered (translated, reoriented, distorted, transformed, and / or projected onto a reference). This registration step can be computationally very challenging, for example, due to different contrasts. In addition, MR images are often also registered to CT images, which adds to the difficulty, since the contrast between these two imaging modalities is significantly different, making it even more difficult to properly register the images. All of these factors can contribute to errors that propagate throughout the treatment, thus increasing the overall uncertainty and inaccuracy.

[0008] In the need to use medical imaging for planning or guidance, such as MRI for radiotherapy or high intensity focused ultrasound (HIFU), image registration adds uncertainties and / or inaccuracies that propagate throughout the entire required workflow, potentially reducing the overall required quality, as every registration step may result in more healthy tissue being irradiated than strictly required in radiotherapy. Furthermore, not only is it often necessary to have images with good image quality and limited (respiratory) motion artifacts, but respiration may also need to be characterized, so that this respiration pattern can be taken into account during planning, for example, to achieve a better conformal radiation plan. For example, in radiotherapy planning, a margin (e.g., defining the internal target volume) is typically added to the gross target volume to take into account the uncertainty induced by respiration. Thus, better knowledge of the respiration pattern can enable the radiotherapist to plan treatment within tighter margins.

[0009] Thus, accurately determining the respiratory state for each acquisition can reduce uncertainty and inaccuracy. This "respiratory state" can be described by a combination of amplitude and phase within a respiratory cycle. Respiratory cycles are generally not symmetrical, for example when comparing inhalation and exhalation, and can vary substantially across cycles, for example due to hysteresis within each respiratory cycle, changes in behavior, fatigue, and / or other influencing factors.

[0010] The respiratory state corresponding to each image (or raw image data) can be detected by continuously tracking relevant physiological and / or anatomical features, for example using a respiratory surrogate signal. A suitable signal can be determined by processing a camera observed video stream, by using a respiratory belt sensor, and / or by various other sensor techniques known in the art. It is also known to derive a (semi-)quantitative surrogate signal indicative of respiration directly from the output of the imaging system, for example the MR navigator signal, the noise navigator signal, or another surrogate derived from the acquired data itself.

[0011] It is known in the art that the registration step between MRI and CT can introduce systematic registration errors, for example in the range of 2-5 mm, which propagate throughout the radiation treatment plan and treatment. This provides a good impetus for relying on substantially MRI-only treatment planning, for example, where only MRI data is used to delineate relevant anatomical features and plan radiation treatment. However, registration between different acquisitions within a single MRI examination also introduces registration errors.

[0012] For MRI-guided treatments, such as MR-guided radiotherapy delivered by a combination of an MRI system and a radiotherapy accelerator (MR-Linac), it is particularly important to scan the patient in the same way as during the simulation. As commonly applied, by minimizing the positioning differences by various means, image registration can be performed relatively easily and accurately, e.g., identified contours can be quickly and accurately projected when generating a new radiotherapy plan. However, the respiratory state remains a constraining factor. Summary of the Invention [Problem to be solved by the invention]

[0013] It is an object of embodiments of the present invention to provide good and / or efficient (eg automated, eg accurate) means and methods for correlating current respiratory status with Magnetic Resonance (MR) imaging acquisitions. [Means for solving the problem]

[0014] An advantage of embodiments of the present invention is that image registration (e.g., registration of images within a single, e.g., longitudinal and / or multi-sequence MRI study) can be avoided or can be performed accurately and / or efficiently.

[0015] An advantage of embodiments of the present invention is that good accuracy, efficiency and / or effectiveness can be achieved in radiation therapy and / or for other image-guided treatments, such as HIFU.

[0016] An advantage of embodiments of the present invention is that MRI images can accurately take into account patient movement due to breathing.

[0017] It is an advantage of embodiments of the present invention that MRI imaging of the thoracic and / or abdominal region can be performed accurately with good image quality, e.g. such that image artifacts due to motion can be reduced.

[0018] An advantage of embodiments of the present invention is that better, e.g. short, scan times can be achieved in MRI imaging in contrast to some prior art methods, such as some prior art methods that rely on respiratory gating techniques.

[0019] An advantage of embodiments of the present invention is that MRI images (and / or MRI raw image data) may be acquired with good knowledge of the associated respiratory state of the imaged subject.

[0020] An advantage of embodiments of the present invention is that multiple MRI images (and / or raw MRI data) may be acquired, e.g., in different sequences in the same examination session, at substantially the same respiratory state (e.g., the same respiratory phase and / or amplitude, or abstractions thereof), e.g., precisely at the same point in different respiratory cycles.

[0021] An advantage of embodiments of the present invention is that image registration of different MRI images, e.g., acquired within the same study and / or in different MRI studies, and / or other imaging modalities, e.g., (one or more) additional CT images, may be accurately, efficiently, and / or more easily registered to each other (or to another reference image).

[0022] An advantage of embodiments of the present invention is that image registration errors and / or artifacts in the registered images can be reduced and therefore the propagation of such errors / artifacts to further images, plans, treatments and / or other types of derived information when derived from and / or based on the registered images can also be reduced.

[0023] An advantage of embodiments of the present invention is that image data can be acquired with good knowledge of the associated respiratory state, so that this side information can be taken into account in further use, e.g. when planning and / or assessing treatment.

[0024] An advantage of embodiments of the present invention is that the uncertainty margin taken into account when planning a treatment, e.g., a radiation treatment, can be reduced due to a lower error margin for the determined respiratory state related to the imaging data, and thus a better treatment can be achieved, e.g., potentially irradiating healthier tissue and / or achieving a better dose delivery at the treatment target.

[0025] It is an advantage of embodiments of the present invention that a signal indicative of a respiratory state can be determined and associated with corresponding acquired imaging data and / or images.

[0026] An advantage of embodiments of the present invention is that registration errors (eg related to spatial mismatch) may be reduced and / or eliminated, eg systematic and / or random errors may be reduced.

[0027] An advantage of the embodiments of the present invention is that high quality MRI-guided treatments, such as MRI-guided radiation treatments, can be achieved. For example, differences in respiratory status during pre-treatment imaging operations for simulation purposes and during treatment can be reduced. For example, by including accurate knowledge of the respiratory status for each acquired image (and / or the raw data on which it is based) for the same patient, it is more certain that differences between treatment and simulation are due only to underlying changes that should be taken into account when adapting the plan to take into account treatment response and / or other anatomical / physiological changes.

[0028] An advantage of embodiments of the present invention is that a motion model can be combined with a respiratory state signal (eg, a "surrogate signal") to accurately identify the respiratory state for each acquisition.

[0029] Advantageously, image acquisition can be automatically aligned to the (or each) respiratory phase of interest, e.g., when reconstructing k-space data (e.g., by adding the determined phase offset) and / or by taking this into account in image registration.

[0030] A method, computer program device, magnetic resonance imaging system, and magnetic resonance imaging workstation according to embodiments of the present invention achieve the above objectives.

[0031] In a first aspect, the present invention relates to a method for respiratory monitoring of a subject's breathing during a magnetic resonance imaging session. The method includes acquiring magnetic resonance imaging data comprising k-space values ​​along at least one trajectory segment in a k-space matrix. The method also includes monitoring the subject's breathing by measuring one or more respiratory parameters. The method includes determining, during acquisition of the magnetic resonance imaging data, a time instant at which a centre of the k-space matrix and / or a nearest point of the k-space trajectory segment is sampled relative to the centre of the k-space matrix, and using said monitoring of breathing to determine at least one respiratory parameter for said time instant.

[0032] Methods according to embodiments of the present invention can include providing feedback to the subject to guide the subject's breathing behavior towards a predefined breathing pattern. The predefined breathing pattern may be fixed, e.g., invariant in the context of a particular use case of the method, configurable by a user (e.g., selectable or definable in terms of parameters by a technician or medical professional), and / or determined by automated methods, e.g., by taking into account desired characteristics of the images to be acquired and / or the medical condition of the subject.

[0033] A method according to an embodiment of the invention may include using at least one monitored respiratory parameter to prospectively determine a time point at which respiration corresponds to a reference respiratory state, and controlling at least one acquisition parameter of the acquisition of the or each trajectory segment such that said time point determines the time of sampling of the centre of the k-space matrix and / or the closest point of the k-space matrix segment to the centre of the k-space matrix.

[0034] Methods according to embodiments of the invention may include determining the reference respiratory state by receiving reference information via a user interface and / or by retrieving the reference respiratory state from a data storage device.

[0035] In a method according to an embodiment of the present invention, at least one monitored respiratory parameter can be used to determine a number of time points at which respiration corresponds to a reference respiratory state for acquiring a number of trajectory segments respectively, such that each determined time point determines the time of sampling of the center of the k-space matrix and / or the closest point on each k-space trajectory segment relative to the center of the k-space matrix.

[0036] In a method according to an embodiment of the invention, controlling at least one acquisition parameter for each of the plurality of trajectory segments may include taking into account a predetermined time window for the collective acquisition of the plurality of trajectory segments and adaptively applying sparse sampling and / or compressed sensing techniques towards and / or past the predetermined time window. For example, sparse sampling may be combined with an iterative minimization approach for reconstruction or with another reconstruction technique suitable for sparsely sampled data acquisition.

[0037] A method according to an embodiment of the invention may comprise, after said acquisition, comparing the determined at least one respiratory parameter corresponding to the time of sampling a point at or closest to the center of the k-space matrix of the trajectory segment with a predefined reference value, determining a phase offset from this comparison, and applying said phase offset to the k-space values ​​obtained for the trajectory segment to retroactively compensate for translations in real space due to respiratory motion determined from said comparison.

[0038] A method according to an embodiment of the invention may include registering previously acquired images of the subject, having respiratory information associated with the previously acquired images, modeling a transformation acquired in the registration as a function of respiratory information to obtain a motion model, and determining the phase offset using the motion model.

[0039] A method according to an embodiment of the present invention may include reconstructing the acquired magnetic resonance imaging data and / or the magnetic resonance data corrected by the determined phase offset into at least one slice image, the slice image being annotated with a respiratory state determined by at least one respiratory parameter determined to correspond to the time point for sampling the center of the k-space and / or the predetermined reference value used to determine the phase offset.

