Method for reconstructing a time series of magnetic resonance data sets for water-fat separation based on the Dixon method

Abstract

The invention relates to a computer-implemented method for reconstructing a time series of magnetic resonance data sets for water-fat separation based on the Dixon method (14), wherein a magnetic resonance data set (14) comprises complex image data acquired at at least two different echo times. The method comprises (a) determining (32) one or more phase maps (10) of the phase errors of the complex image data caused by B0 field inhomogeneities for the two or more magnetic resonance data sets (14) acquired at different times, wherein the determination comprises a phase unfolding (12) which is carried out taking into account all magnetic resonance data sets (14) of the time series; and (b) calculating fat and / or water images from the magnetic resonance data sets (14) corrected by the phase map(s) for the time series.The invention also relates to a computer program and a computer (5), in particular a control computer of an MRI device (5).

Description

[0001] Regardless of the grammatical gender of a particular term, persons with male, female or other gender identities are included.

[0002] The invention relates to a computer-implemented method for reconstructing a time series of two or more magnetic resonance imaging (MRI) datasets acquired at different times during an examination of a subject for water-fat separation based on the Dixon method. The invention also relates to a computer program and a computer, in particular a control computer of a magnetic resonance imaging (MRI) device.

[0003] Magnetic resonance (MRI) images of the human or animal body typically contain signals from both water and fat. To differentiate these tissue types in the image, one can exploit the fact that water and fat have slightly different resonant frequencies, i.e., a chemical shift difference of approximately 3.4 ppm. This can be used, for example, to suppress the fat signal with a frequency-selective pulse. However, this is difficult, especially at lower field strengths, because the difference in chemical shift becomes too small in absolute terms. Another way to separate the fat signal from the water signal is the Dixon technique. This involves acquiring images in which the magnetic resonance signal is recorded at different echo times. The different echo times result in a phase difference between the fat and water signals.In an ideal example, the echo times could be chosen such that the grease and water signals have the same phase at the first echo time, and a phase difference of π at the second echo time. In this case, the images can then be combined in a post-processing step to generate a grease image and a water image.

[0004] However, phase ambiguities must be removed and the actual phase profile reconstructed so that fat and water can be correctly separated. In particular, the problem is that even small inhomogeneities in the B0 field lead to such large phase shifts that the phases of the fat and water signals vary across the image, even beyond the maximum resolvable limit of +π and -π. This results in phase ambiguities because every time the phase exceeds +π or -π, it folds, jumping from +π to -π and vice versa. This is also called "phase wrapping." Therefore, Dixon imaging employs various phase unwrapping techniques to remove these phase ambiguities.Corresponding methods are described, for example, in the article by Jingfei Ma, "Dixon Techniques for Water and Fat Imaging," Journal of Magnetic Resonance Imaging 28: 543-558 (2008), Holger et al., "Dual-echo Dixon Imaging with Flexible Choice of Echo Times," Magn. Reson. Med. 65(1):96-107 (2001), and also in the article by Johann Berglund et al., "Two-point Dixon Method with Flexible Echo Times," Magnetic Resonance in Medicine 65: 994-1004 (2011). These techniques make it possible to estimate the correct phase profile within an image and thus perform phase correction of the Dixon images.

[0005] The Dixon technique is also frequently used in time-resolved MRI scans. In this technique, MRI images are acquired, for example, before, during, and after contrast agent administration to observe the contrast agent's inflow into the tissue. Since malignant tumors, for instance, accumulate contrast agents particularly quickly, this can help to identify and localize them as precisely as possible. A fat and / or water image is then generated for each MRI scan, and the respective fat and / or water images from the time series are subtracted from each other. The contrast agent inflow effect is particularly evident in these difference images.

[0006] In the current state of the art, phase correction is performed individually for each time point. This is also sensible because the B0 field inhomogeneities, which trigger the phase errors, can differ between individual time points, caused by patient movement.

[0007] Despite sophisticated phase-development techniques, artifacts occasionally occur in MR images acquired using the Dixon technique, and especially in dynamic MR images. The invention therefore aims to further improve the Dixon technique.

[0008] In particular, artifacts caused by B0 field inhomogeneities should be reduced even further.