[0040] In a method according to an embodiment of the present invention, reconstructing the at least one slice image may include selecting a subset of magnetic resonance imaging data to fill a k-space matrix with the magnetic resonance imaging data selected for reconstruction, where acquisition of k-space trajectory segments is selected based on a relationship of at least one respiratory parameter determined for each k-space trajectory segment to a reference respiratory state for reconstruction.

[0041] In a method according to an embodiment of the present invention, acquiring magnetic resonance imaging data includes repeatedly acquiring the same k-space trajectory segment at different points in the respiratory cycle, so that images can be (and / or are) reconstructed for different respiratory states by selecting from, interpolating between, and / or weighting different data sets acquired for said same k-space trajectory segment.

[0042] In a second aspect, the invention relates to a computer program product for carrying out the method according to the first aspect of the invention when the computer program product is executed on a computer.

[0043] In a third aspect, the invention relates to an apparatus for respiratory monitoring of a subject's breathing during a magnetic resonance imaging session. The apparatus comprises a sensor input for receiving sensor data and determining at least one respiratory parameter therefrom to continuously monitor the breathing of the subject during a magnetic resonance imaging session. The apparatus comprises a data input for receiving magnetic resonance imaging data comprising k-space values ​​along at least one trajectory segment in a k-space matrix, and a processor. The processor is adapted to determine a time instant at which the centre of the k-space matrix and / or the nearest point of each k-space trajectory segment to the centre of the k-space matrix is ​​sampled in the magnetic resonance imaging data received via the data input, and to determine the at least one respiratory parameter for said time instant using the monitored at least one respiratory parameter received via the sensor input.

[0044] An apparatus according to an embodiment of the present invention may comprise a controller output, and the processor may be adapted to use the at least one monitored respiratory parameter received via the sensor input to predictively determine the point in time at which respiration will correspond to a reference respiratory state, and to control via the controller output at least one acquisition parameter of the acquisition of the or each trajectory segment such that the point in time determines the time of sampling of the centre of the k-space matrix and / or the closest point of the k-space matrix to the centre of the k-space matrix.

[0045] In an apparatus according to an embodiment of the invention, a controller output may be adapted to provide a trigger signal to a magnetic resonance imaging system in which a magnetic resonance imaging session is performed to initiate acquisition of magnetic resonance imaging data of said k-space trajectory segment at a time determined such that the time of acquisition of the centre of the k-space data, or the nearest point on that trajectory segment, coincides with the time at which at least one respiratory parameter is predicted to correspond to said reference respiratory state.

[0046] In an apparatus according to an embodiment of the present invention, the processor may be adapted to use the monitored at least one respiratory parameter to determine a number of time points whose respiration corresponds to a reference respiratory state for acquiring a number of trajectory segments respectively, such that each determined time point determines the time of sampling of the centre of the k-space matrix and / or the nearest point on each k-space trajectory segment relative to the centre of the k-space matrix.

[0047] In an apparatus according to an embodiment of the present invention, a processor may be adapted to control at least one acquisition parameter for each of the plurality of trajectory segments taking into account a predetermined time window for the collective acquisition of said plurality of trajectory segments, and to adaptively apply sparse sampling (e.g. combined with a suitable reconstruction technique) towards and / or beyond the predetermined time window.

[0048] In an apparatus according to an embodiment of the invention, a processor may be adapted to retrospectively compare the determined at least one respiratory parameter corresponding to said time of sampling of the centre of the k-space matrix or the nearest point on the trajectory segment with a predetermined reference value, determine a phase offset from this comparison, and apply the phase offset to the k-space values ​​obtained for the trajectory segment to compensate for translations in real space due to respiratory motion determined from said comparison.

[0049] An apparatus according to an embodiment of the invention may comprise a data storage device, and the processor may be adapted to retrieve previously acquired images of the subject from the data storage device, in which respiratory information is associated with each of the previously acquired images. The processor may be further adapted to register the previously acquired images, model a transformation obtained by said registration as a function of respiratory information to obtain a motion model, and use said motion model to determine the phase offset to compensate for translations in real space due to respiratory motion. Alternatively, the apparatus may be adapted to receive such registration and / or such motion model directly from a data storage device, which may for example be stored by an external image processor.

[0050] An apparatus according to an embodiment of the invention may comprise a reconstructor for reconstructing the received magnetic resonance imaging data and / or the phase corrected magnetic resonance imaging data into at least one tomographic image.

[0051] In an apparatus according to an embodiment of the present invention, the reconstructor may be adapted to annotate the reconstructed slice image with a respiratory state determined by at least one respiratory parameter determined to correspond to the time point at which the center of k-space is sampled, which is used to determine a phase offset for adjusting the magnetic resonance imaging data used for reconstruction, and / or by the predetermined reference value.

[0052] In an apparatus according to an embodiment of the present invention, a reconstructor may be adapted to select a (suitable) subset of the magnetic resonance imaging data so as to fill a k-space matrix with the magnetic resonance imaging data selected for reconstruction, the k-space trajectory segments being selected based on a relationship of at least one respiratory parameter determined for each k-space trajectory segment to a reference respiratory state for reconstruction.

[0053] In an apparatus according to an embodiment of the invention, the processor may be adapted to control the acquisition of magnetic resonance imaging data by repeatedly acquiring the same k-space trajectory segment at different points of the respiratory cycle in an oversampling strategy, and the reconstructor may be adapted to reconstruct images for different respiratory states by corresponding selection from, interpolation between and / or weighting of data acquired for said same k-space trajectory segment.

[0054] Apparatus according to embodiments of the present invention may include a feedback output for providing sensory feedback to a subject during a magnetic resonance imaging session to guide the subject's respiratory behavior towards a predetermined breathing pattern.

[0055] Apparatus according to embodiments of the present invention may be configured such that sensor data is synchronized with magnetic resonance imaging data.

[0056] The apparatus according to the embodiment may have a processor configured to determine the closest point of each k-space trajectory segment to the center of the k-space matrix, at least in part using at least one acquisition parameter. The processor may have access to, for example, pulse sequence commands that define the timing of various gradient and radio frequency pulses, as well as the acquisition time of the magnetic resonance data. This may enable the closest point of each k-space trajectory segment to the center of the k-space matrix, at least in part using at least one acquisition parameter.

[0057] In a fourth aspect, the present invention relates to a magnetic resonance imaging workstation and / or a magnetic resonance imaging system comprising an apparatus according to an embodiment of the third aspect of the invention.

[0058] The independent and dependent claims set out particular and preferred features of the invention. Features from the dependent claims may be combined with features of the independent claims and with features of other dependent claims as appropriate and not necessarily as explicitly set out in the claims. [Brief description of the drawings]

[0059] [Figure 1] 1 illustrates a method according to an embodiment of the present invention. [Diagram 2] 1 shows an apparatus according to an embodiment of the present invention. [Diagram 3] 1 illustrates a magnetic resonance imaging system and a magnetic resonance imaging workstation according to an embodiment of the present invention; DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0060] The drawings are schematic and non-limiting. Elements in the drawings are not necessarily drawn to scale. The present invention is not necessarily limited to the particular embodiment(s) of the invention illustrated in the drawings.

[0061] Notwithstanding the exemplary embodiments described below, the present invention is limited only by the claims appended hereto, which are expressly incorporated into this detailed description, and each claim, and each combination of claims permitted by the substructure defined by the claims, forms a separate embodiment of the present invention.

[0062] The term "comprising", when used in the claims, is not limited to the features, elements, or steps listed below, but does not exclude additional features, elements, or steps. Thus, it specifies the presence of a stated feature without excluding the further presence or addition of one or more features.

[0063] In this detailed description, numerous specific details are presented. Embodiments of the present invention may be practiced without these specific details. Additionally, well-known features, elements, and / or steps are not necessarily described in detail for the sake of clarity and conciseness of the present disclosure.

[0064] In a first aspect the invention relates to a method for respiratory monitoring of a subject's respiration during a magnetic resonance imaging session, e.g. for annotating image data and / or for taking the subject's respiration into account during the procedure and / or during post-processing of acquired data.

[0065] In particular, the method may be a computer-implemented method, i.e., an automated or semi-automated method that may be performed by dedicated processing hardware (e.g., using application specific integrated circuits or circuits) and / or configured / programmed processing hardware, e.g., a computer or other general purpose processor programmed for the specific task of performing the method, and / or a configurable hardware platform (e.g., a Field Programmable Gate Array) configured for the specific task of performing the method. Combinations of application specific hardware (e.g., ASICs) and / or configured hardware (e.g., FPGAs) and / or one or more programmed processing devices (e.g., using a CPU, GPU or other suitable processor, possibly in combination with supporting hardware such as is typically found in a computer, e.g.).

[0066] "Automated" and "semi-automated" may refer to a computer-implemented method implemented by a (e.g., digital) processor, controller, and / or other such hardware, where the method is performed in an autonomous or supervised autonomous mode requiring only limited input from and / or interaction with an operator, e.g., to select or enter relevant parameters and / or configuration options, to start, stop, and / or interrupt a procedure, e.g., to oversee the procedure and / or other such limited interaction.

[0067] It will be understood that a magnetic resonance imaging (MRI) session refers to an examination using an MR imaging scanner, for example, for diagnostic and / or medical purposes, to diagnose a condition of a subject, to plan a treatment, to monitor and / or control a treatment being performed simultaneously, and / or for other related, e.g., medical, purposes. The examination may include, for example, magnetic field gradient and radio frequency pulse application and data acquisition for one or more MRI sequences to reconstruct one or more images, e.g., tomographic and / or 3D images, possibly with different contrast characteristics, and / or to show dynamic changes in the body, e.g., one or more 4D scans. An example of a use case in which a treatment is performed simultaneously is, for example, (MR) image-guided radiation therapy, using a combination of a magnetic resonance imaging system and a radiation therapy accelerator (MR-Linac). When a session is referred to, this may refer to a procedure in which a subject is placed in a magnetic resonance imaging system to perform an imaging procedure, after which multiple different sequences may be applied, for example, to acquire MRI data without removing the patient from the imaging environment (not necessarily excluding potentially repositioning the patient in a different posture or to better align the body). Although embodiments of the present invention may be advantageous to improve consistency and / or improve image quality when applied equally to multiple sessions of such a multi-session sequence, a case in which the subject is removed from the imaging system to return for a further session at a later time or date may generally be considered multiple sessions.