[0009] The invention solves this problem by means of a method according to claim 1, a computer program according to claim 8 and a computer according to claim 9. Advantageous embodiments are specified in the dependent claims.

[0010] According to a first aspect, the invention provides a computer-implemented method for reconstructing a time series of two or more magnetic resonance data sets recorded at different times during an examination of a subject for water-fat separation based on the Dixon method, wherein a magnetic resonance data set comprises complex image data, and wherein the method comprises the following steps: (a) Determining one or more phase maps of the phase errors caused by B0-field inhomogeneities of the complex image data for the two or more magnetic resonance datasets acquired at different times, wherein the determination includes a phase unfolding taking into account all magnetic resonance datasets of the time series; (b) Calculating fat and / or water images from the time series magnetic resonance datasets corrected by the phase map(s).

[0011] The Dixon methods are based on the assumption that the B0 field drifts slightly across the field of view. Therefore, the magnetic resonance datasets, which contain complex image data, are reconstructed using a model that includes the water signal, the fat signal, and a background phase, also referred to here as phase error. In the prior art, the background phase is determined independently for each time point in a time series, which can lead to artifacts at individual time points. According to the invention, the background phase is now determined correlated for all time points, thereby avoiding artifacts.

[0012] The invention thus solves the described problem by determining the phase maps of the phase errors in the image data of a time series of magnetic resonance (MRI) datasets caused by B0 field inhomogeneities using a phase deconstruction that takes into account all MRI datasets in the time series. In other words, unlike in the prior art, a separate phase map is not determined for each MRI dataset in the time series, independent of the phase maps of the other MRI datasets in the time series. Instead, the phase information from the different time points is processed in combination. The invention recognizes that artifacts in the subtraction images described above can be caused by slight fluctuations in the B0 field between individual time points, for example, due to movement, pulsation, or other physiological reasons.This can then lead to different results in the phase unfolding in successive images, and this can lead to artifacts, especially in subtraction images of a measurement series.

[0013] Phase propagation is often performed under the assumption that the spatial variation of the phase information is weak. This assumption has also been shown to be helpful for the temporal dimension, leading to a reduction in artifacts. In particular, it is important that the phase errors of the magnetic resonance datasets acquired at different times are corrected jointly. This has been shown to result in fewer artifacts within a time series. This becomes especially clear when comparing the fat and / or water images of the time series, particularly when subtracting them to generate subtraction images.

[0014] The subject is preferably a human or animal, in particular a patient. The MR datasets can be acquired from any part of the body, for example the skull, abdomen, heart, lungs, liver, or limbs.

[0015] A magnetic resonance imaging (MRI) dataset for water-fat separation based on the Dixon method refers specifically to a dataset acquired during an MRI scan. This MRI dataset allows, in particular, the separation of the water signal and the fat signal due to their different phases. Such an MRI dataset can contain data acquired at several different echo times. A magnetic resonance imaging (MRI) dataset for water-fat separation based on the Dixon method is also referred to as a "magnetic resonance imaging (MRI) dataset" or "Dixon dataset" in the following.

[0016] A time series, as used here, is understood to be a series of two or more magnetic resonance imaging (MRI) datasets acquired at different times using the Dixon method. The time series can, for example, contain 2–50, preferably 10–30, images of the subject. These are preferably acquired consecutively, i.e., during the same examination of the subject. The MRI datasets are preferably image datasets, hereinafter also referred to as images. Preferably, all MRI datasets are acquired with the same parameters, in particular with the same field of view and the same acquisition parameters. However, a time series can also refer to an arrangement of different image contrasts, such as different inversion times.In this case, at least some magnetic resonance datasets of the time series are acquired with different acquisition parameters, for example, with different inversion times. This makes it possible to acquire image datasets using a combination of T1 mapping and Dixon, where the Dixon method is used separately for different inversion times.

[0017] According to one embodiment, a contrast agent, particularly a gadolinium-based contrast agent, is injected into the subject at the beginning of the time series. This is also referred to as DCE imaging ("Dynamic Contrast Enhanced"). In particular, this allows for the examination of blood flow. DCE is based on the T1-shortening effect of gadolinium-based contrast agents. For example, an intravenous bolus of the contrast agent is injected, followed by the acquisition of a rapid time series of T1-weighted MRI images. Regionally increased signal indicates a concentration of the contrast agent, which in turn suggests, for example, increased blood flow or tissue permeability into the extravascular space.

[0018] Within the time series, the individual magnetic resonance (MRI) datasets are preferably acquired with a time resolution of 2–20 seconds, preferably 5–10 seconds per image. Different sequence types can be used, in particular gradient echo sequences. For example, a spoiled gradient echo sequence of type "VI BE" can be used. Preferably, each MRI dataset contains image data acquired at at least two different echo times. Preferably, these were acquired at exactly two different echo times (2-point Dixon technique), but it is also possible to measure at three different echo times (3-point Dixon technique). Furthermore, the method according to the invention can also be carried out with MRI datasets acquired at only one echo time, i.e., with the so-called "1-point Dixon technique".

[0019] The method according to the invention comprises the determination of one or more phase maps of phase errors caused by B0 field inhomogeneities in the complex image data for two or more magnetic resonance datasets acquired at different time points for water-fat separation based on the Dixon method. The determination includes phase unfolding, which is performed taking into account all magnetic resonance datasets in the time series. The determination of the phase map(s) involves determining the phase errors of the complex image data caused by B0 field inhomogeneities. Because the determination is performed taking into account all magnetic resonance datasets in the time series, fluctuations in the B0 field between the individual time points are also adequately considered, particularly during phase unfolding. According to the invention, the phase information of the respective time points is thus processed in combination.This allows B0 field changes to be balanced between individual points in time.

[0020] The determination of phase maps can be carried out using known methods, such as those described in the aforementioned articles by Ma et al., Eggers et al., or Berglund et al. Ma et al. describe various possible procedures. In one example, a phase correction algorithm is used that determines the phase error in the form of a phase vector. This involves a region-growing method. These and other methods are based on minimizing a cost function. For example, minimum-norm methods or path-following methods can be used for phase development.

[0021] The determination of phase maps includes, in particular, phase unwrapping. This refers specifically to handling the phase ambiguities that arise because the phase of the complex image data can have a maximum value from +π to -π.

[0022] After determining one or more phase maps, taking into account all magnetic resonance (MRI) datasets of the time series, fat and / or water images are calculated from the MRI datasets corrected by the phase maps. In particular, water images can be calculated, which then advantageously show the infusion of any contrast agent. Specifically, a water image and / or fat image is calculated for each complex dataset of the time series.

[0023] According to one embodiment, the one or more phase maps for the various time points are determined under the assumption that the phase errors change only slightly over time, with the phase maps being averaged or smoothed over the time series. This assumption has the advantage of avoiding sudden jumps in the phase profile over time. It has been shown that this reduces the susceptibility to artifacts, particularly in subtraction images. The assumption is based, among other things, on the fact that the patient does not move, or moves only slightly, between the individual acquisitions in the time series, so that the B0 field inhomogeneities also change only minimally. It is therefore possible to calculate the phase maps for the various time points using the different methods described here, which smooth or average the complex image data and / or the phase maps over the time series.This allows the phase to be homogenized across different time points and avoids artifacts in the subtraction image of two time points.

[0024] According to an advantageous embodiment, one phase map is determined by averaging the magnetic resonance data sets over the different time points and determining a phase map from the complex image data of the averaged magnetic resonance data set, and wherein this phase map is used to calculate the fat and / or water images for the entire time series.

[0025] In this embodiment, only a single phase map is used for correcting the magnetic resonance (MRI) datasets. This map is determined by averaging the MRI datasets acquired at different time points. The averaging can, for example, calculate the arithmetic mean over time for each pixel. A phase map is then determined from the averaged MRI dataset using known phase deconstruction methods. This phase map is then used for the entire time series; however, the fat and / or water images are determined from the unaveraged MRI datasets using this phase map. This preserves the image information while advantageously homogenizing the phase correction over time. Besides reducing artifacts, this has the advantage of reducing the computation time for the phase maps.In addition to avoiding artifacts, this also results in a reduction in computing time.