[0068] 1, an exemplary method 100 according to an embodiment of the present invention is shown. The method includes monitoring 102 the respiration of a subject by measuring (e.g., continuously and / or continuously, e.g., at least frequently), e.g., using a sensor, one or more respiratory parameters P(t). Thus, the respiratory parameters may be sampled continuously in a consistent manner.

[0069] In general, a sensor may refer to any suitable detection device, i.e. a device or system capable of determining relevant parameters that can characterize aspects of breathing from observed physical quantities. The breathing parameters may be parameters clearly directly related to breathing, such as air volume in the lungs, expiratory flow and / or breathing frequency, and / or parameters sufficiently related to and / or indicative of (one or more) characteristics of the breathing cycle (e.g. "surrogate" signals). A combination of different parameters may be used to characterize breathing more completely, for example characterizing phase values ​​within the breathing cycle and breathing amplitude values ​​(e.g. indicative of instantaneous chest expansion). Examples of suitable signals that can be considered as breathing parameters and / or from which such parameters can be derived include monitored predefined reference positions (or their changes) on the chest due to (one or more) breathing, velocity or acceleration measurements of such one or more reference positions, measurements of chest expansion, blood pressure measurements, and / or relevant information derived from camera observed images (e.g. video streams) by image processing techniques. This also means that the sensor or detection device can take many forms and can comprise, for example and without being limited to these examples, a spirometer, a proximity and / or ranging sensor, a breathing belt sensor, an accelerometer, a radar sensor or similar technology (lidar, sonar, etc.), a pressure sensor or an array of pressure sensors (e.g. arranged in a pressure sensitive mattress), and / or a camera or a combination of cameras (e.g. arranged for depth imaging). If the sensor or detection device comprises one or more cameras, this / these can comprise monochrome cameras, color imaging cameras, and / or multispectral cameras and / or can operate in the (human) visible spectrum, the infrared spectrum, a combination thereof, or a small sub-band thereof (e.g. only certain color components).The at least one respiratory parameter may be derived in a direct manner from data provided by a sensor or detection device (e.g., a sensor may provide an output that can be directly used as the respiratory parameter of the subject), or may require some processing, for example from data from a video stream, to determine a robust parameter (e.g., that correlates with or at least is indicative of the current amount of air in the lungs, the level of chest expansion, breathing frequency, etc.). However, many suitable methods for determining one or more respiratory parameters from a variety of data sources (including, for example, video camera observations) are known in the art and may thus be readily implemented in embodiments according to the present invention.

[0070] The method may therefore comprise the steps of acquiring sensor data from one or more sensors and / or one or more detection devices (which may also include, for example, a camera observation system from which relevant information can be derived) and determining one or more respiratory parameters from the acquired sensor data, for example by applying appropriate processing, and determining, for example, one or more parameters indicative of the current amount of air filling the lungs, such as respiratory rate, respiratory rate, distance between reference points where at least one point is subject to respiratory movement, and / or other such information indicative of breathing.

[0071] The MRI data may be acquired during free breathing or may be combined with other motion suppression techniques known in the art. For example, respiratory monitoring may be used to provide feedback to the patient (105) to guide the respiratory behavior towards a predefined respiratory pattern, as known in the art. Thus, feedback may be provided to the subject being imaged to influence the subject to breathe according to a predefined respiratory pattern. This may allow the acquired data to be better aligned with the predefined respiratory state(s) and improve the robustness and reproducibility of the imaged respiratory state(s). For example, the respiratory information may be shown or otherwise communicated to the patient, for example, in combination with information about a desired respiratory pattern, for example, by visualizing deviations therefrom or by providing other sensory cues to indicate such deviations. This may be applied during an imaging procedure, for example, simulation imaging for a therapy, and later repeated during therapy so that the breathing pattern corresponds, i.e., the collected imaging data forms an accurate representation of the condition during therapy.

[0072] Thus, the method may include a step 105 of providing feedback to the subject to influence the subject's breathing while monitoring the breathing to meet or approximate a predetermined or selected breathing goal. Thus, the subject may be actively guided toward a breathing pattern corresponding to the goal. For example, the measured breathing parameter(s) may be compared to a target defined by the target, and feedback may be provided to the subject representing the currently observed parameter(s) versus the target, the difference or other type of deviation measurement between the current value and the target value, or otherwise allowing the subject to gain insight in the alignment of the current breathing pattern with the predefined or selected target.

[0073] Such coaching can take various forms and shapes, potentially based on multi-modality inputs, e.g., different kinds of sensor inputs, and / or different measured or derived parameters representative of breathing behavior. A typical example is displaying a desired breathing pattern and providing insight or feedback on the accuracy of the actual breathing pattern. In the latter case, the method can check how close the actual behavior is to the desired target behavior and provide appropriate feedback.

[0074] As an example, WO 2020 / 069948 A1 discloses a suitable method for influencing a respiratory parameter of a subject's breathing pattern to achieve a goal. In this exemplary approach, the respiratory parameter can be decreased or increased (along with and toward the goal) by monitoring the current value of the respiratory parameter and providing the subject with sensory feedback about a determined preferred value of the respiratory parameter (different from the current value) for achieving the goal. For example, the preferred value of the respiratory parameter can be determined by an artificial intelligence intelligent agent, as described in the references. However, other approaches for coaching a breathing pattern toward a particular goal are also known in the art and are not necessarily excluded.

[0075] The method 100 includes, at 101, acquiring magnetic resonance imaging data including k-space values ​​along trajectories, e.g., lines, in k-space in one or more acquisitions, and reconstructing one or more images, e.g., by sampling and filling a k-space matrix or matrices, e.g., by filling a number of lines and / or trajectory segments through the matrix, until sufficient (or all) values ​​in the matrix have been sampled to allow MRI image reconstruction. As is known in the art, this may include applying a strong magnetic field around the body region to be imaged, e.g., by placing the patient in the magnet bore of a strong superconducting magnet, in which there is a strong magnetic field, e.g., of at least 0.1 T, e.g., at least 0.5 T, typically 1 T to 4 T, or even higher, e.g., on the order of 4 T to 10 T. This main magnetic field generally aligns, e.g., polarizes, the magnetic moments of susceptible atoms, e.g., hydrogen atoms, in the body.

[0076] Various magnetic field gradients can be applied to create differences in the magnetic field across the body, e.g., different planes, typically regions, in the body can be selected to be imaged, and / or spatial information in the signal response can be encoded by various possible encoding mechanisms, e.g., phase and / or frequency encoding. For example, typically, a radio frequency pulse as a series of pulses or other transmitted RE signal portions can be applied such that an oscillating magnetic field is temporally obtained at a suitable resonant frequency, e.g., by Larmor precession. For example, a gradient coil can adjust a selected region of the patient's body to the required magnetic field state so that it absorbs the energy of the pulse and resonates, i.e., a radio frequency (RF) signal is emitted in response, which can then be measured by a receiving coil. Response RE signal processing allows information about the material at its resonant origin to be deduced. By further varying the applied gradient field and applying and acquiring RE signals, information about local compositional and / or structural changes in the body can be reconstructed, e.g., one or more tomographic images of the body's internal structure, e.g., 2D, 3D and / or 4D images (i.e., a temporal sequence of 3D images).

[0077] By combining the time with information from step 102 of monitoring respiration with MRI acquisition 101, the respiratory state can be characterized at the time of data acquisition. Images can therefore be reconstructed from data acquired in the same respiratory phase. Thus, the need for a registration step can be reduced or eliminated, or made easier to perform (e.g., by reducing the initial difference between the images to be registered). By monitoring respiration using an appropriate respiratory (e.g., surrogate) signal, e.g., a signal that is preferably not automatically rescaled and that is captured with a sufficiently high spatial resolution (or that can be accurately interpolated or modeled based on less frequent sampling), an accurate characterization of respiration can be obtained that, in combination with accurate knowledge of when the MRI data is being acquired, allows the respiratory state of each acquisition to be defined.

[0078] The instant in time at which the center of k-space is sampled for the data from which an image is reconstructed can be used as a good representation of the time at which the image is acquired as a whole, such that the respiratory state associated with an image (e.g., 2D and / or 3D) can be determined by correlating the instant in time of central k-space sampling with a respiratory surrogate signal.

[0079] Generally speaking, in MRI, k-space refers to the (complex-valued) two- or three-dimensional Fourier space of the MR image (volume image or 2D image slice) being acquired. The k-space of the (planar or volumetric) image is sampled by applying a carefully controlled pulse sequence, e.g., a precisely timed sequence of radio frequency pulses and magnetic field gradients, to obtain a matrix of k-values ​​that form the temporal image space. Typically, the k-space matrix is ​​filled progressively (and thus spread out in time) along a sampling trajectory, e.g., row-by-row (or Cartesian), in a radial sampling pattern, in a zigzag pattern, in a spiral pattern, and / or possibly in less general patterns. Each line or trajectory segment sampled through the k-space matrix has a point passing as close as possible to the center of k-space where the lowest spatial frequency of that line or segment is encoded.

[0080] The time at which data for the center of the k-space matrix is ​​acquired is known precisely in advance (and a posteriori), e.g., with respect to the applied sequence. The temporal image of k-space values ​​can then be processed, e.g., reconstructed, to obtain the desired image (or a portion thereof). For example, the k-space matrix can have the same number of rows and columns as the final image to be reconstructed, and is filled line-by-line (or using another trajectory) with raw data during the scan, e.g., one row (not necessarily limited to) per repetition time TR.

[0081] The method 100 includes determining 103 a time tc at which the center of the k-space matrix (e.g., the origin) and / or the closest point (e.g., row) of a k-space line or trajectory segment to the center of k-space is sampled when acquiring 101 magnetic resonance imaging data, e.g., retrospectively (e.g., by adjusting acquisition parameters and / or triggers) for acquisitions that have already been performed and / or are to be performed.

[0082] It will be understood that the matrix does not necessarily include an odd number of rows and / or columns, or individual lines or an odd number of pixels, so that the center (and / or the closest passing point) may refer to the center or an interpolation point between the closest pixels (e.g., segmenting the average of the time points at which the four pixels of the central 2x2 block are sampled). Alternatively, one of the pixels closest / closest to the center (and equally close / close to each other) may be selected as a reference. Regardless of the matrix having an even or odd number of rows / columns, a small neighborhood of the center may be used to average (or otherwise aggregate) the sampling times to obtain a representative time point at which the center is sampled to increase robustness. The neighborhood may be symmetric, for example a block, a hard disk approximation, or another suitable form, but may also be asymmetric, for example considering multiple central pixels of a single central row. Preferably, when such averaging (or otherwise aggregate) is used, the number of pixels (k values) considered is rather small, for example less than 20 pixels, preferably less than 10 pixels, for example 9, 4 or 2. Also note that the time point for sampling the center of k-space (or the nearest path of the trajectory segment) may be determined directly from the sequence and may often (but need not) correspond to the center of the acquisition time window for a single acquisition.

[0083] The time to sample the center of k-space may be determined for each line in the k-space matrix, for example, corresponding to only the center of one k-space dimension (e.g., the center column of any row) or only the center line (e.g., the center column of the center row). In other words, the center of k-space may refer to the center in one, two, or three dimensions, as determined for each line, respectively, across all lines. The embodiments are not necessarily limited to row-by-row sampling, i.e., the k-space line may generally refer to any trajectory (or part thereof) through the k-space being sampled, and the determined time point 103 may correspond to the time at which the data acquisition sequence intersects or is closest to the center of k-space in at least one k-space dimension, e.g., the lowest spatial frequency component for that trajectory (segment) is captured.

[0084] At least one respiratory parameter P(tc) monitored 102 is also determined for this time tc, for example by predicting the respiratory parameter at the time of interest based on previous measurements of the parameter (i.e., when determining P(t) prospectively) and / or by looking up, interpolating and / or otherwise relying on observed values ​​of the respiratory parameter (when determined retrospectively) (104). Thus, a respiratory state representing each acquisition (each k-space line or trajectory segment, and / or each k-space matrix as a whole) is determined, which can be used, for example, in the reconstruction, to combine all acquisitions for the same respiratory phase and / or to annotate the reconstructed image with the corresponding respiratory state.

[0085] It will be appreciated that values ​​near the center of k-space correspond to low frequency (spatial) Fourier components, and therefore these values ​​have the strongest effect on the reconstructed image, e.g., affect large scale features of the reconstructed image. Thus, even though the data acquisition for an image is generally spread out in time, the time of acquisition of these low frequency components forms the best representation of the time of acquisition for the image as a whole, and similarly, the crossing or closest passage to the center of k-space for each acquisition sampling a single line or trajectory segment in k-space forms the best representation for the time of acquisition of that line segment. Thus, the respiratory state determined by the monitored respiration at this time is also a good approximation of the respiratory state associated with the particular trajectory segment as a whole, or similarly.

[0086] In a method according to an embodiment of the invention, the at least one respiratory parameter P(t) monitored 102 can be used to prospectively determine 106 the time tc at which the center of k-space and / or the point closest to the center of a line or trajectory (or trajectory segment) in k-space is sampled, e.g. to predict the time tc at which the respiratory parameter P(tc) is expected to be in a given state, e.g. P(tc)=PR, where PR represents the value of the respiratory parameter corresponding to that given state. This can be as simple as applying a fixed time offset (between the start of the acquisition and reaching the most central k-space sampling position for that acquisition) or can use a more elaborate model of the acquisition process and / or the time evolution of the respiratory parameter. The MRI acquisition can be triggered accordingly, such that the target respiratory state coincides with the time to sample the center of k-space, e.g. taking into account a predefined delay between the trigger and the time at which the center of k-space is sampled. Additionally or alternatively, for example, instead of or in addition to controlling the time at which acquisition begins (e.g., by generating a trigger), the characteristics of the applied magnetic field gradient may be altered to shift the time of acquisition of the center of k-space to the target, as determined from respiratory monitoring.

[0087] This approach allows for the acquisition of (most) central k-space data corresponding to the body in a given respiratory state. Optionally, actual measurements acquired after the acquisition can be used a posteriori to further more accurately define the respiratory state, e.g., respiratory parameters associated with the acquisition can be adjusted as necessary to the actual respiratory state at the time of the central k-space acquisition, e.g., predicted target values ​​can be replaced by values ​​based on actual measurements (102).

[0088] In other words, the monitored 102 at least one respiratory parameter may be used to determine 104 a time point at which respiration corresponds (at least approximately) to a reference respiratory state, and this time point may be used to determine and apply 108 acquisition parameters (e.g., time for triggering acquisition and / or other acquisition parameters) for acquiring MRI data 101, such that sampling the center of the k-space matrix and / or the nearest point on the k-space line or trajectory segment at the determined 103 time point to the k-space center according to the determined acquisition parameters corresponds to a time point at which respiration is in the reference respiratory state.

[0089] Thus, the time of acquisition of the center of k-space of the raw k-space image (and / or the most central point of each trajectory segment being sampled for the image) can be timed to the co-occurrence of a given state of respiration. The reference respiratory state may be determined as a phase of the respiratory cycle (e.g., a time point relative to a reference starting point of the cycle) and / or other parameters (e.g., a respiratory amplitude of clinical interest) or may be defined by a previously acquired image, e.g., in a different imaging session. For example, respiratory parameters (or an abstraction thereof) associated with the previously acquired image may be used as a reference for the new acquisition.

[0090] The method may include a step 107 of determining the reference respiratory state (e.g., a corresponding reference parameter value or value PR, or an abstraction thereof) by interacting with a user, e.g. receiving a reference respiratory state or information from a user via a user interface, e.g. a graphical user interface, and / or retrieving the reference respiratory state from a data storage device, e.g. as an annotation added to a previously acquired image and / or stored in association with a previously acquired image.

[0091] An example of the latter approach may concern the application of the method 100 in follow-up examinations, where the evolution of a medical condition with respect to a previous situation is of interest, for example to assess the response to treatment. This allows the new acquisition to be adjusted to acquire images that represent the same anatomical / physiological situation as the respiratory condition visualized by the previously acquired image. Another example may be a radiation therapy workflow, where a first MRI or CT image, appropriately annotated with respiratory information, may be used as a reference to register the respiratory state of a new MRI image to be acquired, so that new information is acquired for the same respiratory state, e.g., by adding MRI information to previously acquired CT and / or MRI data and / or by registering the MRI imaging with the planning master image during the treatment session (e.g., using an MR-Linac combination).

[0092] For example, respiratory information and associated image data may have already been acquired in the same imaging session or in a previous imaging session, and this information may be used to determine a reference respiratory state of interest, such as maximum expiration (e.g., typically most stable for planning treatment) or an intermediate position (e.g., a point at which the volume of interest lies, averaged over a respiratory cycle, that may be most representative for planning treatment). The target reference respiratory state of interest may also be determined directly, for example using a predictive model, and / or from a monitored respiratory surrogate signal, and / or simply by selection by an operator, e.g., via a user interface.

[0093] Thus, the acquisition of k-space MRI data may be prioritized for a particular point in the respiratory cycle, i.e., a reference state. Data may be acquired such that the center of k-space is always acquired at a desired respiratory phase, while data closer to the center of k-space is acquired at a time just before or just after the most optimal respiratory phase. Thus, peripheral k-space data may be acquired at different respiratory phases, for example, depending on the time required for line acquisition. However, these peripheral data may typically have a less severe impact on the reconstructed image quality, since they are associated with an increasing time difference to the target point in the respiratory cycle as the distance in k-space increases, while they are associated with an increasingly smaller spatial scale relative to the center of k-space.

[0094] For example, the method may be configured to take into account a target or maximum scan time and select appropriate points at times tc1, tc2, ... within this time window to optimize data acquisition and exhibit minimal artifacts, e.g., due to both motion and undersampling. In other words, for each found (i.e. predicted) appropriate time t at which a reference respiratory state will be achieved, one or more k-space lines (or, in general, trajectory segments) may be scheduled to coincide the acquisition of the center of the k-space data (or the closest path of the line) with the target time. Images may thus be acquired by sampling a portion of k-space during different respiratory cycles, each time at or near the same target point of the respiratory cycle.

[0095] If insufficient data is determined to be acquired near the end of the available time window, a compressed sensing algorithm may be applied to help reduce undersampling artifacts, for example using Poisson hard disk distribution sparse sampling. Thus, the method 100 includes, at 106, determining, for a plurality of acquisitions 101 that trace corresponding lines or trajectory segments through k-space, using at least one respiratory parameter, a plurality of times to pass through the center of k-space, or the times closest to it along that line or trajectory segment, so that said times correspond to a target reference respiratory state. This use 106 may include optimizing 108 (e.g., determining, adapting) the acquisition parameters to adequately sample the k-space matrix within a given time window, for example fitting as many of such acquisitions to the time window as possible to at least approximate the intended target reference respiratory state, potentially reducing the time required for acquisitions, for example, toward or past the end of a fixed / preferred time window, by sparse sampling.

[0096] Thus, the method can continuously optimize (108) (e.g., update and / or modify as necessary) sequence characteristics, for example, not necessarily to trigger but also potentially to apply other parameters such as gradients to determine the time delay for sampling the center of k-space (or the nearest pass of the trajectory segment) of the next sequence portion to apply relative to the start, and determine the time at which the next sequence portion should begin to align the central k-space sampling with the desired respiratory state and trigger the start of its acquisition accordingly.

[0097] A method according to an embodiment of the invention may comprise the steps of comparing (104) after (e.g. after each or after all) acquisitions 101 at least one determined respiratory parameter P(tc) corresponding to the center (or the point closest to the center) k-space of the acquisition with a predefined reference value PR, e.g. the desired respiratory state to be examined or the respiratory state in which previously acquired images of the same patient were acquired (see example above), and determining (109) from this comparison a phase offset φ(P(tc)−PR), which is for example proportional to the difference or other comparison measure (e.g. ratio) between the determined value and the reference, and adding (or subtracting) the phase offset to the k-space values ​​(e.g. of a line or trajectory segment) acquired in that acquisition. Thus, retrospectively, an additional phase may be added during (or at least before) the reconstruction to each acquired k-space line or set of lines (or to each trajectory, each subpart of a trajectory, the entire k-space matrix, ...). This step is not necessarily combined with prospective tuning, e.g., by triggering acquisition to a desired reference respiratory state, as described above, and the reference value PR for this correction step is not necessarily the same as the reference value PR when used for such prospective acquisition control. However, it will be appreciated that such a combination may nevertheless advantageously enable efficient correction of (e.g., typically small) deviations from the target, according to some embodiments.