[0026] According to an advantageous embodiment, the one or more phase map(s) are determined by determining a phase map for each magnetic resonance data set of the different time points, and averaging the phase maps of the different time points to obtain an averaged phase map, or smoothing them over time to obtain a smoothed phase map for each time point, wherein the averaged phase map or the smoothed phase maps are used for calculating the fat and / or water images.

[0027] In this embodiment, a phase map is first determined for each magnetic resonance data set in the time series using a known method. The phase maps of the different time points are then averaged to obtain a single averaged phase map. Again, the arithmetic mean can be used. Alternatively, the phase maps of the different time points are smoothed over time to obtain a smoothed phase map for each time point. This also compensates for the B0 field changes between the individual time points and thus avoids artifacts. The phase map averaged in this way, or the multiple smoothed phase maps, are then used for water-fat separation at the individual time points.

[0028] According to an advantageous embodiment, the multiple phase maps are determined by low-pass filtering the magnetic resonance data sets over the various time points and determining a phase map for each time point from the magnetic resonance data set low-pass filtered in the time dimension, wherein the fat and / or water images are determined in particular from the unfiltered magnetic resonance data sets corrected by these phase maps.

[0029] In this embodiment, the two or more magnetic resonance (MRI) datasets are low-pass filtered in the time dimension. The low-pass window can be divided so that, for example, up to five consecutive MRI datasets always contribute. This smooths out any jumps in the phase information. Thus, a low-pass filtered MRI dataset is determined for each time point. The phase map used for correction is then derived from this dataset. However, to ensure the image information remains as accurate as possible, the phase maps are not used to correct the low-pass filtered MRI datasets, but rather to correct the unfiltered MRI datasets. Therefore, averaging over the time series is performed solely for phase correction purposes; the fat and / or water images are determined from the unfiltered MRI datasets.This preserves the image information, but the phase correction is advantageously homogenized over time.

[0030] According to an advantageous embodiment, the multiple phase maps are determined from the complex image data of the magnetic resonance datasets using multidimensional phase unwrapping, where time is one dimension of the phase unwrapping. In this embodiment, the phase unwrapping is performed not only in the spatial direction but also in the temporal direction. The determination of the phase maps by phase unwrapping within the framework of the Dixon method is based on the assumption that the spatial variation of the phase information is weak. This is equally justified for the temporal dimension. Thus, the phase unwrapping in this embodiment can be extended such that time is a (further) dimension of the phase unwrapping. Preferably, the phase unwrapping in this embodiment is not limited to discrete values. Advantageously, this can also result in smoothing.For phase deconstruction, examples of methods can be cited, such as those described in the aforementioned articles by Eggers et al., Ma et al., and Berglund et al. In particular, phase deconstruction methods can be based on techniques like region growing, for example, in conjunction with minimum gradient or minimum variance methods. In this approach, for instance, a pixel is selected, and the region then grows by incorporating the neighboring pixel with the highest quality value.

[0031] According to an advantageous embodiment, an additional step (c) is performed to calculate subtraction images from the fat and / or water images of the magnetic resonance datasets corrected by one or more phase map(s) over the time series. These subtraction images show, for example, the influx of a contrast agent injected at the beginning or before the start of the time series. In particular, the subtraction images are calculated by subtracting the first water and / or fat image of the time series from the second and each subsequent water and / or fat image. This allows the changes compared to the first image to be displayed particularly well. This enables, for example, the assessment of suspicious nodules in the context of DCE, since these accumulate a contrast agent more rapidly.

[0032] According to one implementation, the procedure also includes a step of acquiring a time series of magnetic resonance (MRI) data sets for water-fat separation based on the Dixon method. Specifically, the MRI data sets are acquired from a body part of a subject, particularly a patient. This acquisition is performed, in particular, during an examination using a magnetic resonance imaging (MRI) scanner.

[0033] According to a further aspect, the invention also relates to a computer program comprising software code sections that cause a computer to execute the method according to the invention when the software code sections are executed on the computer. The software code sections can be programmed in any programming language, preferably in a programming language that can be implemented by a control computer of an MRI device.

[0034] The invention also relates to a digital storage medium containing a corresponding computer program. The digital storage medium can be any type of data storage medium, for example, an optical, magnetic, or solid-state storage medium. Examples include a hard drive, a cloud computer, an SD card, an SSD, a USB flash drive, or a CD-ROM.