[0098] Since different respiratory conditions cause shifts in the underlying anatomical structures, large components of the respiratory induced difference can be compensated for by a translation corresponding to the phase shift in k-space. The appropriate correspondence between the difference of the monitored respiratory parameter relative to its reference and the phase offset to be applied can be easily determined by experiment and / or simulation. This correspondence can take into account additional information, such as, for example, the orientation and / or position of the k-space lines in k-space, as well as potentially other parameters, such as patient characteristics (e.g., weight, volume, chest circumference, etc.), patient posture, type of examination, and / or other such parameters.

[0099] It is therefore an advantage that compensation for (e.g., residual) motion relative to a reference respiratory state can be easily performed in Fourier space. It should be noted that although this only compensates for translation, this approach can be applied to each line or set of lines acquired in k-space, and the dependency of the phase offset applied to the determined difference of respiratory state can be adjusted for each line or set separately (e.g., taking into account the orientation and / or position of the line). Thus, the result can be closer to a non-rigid transformation than a simple translation in real image space (not necessarily excluding a simpler embodiment in which the same phase / respiratory parameter difference relationship is used for all lines forming the raw image).

[0100] By way of example and not limitation, k-space may be sampled by multiple radial lines each passing through the center, and a phase offset for each line may be determined as a function of the radial angle and the observed difference between the respiratory parameter and its target. This allows for simple compensation of respiratory motion by taking into account the distribution of chest volume expansion / contraction during breathing, e.g., by taking into account at least small differences between the target state and the determined respiratory state.

[0101] For example, a phase shift determined from a detected deviation between the observed respiratory parameter(s) at the time of sampling the center of k-space relative to a predefined reference (e.g., target) / their respiratory parameter(s) can be applied to the data prior to performing a Fourier transform (or similar reconstruction algorithm) in order to correct the data to (or towards) the correct respiratory phase.

[0102] In an exemplary approach, the required phase offset, e.g., the relationship between the phase offset and the respiratory parameters, can be easily determined, e.g., by continuously (or continuously, e.g., with a sufficiently high time resolution) acquiring a single k-space line or set of lines over a single (or several) respiratory cycle to form a calibration. Any change in the signal observed over the respiratory cycle can be attributed to anatomical motion, at least as long as the gradients are configured the same while repeatedly acquiring this same line. Thus, the phase difference observed over time in the k-space data is the result of (in a first approximation) translation of the underlying anatomical structures. Such a calibration can be performed for a particular patient or a particular session, or can be performed on a test population to determine a relationship suitable for similar patients. This can be easily extended to incorporate variable parameters such as patient weight, chest circumference, sex, and / or other patient characteristics, and / or to take into account the position and / or angulation of the k-space lines and / or other sequence parameters, e.g., to determine an appropriate relationship that depends on such patient and / or sequence parameters.

[0103] Also note that, for example, if an oversampling and / or free-breathing strategy is used, in a first stage, points in the respiratory cycle for which sufficient k-space lines are available may be selected (e.g., sampling of the k-space center, or its nearest passage that is sufficiently close in time to the desired point in the respiratory curve). Images of these respiratory reference points may then be reconstructed and used to build a motion model (110), for example, by aligning the reconstructed images with each other (e.g., relative to any reference) and / or by modeling the resulting transformation (e.g., a deformable vector field) as a function of respiratory parameter(s), e.g., respiratory phase. This allows determining a phase offset (109) for any k-space trajectory, for example, to project one time point on the respiratory points to another arbitrary point using such a motion model. This empirical relationship may be used to adjust the k-space lines to any respiratory condition, so that in a second stage, phase-corrected data may be used to refine the first reconstruction and / or to reconstruct intermediate phases for which not enough data was initially available. In other words, calibration can be performed in-line without the need for an acquisition specifically aimed only at calibration, and so far enough data is sampled throughout the respiratory cycle to reconstruct a limited number of images that are well distributed throughout the respiratory cycle.

[0104] Thus, the method 100 may include, for example, determining a motion model from a registration 110 of a previously acquired image and its associated respiratory state information, included as metadata in the image, and determining a phase offset φ(P(tc)−PR) to apply to the k-space line or segment from the motion model. The previously acquired images may be images of the same subject acquired in the same or different imaging sessions, possibly with different modalities (e.g., CT). For example, multiple images (images of the same patient showing the same, or at least sufficiently overlapping, anatomical structures) may have been previously acquired, for example as a 4D scan, such as a time series of 3D images over a respiratory cycle. These images may be, for example, registered to each other using image registration techniques known in the art, such that different respiratory phases in the 4D set are registered to each other (e.g., selected as references) to define a transformation from one (respective) respiratory state to (e.g., any) other respiratory state. Such a transformation may be defined, for example, as a deformable vector field, or another suitable rigid or non-rigid transformation.

[0105] Since the respiratory state for the acquired data is precisely defined according to an embodiment of the invention, a transformation, e.g., a deformable vector field, determined from previously collected data can be used to convert from a known respiratory state to any other desired state, e.g., corresponding to one of the previously collected and registered images, or even by interpolation between those states. Even if the MRI data is not collected at exactly the same respiratory state as one of the previously collected images, interpolation techniques and / or motion models adapted to the data, e.g., based on the deformable vector field, can be used.

[0106] An advantage of the above-mentioned phase alignment technique 109, in contrast to conventional (real space) registration, is that the raw data can be constructed from phase aligned data from different respiratory phases, even if, for example, sufficient data for reconstruction for any of the respiratory phases is not available.

[0107] However, it will be appreciated that conventional (real space) registration may be used as well, additionally or alternatively, for example, using a motion model (e.g., or direct transformation) determined from the registration of previously acquired images. Thus, the MRI data is reconstructed (111) into an image having a respiratory state corresponding to a time sampling the center of k-space associated with the reconstructed image, and the reconstructed image is transformed (112) by a suitable transformation determined from the previously registered image, for example, using a parametric model, vector field interpolation, or any other suitable technique, to obtain a new image with any desired respiratory state different from that associated with the original reconstructed image. Since the previously acquired image data, for example, a 4D image set (e.g., a 4D synthetic CT image constructed from the MRI data, not limited to this example), may have been acquired under consistent (intra-sequence) conditions, for example, each 3D image (time point) having the same resolution, field of view, and / or substantially the same signal-to-noise ratio and contrast characteristics, the registration performed on these images may be robust and / or relatively easy. Information from this alignment may be utilized, for example, to transform newly acquired data in forming a motion model, even when acquisition conditions are substantially different, for example, using different contrast settings, differences in field of view selection, etc., on different imaging systems.

[0108] Since the currently acquired image data can be transformed into essentially any desired respiratory state after acquisition, the time required for imaging can be reduced and / or the selection of the appropriate respiratory state for imaging can be postponed until after the imaging procedure. For example, if in the inhalation phase, e.g. in radiation therapy planning, it is found that there is a good separation between the treatment target and the organs at risk, this respiratory state can be selected for use in planning and gating, even if the raw data does not correspond to imaging of this exact state. This provides substantial freedom to the clinician planning the procedure without increasing the burden of the imaging workflow. If the clinician decides in this example that the exhalation phase still provides better benefits for treatment, e.g. because it may be more stable to maintain and / or the patient may naturally dwell longer in this state, the option remains open to use that phase for planning and treatment by a simple selection of different transformation settings. Similarly, the same approach can be used to select an intermediate position (e.g. generating an image representation of it), i.e. the visualization of the time-averaged position of the treatment target, e.g. a tumor.

[0109] It is noted that while it is mentioned that the acquired data may first be reconstructed 111 into an image and then transformed 112 using previously collected registration information, other approaches (not necessarily mutually exclusive) described herein above may use, for example, the registration information 110 already available in the formation of the motion model to simulate appropriate phase offsets 109 for the different k-space lines (trajectories, trajectory segments) and transform each acquired k-space line (with its associated respiratory parameters or parameters indicative of the crossing of the center of k-space) to the respiratory state of the subject. Thus, while the data for reconstructing an image may be collected over a non-negligible time window, each separate k-space line may be easily transformed to bring the entire reconstruction to a desired respiratory state by determining the appropriate phase offset for that line. It is noted that this allows for a very efficient transformation process, since it only requires that a phase value be added (or subtracted) to each line in Fourier space. For example, for each line, the respiratory parameters associated with the crossing of the center of k-space (or the point closest to the center of that line) may be used to look up the current state in the motion model, while the desired output state defines a second target (reference) state in the motion model.

[0110] Alternatively or additionally (e.g., to compensate for residual motion), a registration technique as known in the art may also be applied to the reconstructed images using such a predefined motion model from a previous image sequence. The previously collected registration information may be acquired in a different imaging session, possibly even with a different imaging modality, or may relate to a different image acquired in the same imaging session, e.g., a 4D-MRI scan with sufficient time resolution. The previously collected image information (previous or same session) is preferably annotated with sufficiently accurate respiratory information, so that the motion model determined by the registration technique from this previously collected imaging data may, for example, be correlated with respiratory parameters, so that the transformation from any respiratory state to any other respiratory state can be easily determined.

[0111] Without being limited to this exemplary approach, the difference in the vector field between these two states can be easily calculated, and the average displacement across the k-space lines at the line can be easily determined. Using the known displacement to the phase relationship (e.g., determined from the acquisition parameters), the phase to be added for that line can be determined. Note that while this approach essentially only describes the displacement, this displacement can vary across the k-space lines, and thus can approximate the effect of deformable fields (e.g., non-rigid transformations). Furthermore, the acquisition sequence can be adjusted to take this simple registration process into account, for example, using (without being limited to) a radial k-space sampling pattern to better approximate the concentric displacement of the thoracic and / or abdominal anatomical structures due to breathing due to phase offsets of the radial lines. Other transformations in k-space can also be considered, such as, for example, interpolation to increase and / or decrease the overall bandwidth to better account for foreshortening (changes in scale across each line), to be determined for each line and applied to each line individually.

[0112] For example, when the respiratory state of each acquisition within an MRI examination is accurately determined, according to an embodiment of the present invention, the information collected for one scan (or a series of scans) can be exploited for other scans. For example, motion information from a 4D-MRI sequence can be used for other scans. Such a 4D-MRI sequence can be used to determine a deformation vector field (or another type of motion model) associated with the corresponding respiratory state information. Another image acquisition, for example an MRI image with different contrast characteristics, can be transformed according to this motion model to obtain an approximation of the 4D sequence (or, in general, an image of any desired point in the respiratory cycle different from that associated with the acquired image being transformed). Thus, a 4D respiratory cycle image series can be obtained for any desired contrast setting, for example, a 4D diffusion weighted image (DWI), a T2-weighted 4D scan, or even a 4D spectroscopic scan. On the other hand, it can also be used to speed up the acquisition of a particular 4D-MRI scan and to estimate information not captured in the fast 4D-MRI acquisition using information from other scans. A relatively fast sequence can be used to characterize the temporal dynamics of the respiratory cycle, and this information can then be used to obtain a full 4D representation of any desired respiratory state and / or other acquired images, which would potentially take substantially longer to acquire. It should also be noted that such time-intensive image acquisitions need not be collected at exactly the same points during the respiratory cycle, since the individual k-space lines (or trajectories, or trajectory segments) can be at least approximately aligned to a reference point by using the phase adjustments described above.

[0113] It is also advantageous that the above-mentioned phase adjustment technique can be used without the above-mentioned triggering approach, but can also be used in combination, e.g., such that this adaptive triggering (or adjustment of the sequence characteristics performed at the time of the center of k-space to the target point in the respiratory cycle) can predictively prevent large deviations of the acquired image from the anatomical structure at the intended respiratory phase to be visualized, and the phase adjustment can retroactively compensate for smaller deviations, e.g., further improving the quality of the reconstructed image. It is also noted that the latter is particularly easy to implement and / or requires little computational resources.

[0114] A method according to an embodiment of the invention may include a step 111 of reconstructing the acquired k-space data to obtain one or more (spatial; real-space) images, for example using an (inverse) Fourier transform, e.g. a fast Fourier transform, and / or another tomographic reconstruction technique known in the art. A respiratory state (e.g. a respiratory parameter or parameter P(tc), and / or an abstraction thereof) 104 is determined to correspond to a time instant tc of sampling the center of k-space (e.g. in 2D or 3D, e.g. the center of a k-space matrix) and may be associated with the reconstructed image, for example by annotating 113 the reconstructed image to indicate the corresponding respiratory phase. The time instant representing the reconstructed image may correspond to the center of the k-space sampling of the most central line (e.g. when row-wise sampling is used), or the average of the centers of the k-space sampling of all lines passing through the k-space matrix center (e.g. for radial sampling). Similarly, images acquired by reconstruction of phase-adjusted data 109 and / or by applying another registration technique 112 as described above may be annotated 113 with respiratory state information as well.

[0115] Advantageously, the respiratory state associated with an image can be accurately determined and added to the image, e.g. as side information incorporated in metadata, and this information can be taken into account in the analysis of the image, e.g. by a clinician and / or by an automated algorithm, and / or in further processing applied to the image. For example, the respiratory state may be characterized by at least one respiratory parameter related to the time point determined for the image, and may include, for example, a value(s) indicating the time point in the respiratory cycle (e.g., a phase value) and / or other parameters of the respiratory cycle, e.g., an amplitude value. The respiratory state may also be represented, for example, by a respiratory phase indicator, where the respiratory cycle is divided into a discrete number of phases, e.g., 4 or 10 phases (not limited to these numbers). Once the respiratory state or phase of the reconstructed image is determined, this information may be added as an annotation, e.g., in the header of a DICOM file that stores the image in a DICOM tag. In this way, a treatment planning system, a depiction program, other processing software, and / or a user can determine to which respiratory state an image corresponds, for example, to determine whether the image was acquired in a particular respiratory state / phase or the same respiratory state / phase and is therefore suitable for the intended processing. This information can also be used to determine in follow-up or treatment sessions, such as daily imaging on an MR-Linac system, which respiratory states / phases new images should be acquired to allow better comparison with previously acquired images.

[0116] It is also advantageous that the point at which each line (or trajectory segment, e.g. a row in a zigzag pattern, a radial line, a straight line segment, a spiral line...) passes through (or passes as close as possible to) the center of k-space can be determined so as to associate a specific respiratory state with each such line (or line segment). For example, MRI data may be acquired multiple times for the same k-space line (or trajectory segment), typically at different points of the respiratory cycle, e.g. with an oversampling approach. Thus, the step 111 of reconstructing the image can include a step 114 of selecting a set of k-space data such that the entire raw data (k-space) matrix (or at least a part thereof to be reconstructed) is filled by a combination of the selected data, the selected line being sufficiently close to the intended respiratory point (or phase corrected to that point).

[0117] In an embodiment according to the present invention, the (same) line (or trajectory, or trajectory segment) may be repeatedly acquired (115), for example to acquire oversampled data. Since the respiratory state corresponding to each sampled k-line can be accurately determined, and each sample may generally be obtained for a different position along the respiratory cycle, the reconstruction may determine the k-space segment to reconstruct for a given respiratory state by selecting suitable k-space samples to fill the segment (114), for example by interpolating between the closest respiratory states and / or by using a weighted average or other aggregation technique to take into account a distance measure between the desired respiratory state and the respiratory state determined for the (e.g.) line. Thus, retrospectively, any desired point in the respiratory cycle may be reconstructed using the oversampled data with respiratory state annotations. The same data for the same k-space matrix (2D or 3D) may be acquired multiple times during a single respiratory cycle. For example, a 2D image or a portion of 3D k-space that takes 200 ms may be sampled 25 times during a typical respiratory cycle of about 5 seconds. If the same portion of k-space is sampled during a single respiratory cycle and changes which region is acquired for the next respiratory cycle, any respiratory state can be reconstructed during reconstruction. When using oversampling techniques, more data may need to be collected, but the benefits may outweigh this aspect. For example, not only can any desired respiratory state be accurately reconstructed, but two or more lines close enough to any desired state can be sampled (e.g., by averaging, weighted averaging, etc.) for use in any state reconstruction, thus improving the signal-to-noise characteristics. This may be further improved by combining with the phase offset technique described above. The acquisition sequence may also be adjusted to take into account this oversampling and flexible reconstruction.For example, the entire k-space can be sparsely sampled in each respiratory cycle to create a fully sampled k-space matrix over multiple cycles, only one or a few k-space portions can be sampled in each respiratory cycle, and / or other variations can be considered to balance the total acquisition time against the desired level of oversampling at each or particular points in the respiratory cycle.

[0118] Typically, for example, in current radiation therapy practice, a given respiratory phase is selected to be imaged and used in treatment planning, and is often combined with the goal of aligning radiation treatment delivery to this respiratory phase. Alternatively, a relatively small number of images are acquired for different respiratory phases to provide more flexibility in planning.

[0119] However, although the central position is often, in a statistical sense, the most suitable for treatment, this is not often used in practice, for example relying instead on a single respiratory phase (e.g. selected from a 4D-CT sequence) as a reference for planning and treatment. Instead of a single respiratory phase, internal target volumes (ITVs) can also be commonly relied upon, which use time-integrated images (i.e. the sum, or, when properly normalized, the average of phase images in a 4D scan) to define the tumor volume to be treated. Obviously, the latter can also result in irradiation of a larger volume than strictly necessary, and thus potentially unnecessary damage to healthy tissue in the vicinity of the tumor.

[0120] For example, by providing a convenient approach to accurately characterize breathing for multiple images, possibly with a relatively large number of reconstructions and / or by allowing any point in the respiratory cycle to be reconstructed (e.g., by using phase correction, oversampling, and / or by taking into account motion models determined from previously acquired imaging data), an algorithm can be used to determine the most appropriate respiratory phase for treatment delivery. This can be based on traditional modeling, a combination of expert knowledge and algorithmic design and / or machine learning techniques. For example, since a large amount of image data across the respiratory cycle can be obtained for each patient / subject, a machine learning algorithm can be trained to select the best candidate respiratory phase for treatment, for example, based on supervised learning with the clinician's decision as the output to train, and / or by defining an objective function constructed to take into account relevant features such as separation between the treatment target and the organ at risk, dwell time of the target in a stable position, and / or other such potentially relevant factors. Thus, from the collected data, it can be determined which respiratory phase from 4D-CT, 4D-MRI, or 4D-PET (without being limited to these exemplary modalities) is most representative for the patient. It should also be noted that embodiments of the present invention can enable inherent aggregation of information from a wider time window and / or across multiple respiratory cycles to characterize a particular point (and multiple such points) in the respiratory cycle, for example by using oversampling, such that the acquired (and reconstructed) images can provide a better representation of the average condition in the body at that point in the respiratory cycle than conventional 4D-CT and / or 4D-MRI.

[0121] Although conventional respiratory motion can be characterized by such 4D-CT or 4D-MRI scans, the acquisition time may be insufficient to fully characterize the respiratory motion, e.g. providing only a small number of acquisitions and reconstructed phases. Nevertheless, the required acquisition time is often relatively long, e.g. on the order of 3 to 10 minutes, compared to other scans. The intermediate position defined by the time-averaged position of the tumor and / or organ at risk may be different when determined from a more detailed 4D imaging procedure (e.g. in terms of time resolution and / or acquired over a longer acquisition time window) than when determined using such conventional 4D-CT or 4D-MRI scans. According to an embodiment of the present invention, the respiratory motion can be characterized in detail, so that the intermediate position of the volume of interest in the body, e.g. the time-averaged position of the tumor and / or organ at risk over a respiratory cycle, can also be accurately determined. This may be achieved by prospectively acquiring MRI data at or near a time point representing the intermediate position (e.g., determined by continuously analyzing the respiratory signal) and then (optionally) registering the different intermediate positions with each other, e.g., acquired at different time points (e.g., at different respiratory cycles), potentially also using the phase adjustment approach discussed above. Note that data may also be acquired without adjusting the time point to the estimated intermediate point state, e.g., if sufficient sampling is achieved over a respiratory cycle (and potentially over multiple respiratory cycles), such that appropriate data may be retrospectively selected and / or appropriately weighted. Images representing the intermediate positions may also be determined, e.g., by weighting reconstructed images from different time points, taking into account the actual respiratory state and / or relative position of the target volume at that acquisition time with respect to other intermediate position scans (e.g., selected reference scans). Intermediate position information may be updated during MRI-guided treatment, e.g., using an MR-linac combination to obtain daily (or other frequency) updates of the intermediate positions.