[0035] According to a further aspect, the invention also relates to a computer, which is in particular a control computer of a magnetic resonance imaging (MRI) device. The computer comprises: an input interface for receiving a time series of two or more magnetic resonance data sets acquired at different times during an examination of a subject for water-fat separation based on the Dixon method, wherein a magnetic resonance data set comprises complex image data, and a processor unit configured to execute the method according to the invention. The computer is in particular connected to or part of an MRI device.

[0036] The processor unit can be any computing unit, in particular a CPU or GPU. However, the method according to the invention can also be executed on a mobile computer, for example a laptop, tablet, or mobile device such as a smartphone. Furthermore, the method according to the invention can also be performed on a cloud computer. In particular, it is possible to export the complex magnetic resonance data sets to a remote computer after acquisition, e.g., via an internet connection, and evaluate them there. The fat and / or water images can then be provided by this remote computer and, if necessary, displayed and / or analyzed at a third location.

[0037] According to one embodiment, the processor unit is configured to calculate subtraction patterns from the fat and / or water images of the magnetic resonance datasets corrected by one or more phase map(s) over the time series, the computer further comprising an output interface for outputting and / or displaying the subtraction patterns. This advantageously allows the subtraction patterns determined according to the invention to be displayed and evaluated immediately.

[0038] According to another aspect, the invention is also directed to a magnetic resonance imaging (MRI) device which contains the computer according to the invention. Otherwise, the MRI device can be equipped as usual. All features and advantages mentioned with regard to the method can also be applied to the computer program, the digital storage medium, and the computer, and vice versa.

[0039] The invention will now be described in more detail with reference to exemplary embodiments and the accompanying drawings. The drawings show: Fig. 1 an embodiment of an MRI device according to the invention; Fig. 2 a flowchart of an embodiment of a method according to the invention; and Fig. 3 an illustration of a combined phase development.

[0040] Fig. Figure 1 shows a section of an MRI scanner 4 according to an embodiment of the invention. The MRI scanner 4 is configured to acquire magnetic resonance images of a patient 2. For this purpose, the patient is moved on a table 3 into the tunnel 6 of the main magnet 1. The main magnet generates the B0 field, i.e., the main magnetic field, which typically ranges between 0.5 T and 7 T, in particular between 1.5 T and 3 T. Other important components (not shown), such as gradient coils and a radio frequency coil for receiving the MR signal, are also arranged within the main magnet 1. A time series of Dixon image data sets can be acquired from the patient 2. The MRI scanner is controlled by the computer 5. The complex image data of the Dixon images can be transferred to the computer 5 via an interface 7, which is configured to execute the method according to the invention.For this purpose, a suitable computer program can be installed on computer 5. The fat and / or water images or subtraction images can be displayed on a screen 8. Computer 5 also has input devices 9 in the form of a mouse and keyboard.

[0041] Fig. Figure 2 shows a flowchart of an embodiment of the method according to the invention. In step 30, a time series of two or more magnetic resonance (MRI) datasets for fat-water separation based on the Dixon method is acquired from a subject, particularly during an MRI examination. The MRI datasets comprise complex image data, which were acquired at at least two different echo times. In step 32, one or more phase maps of the phase errors in the complex image data caused by B0 field inhomogeneities are calculated from these data. This is done using one of the phase deconstruction methods described herein, taking into account all MRI datasets of the time series. The one or more phase map(s) calculated in this way are then used in step 34 to calculate fat and / or water images.In step 36, the fat and / or water images are further processed; in particular, subtraction images of the time series are calculated by subtracting the first water or fat image of each time series. In step 38, the subtraction images thus determined are displayed, e.g., on a screen 8.

[0042] Fig. Figure 3 schematically shows a series of two-dimensional phase maps 10 over time t, which were generated from a corresponding time series of two-dimensional magnetic resonance datasets 14. Using a suitable data model, it is possible to calculate corresponding phase maps 10 from the complex image data of the magnetic resonance datasets 14. According to one embodiment, these are then subjected to phase unfolding not only in the two dimensions x, y of the image plane, but also in the time dimension t. This can be done, for example, by a pixel-wise region growing method, as indicated in Figure 12. QUOTES INCLUDED IN THE DESCRIPTION

[0000] This list of documents cited by the applicant was automatically generated and is included solely for the reader's convenience. The list is not part of the German patent or utility model application. The DPMA accepts no liability for any errors or omissions. Cited non-patent literature

[0000] Jingfei Ma “Dixon Techniques for Water and Fat Imaging”, Journal of Magnetic Resonance Imaging 28: 543 - 558 (2008

[0004] Holger et al. “Dual-echo Dixon Imaging with Flexible Choice of Echo Times,” Magn. Reson. Med. 65(1):96-107 (2001

[0004] Johann Berglund et al. “Two-point Dixon Method with Flexible Echo Times,” Magnetic Resonance in Medicine 65: 994-1004 (2011

[0004]

Claims

[1] Computer-implemented method for reconstructing a time series of two or more magnetic resonance datasets acquired at different times during an examination of a subject for water-fat separation based on the Dixon method (14), wherein a magnetic resonance dataset (14) comprises complex image data acquired, in particular at least two different echo times, wherein the method comprises the following steps: (a) Determining (32) one or more phase maps (10) of the phase errors caused by B0 field inhomogeneities of the complex image data for the two or more magnetic resonance data sets (14) acquired at different times, wherein the determination includes a phase unfolding (12) which is carried out taking into account all magnetic resonance data sets (14) of the time series; (b) Calculating fat and / or water images from the phase map(s) corrected magnetic resonance data sets (14) for the time series. [2] Method according to claim 1, wherein the one or more phase map(s) (10) for the different time points are determined under the assumption that the phase errors change only slightly over time, wherein in particular the phase maps (10) are averaged or smoothed over the time series. [3] Method according to one of the preceding claims, wherein the phase map (10) is determined by averaging the magnetic resonance data sets over the different time points and determining a phase map from the complex image data of the averaged magnetic resonance data set, and wherein this phase map is used to calculate the fat and / or water images for the entire time series. [4] Method according to any of the preceding claims, wherein the one or more phase map(s) (10) are determined by determining a phase map (10) for each magnetic resonance data set (14) of the different time points, and the phase maps of the different time points are averaged (32) to obtain an averaged phase map, or smoothed over time to obtain a smoothed phase map for each time point, wherein the averaged phase map or the smoothed phase maps are used for calculating the fat and / or water images. [5] Verfahren nach einem der vorhergehenden Ansprüche, wobei die mehreren Phasenkarten dadurch bestimmt werden, dass die Magnetresonanzdatensätze (14) über die verschiedenen Zeitpunkte Tiefpass-gefiltert werden und für jeden Zeitpunkt eine Phasenkarte (10) aus dem in der Zeitdimension Tiefpass-gefiltert Magnetresonanzdatensatz bestimmt wird, wobei die Fett- und / oder Wasserbilder insbesondere aus den durch diese Phasenkarten korrigierten ungefilterten Magnetresonanzdatensätzen bestimmt werden. [6] Method according to one of the preceding claims, wherein the multiple phase maps (10) are determined from the complex image data of the magnetic resonance data sets (14) using multidimensional phase unfolding (12), wherein time is a dimension of the phase unfolding. [7] Method according to one of the preceding claims, comprising an additional step (c) of calculating (36) subtraction images from the fat and / or water images of the magnetic resonance data sets corrected by the one or more phase map(s) over the time series. [8] Computer program comprising software code sections which cause a computer (5) to execute a method according to any one of claims 1 to 7 when the software code sections are executed on the computer. [9] Computer (5), in particular control computer of a magnetic resonance imaging device (4), comprising - an input interface (7) for receiving a time series of two or more magnetic resonance datasets (14) recorded at different times during an examination of a subject for water-fat separation based on the Dixon method, wherein a magnetic resonance dataset comprises complex image data, in particular recorded at at least two different echo times, - a processor unit configured to execute the method according to any one of claims 1 to 7. [10] Computer (5) according to claim 9, wherein the processor unit is configured to calculate subtraction images from the fat and / or water images of the magnetic resonance data sets corrected by the one or more phase map(s) over the time series, and wherein the computer further comprises: - an output interface (8) for outputting and / or displaying the subtraction images.

Images