[0122] In a second aspect, the invention relates to a computer program product for carrying out the method according to the first aspect of the invention when the computer program product is executed on a computer.

[0123] In a third aspect, the invention relates to an apparatus for respiratory monitoring of a subject's breathing during a magnetic resonance imaging session.Figure 2 shows an exemplary apparatus 10 according to an embodiment of the invention.

[0124] For example, the apparatus may comprise a processor, data storage memory, input, output, a user interface, and / or other means commonly known for carrying out the techniques as described above, e.g. programmed and / or configured accordingly. Thus, the apparatus may comprise a computer 20 and a computer program product according to the second aspect of the invention (i.e. adapted to be executed by the computer 20). Additionally or alternatively, the apparatus may comprise hardware specifically designed and / or configured to carry out the method according to the first embodiment of the invention, e.g. comprising an application specific integrated circuit and / or a field programmable gate array configured to carry out the method according to the first embodiment of the invention.

[0125] The apparatus 10 may be included in a magnetic resonance imaging workstation for processing, for example, MRI data, and / or in a magnetic resonance imaging system according to an embodiment of the present invention.

[0126] The device 10 includes an input 11 for receiving sensor data and determining at least one respiratory parameter therefrom, in order to continuously monitor the subject's respiration during an MRI session. The input may include one or more sensors (e.g., a spirometer, a camera, and / or another sensing device suitable for obtaining information capable of characterizing the subject's respiration) and / or is adapted to be operatively connected to such sensors. Thus, the device may also comprise a sensor input processor, for example integrated in the general purpose processor 14, in order to process the received sensor data and determine the respiratory parameter therefrom. Examples of suitable sensors and parameters derived therefrom, without being limited to these examples, have been discussed herein above.

[0127] The device 10 may also comprise a feedback output 12 for providing sensory feedback to the subject during the MRI session to guide the subject's breathing behavior towards a predefined breathing pattern. The sensory feedback may be delivered to the subject by visual, auditory, tactile and / or another sensory feedback mechanism, or may be a combination of different sensory modalities. The feedback output may thus comprise a monitor, a projector, an indicator light, a headphone, a speaker, an actuator, a vibrating element and / or another suitable device for delivering sensory feedback, or a combination of means for delivering various sensory modalities. Alternatively, the output may be adapted to operatively connect to such an output device. The device may thus also comprise a feedback processor, for example integrated in the general purpose processor 14, for determining a feedback signal suitable for providing via the output 12 based on determined breathing parameters determined from the received input 11 in relation to the predefined breathing pattern.

[0128] The apparatus 10 includes a data input 13 for receiving magnetic resonance imaging data including k-space values ​​along at least one trajectory segment in k-space, in one or more acquisitions, for example to reconstruct one or more images from the matrix, e.g., by filling a plurality of lines and / or trajectory segments through the matrix until sufficient (or all) values ​​in the matrix have been sampled to enable MRI image reconstruction.

[0129] The apparatus 10 comprises a processor 14 adapted to determine the time instant at which the center of the k-space matrix and / or the closest point of the (or each) k-space trajectory segment to the center of the k-space matrix is ​​sampled for the received magnetic resonance imaging data, e.g. for each trajectory segment in k-space. For example, the apparatus may comprise a computer, an application specific integrated circuit, and / or a configurable digital processing hardware for performing the processing tasks discussed below, e.g. the method steps of the method according to the first aspect of the invention discussed above. It will be understood that the schematic representation of the apparatus in FIG. 2 does not necessarily correspond to the physical arrangement of the components of the apparatus, e.g. the functions of the reconstructor 15 may also be performed by a processor. Similarly, the functions of the processor 14 as discussed herein may be distributed across several processing units, e.g. multi-core processors, graphical processing unit processors, cell processors, dedicatedly configured or specifically designed hardware devices (e.g. ASICs, FPGAs, etc.), a combination of one or more of the above commonly known components, or even across several computer systems, e.g. running on a computer cluster.

[0130] The processor 14 is further adapted to determine at least one respiratory parameter for said time point using at least one respiratory parameter that monitors respiration.

[0131] The processor 14 may be adapted to use the at least one monitored respiratory parameter to prospectively determine a time point when respiration corresponds to a reference respiratory state and to control, via the controller output 19, at least one acquisition parameter of the acquisition of the or each trajectory segment such that said time point determines the center of the k-space matrix and / or the time at which the closest point of the k-space trajectory segment is sampled to the center of the k-space matrix. For example, the controller output 19 may be adapted to provide a trigger signal to the magnetic resonance imaging system to start the acquisition of data of the k-space trajectory segment at an appropriate time determined such that the time at which the center of the k-space data or the closest point on that trajectory segment is acquired coincides with the reference respiratory state. Additionally or alternatively, other acquisition parameters may be controlled to affect the time at which the closest point to the center of the k-space is sampled, for example by adjusting the magnetic field gradients and / or the pulse configuration. As already mentioned, the device may be integrated into an MRI system (or its operator terminal), so that the output may simply refer to an internal state of the device, e.g. a data memory, accessible by a control system or program of the MRI system to control the MRI system accordingly.

[0132] The device may comprise a data storage device 16. The data storage may be integrated in the device or may refer to an interface for accessing an external data storage device, for example via a data communication network. For example, the data storage unit may comprise a data storage disk, a hard drive, a database, a network drive, or any other suitable means for storing digital information. The processor may be adapted to retrieve the reference respiratory state from the data storage device.

[0133] The device may also include a user interface 17, which may include, for example, a display monitor, a mouse, a keyboard, and / or one or more similar human interface devices known in the art. The user interface may be adapted to control the processes discussed above and below, for example, to configure the respiratory monitor, to set target respiratory states for feedback, to start and / or stop image acquisition, to initiate reconstruction, and / or to initiate other such operations. The processor may be adapted to receive reference respiratory states from a user via the user interface as targets for predictive triggers and / or sequence adjustments discussed above and / or as targets for retrospective phase adjustments and / or reconstructions discussed below. The user interface 17 may be adapted to interact with an operator, such as a medical professional, such as a doctor, nurse, radiation therapist, imaging technician, etc. For example, the user interface may be used by the operator to control the device and / or to observe breathing as monitored by the device.

[0134] The processor 14 may be adapted to use the monitored at least one respiratory parameter to determine a number of time points at which respiration corresponds to a reference respiratory state for acquiring a number of trajectory segments respectively, such that each determined time point determines the time of sampling of the centre of the k-space matrix and / or the closest point on each k-space trajectory segment relative to the centre of the k-space matrix.

[0135] The processor 14 may be adapted to control at least one acquisition parameter for each of the multiple trajectory segments taking into account a predetermined time window for the collective acquisition of the multiple trajectory segments, and to adaptively apply sparse sampling and / or compressed sensing techniques towards and / or past the predetermined time window.

[0136] In addition to, or as an alternative to, predictive control of acquisition taking into account respiratory monitoring, processor 14 may be adapted to retrospectively, i.e. after acquisition, compare the determined at least one respiratory parameter corresponding to said sampling time of the center of the trajectory segment or the nearest point on the trajectory segment with the center of the k-space matrix and a predetermined reference value, determine a phase offset from this comparison, and apply the phase offset to the k-space values ​​obtained for the trajectory segment to compensate for the translation in real space due to respiratory motion determined from the comparison.

[0137] The processor 14 may also be adapted to retrieve previously acquired images of the subject from the data store 16, where respiratory information is associated with each of the previously acquired images. The processor may be adapted to register the previously acquired images and the model translation obtained by the registration as a function of respiratory information to obtain a motion model. The processor may be adapted to use the motion model to determine a phase offset to compensate for translations in real space due to respiratory motion.

[0138] Alternatively, the processor 14 may be adapted to retrieve such registration data or motion models directly from a data store, for example stored by an external registration and / or motion modeling system.

[0139] A previously acquired image may, for example, relate to an image acquired in a previous imaging session, possibly using a different imaging modality, or may have been previously acquired in the same MRI imaging session and stored (e.g., after reconstruction) for the registration and / or motion modeling.

[0140] The device may comprise a reconstructor 15 for reconstructing the received magnetic resonance imaging data and / or the phase-corrected magnetic resonance imaging data into at least one tomographic image. The reconstructor may also be integrated in the processor 14, for example the processor may be adapted to reconstruct the image. The reconstructor 15 may further be adapted to annotate the reconstructed tomographic image with a respiratory state determined by at least one respiratory parameter determined by the processor to correspond to a point in time for sampling the center of k-space and / or a predetermined reference value used for determining a phase offset for adjusting the magnetic resonance imaging data used for the reconstruction.

[0141] The reconstructor 15 may also be adapted to select a (suitable) subset of the magnetic resonance imaging data (e.g., not empty and not including all acquired trajectory segments), such as filling a k-space matrix with selected magnetic resonance imaging data for reconstruction, where the k-space trajectory segments are selected for reconstruction based on a relationship of at least one respiratory parameter determined for each k-space trajectory segment with respect to a reference respiratory state, e.g., selecting the closest candidate, averaging multiple close candidates, using a weighted average taking into account a distance metric between the determined respiratory parameter and the reference, and / or using another similar strategy.

[0142] The processor 14 may also be adapted to control the acquisition of magnetic resonance imaging data by repeatedly acquiring the same k-space trajectory segment at different points in the respiratory cycle (e.g., by actively varying the sampled points in the respiratory cycle and / or by simply uncorrelated sampling with respect to the respiratory cycle). The reconstructor 15 may be adapted to reconstruct images for different respiratory states by corresponding selection, interpolation, and / or weighting of different data sets acquired for the same k-space trajectory segment (or rather, different trajectory segments, e.g., lines, may also be repeatedly sampled in the discussed manner such that sufficient data is available to reconstruct a complete 2D or 3D image for any point in the respiratory cycle). The above-mentioned phasing techniques may also be used to align the data of the k-space trajectory segment to a desired respiratory state for reconstruction. For example, the selection of one or more trajectory segment candidates may select candidates acquired at a time sufficiently close to the target respiratory state, and this selection may be phased to the desired state. Thus, if more candidates are selected and phase-adjusted for the same trajectory segment, the values ​​may be combined by averaging, weighted averaging (e.g., weighted by the inverse distance between the original respiratory parameter or parameters associated with the pre-phase-corrected data and one or more target respiratory parameters of the target respiratory state), and / or another suitable aggregation technique. This approach may thus be applied to different trajectory segments until the k-space matrix is ​​filled, or at least has enough data to allow reconstruction.

[0143] In a fourth aspect, the present invention relates to a magnetic resonance imaging workstation or a magnetic resonance imaging system comprising an apparatus according to an embodiment of the third aspect of the invention.

[0144] 3 illustrates diagrammatically a magnetic resonance imaging system 30 according to an embodiment of the present invention. The MRI system may comprise a primary magnet assembly 40 defining an examination zone 41, which may for example be formed by a volume in which the magnetic field conditions substantially generated and controlled by the magnet assembly are suitable for magnetic resonance imaging. The examination zone may thus correspond to (at least a usable portion of) the volume enclosed by the magnet bore of the system (for example, but not limited to, the principles of the present invention apply equally to open bore systems and other, less frequently used magnet assembly configurations).

[0145] A subject to be examined, e.g. a patient 43, may be positioned on a patient treatment couch 44 within the examination zone when the system is in use. The primary magnet assembly may comprise magnet windings, e.g. coaxial (e.g. superconducting) windings, for generating a fixed homogenous magnetic field within the examination zone. The examination zone may be a cylindrical volume enclosed by these magnet windings.

[0146] The system may include a reconstructor 15 for reconstructing magnetic resonance images, e.g., tomographic MRI images, from magnetic resonance signals acquired by the system in use. The reconstructed images may be provided via output 46 for viewing, processing, or storage. The output (directly or indirectly, e.g., via intermediate storage) may be received by a magnetic resonance imaging workstation 50, where medical personnel may interact with the received images, e.g., to view the images, manipulate the images, diagnose a patient's condition, and / or plan treatment. The workstation 50 may also optionally include the reconstructor 15.

[0147] In use, auxiliary equipment such as an auxiliary RF transmit and / or receive coil assembly 42 may be arranged in the examination zone to acquire magnetic resonance signals from specific parts of the subject's body. An auxiliary coil arrangement may be used to acquire signals from specific body parts or for specific use cases, while typically the signals may also be received (additionally or alternatively) by a receiver coil 47 integrated directly in the housing of the primary magnet assembly. The primary magnet assembly (or in general the magnetic resonance imaging system) also typically comprises gradient coils for generating (controllable) spatial gradients of the magnetic field in the examination region and / or for generating RF excitation pulses.

[0148] The system may comprise a respiratory sensor 52 for acquiring respiratory data indicative of the breathing of the subject 43. The respiratory sensor may comprise a respiratory belt sensor, a spirometer, an observation camera (with a video processing system suitable for deriving respiratory information from camera observations of the subject), and / or another type of sensor suitable for acquiring data from which relevant respiratory information may be obtained, for example as described above in connection with methods according to embodiments of the present invention.

[0149] An apparatus 10 according to an embodiment of the third aspect of the invention may be integrated into the magnetic resonance system 30, into the workstation 50 or elsewhere (or the functions of the apparatus described may be distributed between the MRI system and the workstation). The apparatus may be configured to receive its sensor input 11 from a respiratory sensor 52. It may also be configured to receive its data input from (one or more) receive coils 47 (or an intermediate signal processor or controller) of the MRI system.

[0150] The controller output 19 of the device may be connected to a system controller of an MRI system, for example, to control data acquisition of the system, for triggering acquisition and / or control other acquisition parameters such as gradients, pulses, and / or data acquisition settings.

[0151] The MRI system may also include a sensory feedback system 48 for providing sensory feedback to the subject, for example to guide the subject's breathing in a desired pattern. The feedback system may include visual, auditory, tactile, or other sensory outputs that may be perceived by the subject, such as, for example, a projector, a screen, headphones, speakers, vibrating elements, actuators, and / or other such means, or combinations thereof. Reference is also made to the above description of feedback systems in relation to methods according to embodiments of the present invention.

[0152] Other features or details of the above-mentioned features of the apparatus (e.g., computer program product, MRI system, and MRI workstation) according to embodiments of the invention will be apparent in light of the above description of the method according to embodiments of the invention, and / or vice versa.

Claims

1. 1. A magnetic resonance imaging system having an apparatus for respiratory monitoring of a subject's breathing during a magnetic resonance imaging session, comprising: a sensor input for receiving sensor data and determining at least one respiratory parameter therefrom to continuously monitor the subject's respiration during the magnetic resonance imaging session; a data input for receiving magnetic resonance imaging data having k-space values ​​along at least one trajectory segment in a k-space matrix; a controller output unit; Processor and Equipped with The processor is configured to execute a method for respiratory monitoring of a subject's breathing during the magnetic resonance imaging session, the method comprising: acquiring magnetic resonance imaging data having k-space values ​​along at least one trajectory segment in a k-space matrix; monitoring the subject's respiration by measuring one or more respiratory parameters; predictively determining a time point when the respiration corresponds to a reference respiratory state, the time point when the respiration corresponds to the reference respiratory state being determined by using the monitored at least one respiratory parameter to control at least one acquisition parameter of the acquisition of the or each trajectory segment such that the time point when the respiration corresponds to the reference respiratory state determines a time of sampling of a closest point of the k-space trajectory segment relative to a center of the k-space matrix; determining a time when the closest point of the k-space trajectory segment to the center of the k-space matrix was sampled during the step of acquiring the magnetic resonance imaging data; using said respiratory monitoring to determine at least one respiratory parameter for said time point; retrospectively comparing the determined at least one respiratory parameter corresponding to the time of sampling of the closest point on the trajectory segment relative to the center of the k-space matrix with a predetermined reference value corresponding to the reference respiratory state, determining a phase offset from this comparison, and applying the phase offset to the k-space values ​​obtained for the trajectory segment to compensate for translations in real space due to respiratory motion determined from the comparison; A magnetic resonance imaging system comprising:

2. 2. The magnetic resonance imaging system of claim 1, wherein the processor is adapted to use the monitored at least one respiratory parameter to determine a plurality of time points at which the respiration corresponds to the reference respiratory state for acquiring a plurality of trajectory segments, respectively, such that each determined time point determines a time of sampling of a closest point on each k-space trajectory segment relative to a center of the k-space matrix.

3. 3. The magnetic resonance imaging system of claim 2, wherein the processor is adapted to control at least one acquisition parameter for each of the plurality of trajectory segments taking into account a predetermined time window for the collective acquisition of the plurality of trajectory segments, and to adaptively apply sparse sampling and / or compressed sensing techniques towards and / or beyond the predetermined time window.

4. 2. The magnetic resonance imaging system of claim 1, further comprising a data storage device, wherein the processor is adapted to retrieve previously acquired images of the subject from the data storage device, respiratory information being associated with each of the previously acquired images, and the processor is further adapted to register the previously acquired images with a model translation obtained by registration as a function of the respiratory information to obtain a motion model, and to determine the phase offset using the motion model to compensate for translation in real space due to respiratory motion.

5. 10. The magnetic resonance imaging system of claim 1, further comprising a reconstructor for reconstructing the phase-corrected magnetic resonance imaging data into at least one tomographic image.

6. 6. The magnetic resonance imaging system of claim 5, wherein the reconstructor is adapted to annotate the reconstructed slice images with a respiratory state determined by the at least one respiratory parameter determined to correspond to a time point sampling the closest point on each k-space trajectory segment relative to a center of the k-space matrix by the predetermined reference value used to determine the phase offset to adjust the magnetic resonance imaging data used for the reconstruction.

7. 6. The magnetic resonance imaging system of claim 5, wherein the reconstructor is adapted to select an appropriate subset of the magnetic resonance imaging data to fill the k-space matrix with the selected magnetic resonance imaging data for reconstruction, and the k-space trajectory segments are selected for reconstruction based on a relationship of the at least one respiratory parameter determined for each k-space trajectory segment to a reference respiratory state.

8. 6. The magnetic resonance imaging system of claim 5, wherein the processor is adapted to control the acquisition of the magnetic resonance imaging data by repeatedly acquiring the same k-space trajectory segment at different points in the respiratory cycle in an oversampling strategy, and the reconstructor is adapted to reconstruct images for different respiratory states by corresponding selection from, interpolation between, and / or weighting of data acquired for the same k-space trajectory segment.

9. 10. The magnetic resonance imaging system of claim 1, further comprising a feedback output for providing sensory feedback to the subject during the magnetic resonance imaging session to guide the subject's breathing toward a predetermined breathing pattern.

10. The magnetic resonance imaging system of claim 1 , wherein the sensor data is synchronized to the magnetic resonance imaging data.

11. 2. The magnetic resonance imaging system of claim 1, wherein the processor is configured to determine, at least in part, the closest time point of the or each k-space trajectory segment relative to a center of the k-space matrix using the at least one acquisition parameter.

12. A method for respiratory monitoring of a subject's respiration during a magnetic resonance imaging session, comprising: acquiring magnetic resonance imaging data having k-space values ​​along at least one trajectory segment in a k-space matrix; monitoring the subject's respiration by measuring one or more respiratory parameters; prospectively determining a time point when said respiration corresponds to a reference respiratory state, said time point corresponding to said reference respiratory state being determined by using said monitored at least one respiratory parameter to determine a time of sampling of the closest point of said k-space trajectory segment relative to a center of said k-space matrix; determining a time instant at which the closest point of a k-space trajectory segment to a center of the k-space matrix was sampled during the step of acquiring the magnetic resonance imaging data; using said respiratory monitoring to determine at least one respiratory parameter for said time point; retrospectively comparing the determined at least one respiratory parameter corresponding to the time of sampling of the closest point on the trajectory segment relative to the center of the k-space matrix with a predetermined reference value corresponding to the reference respiratory state, determining a phase offset from the comparison, and applying the phase offset to the k-space values ​​obtained for the trajectory segment to compensate for translations in real space due to respiratory motion determined from the comparison; A method comprising:

13. A computer program product having instructions for causing a magnetic resonance imaging system equipped with a respiratory monitor to perform the method steps described in claim 12.

Images