Computer-implemented method for operating a magnetic resonance device, magnetic resonance device, computer program and electronically readable data carrier

Abstract

The invention relates to a computer-implemented method for operating a magnetic resonance device (20) during an acquisition process according to an acquisition protocol, which uses at least one magnetic resonance sequence in an acquisition period and comprises radio-frequency pulses (2, 4) preceding acquisition periods (3), wherein - for at least some of the high-frequency pulses (2, 4), at least one measured value is recorded in a measuring period (5) describing the reverse power reflected back by a coil element used to output the high-frequency pulses (2, 4), so that a time series (6, 7, 8, 9) of measured values is created over the recording period, - the time series (6, 7, 8, 9) of measured values for detecting an occurring movement of the recorded examination object, which is described by movement information, is evaluated in an evaluation step (S3), and - the movement information is used to control the recording process and / or in a reconstruction of an image data set from recorded magnetic resonance data, characterized in that before the evaluation step (S3), in a preprocessing step (S2), the time series (6, 7, 8, 9) is used to at least partially remove the property from the time series (6, 7, 8, 9) using at least one piece of background information which describes a non-movement-related, expected property of the time series (6, 7, 8, 9).

Description

The invention relates to a computer-implemented method for operating a magnetic resonance device during a recording process according to a recording protocol, which uses at least one magnetic resonance sequence in a recording duration and comprises radio-frequency pulses preceding recording periods, whereinat least one measurement value, which describes the reverse power reflected back by a coil element used for outputting the radio frequency pulses, is recorded in a measurement period for at least a part of the radio frequency pulses, such that a time series of measurement values is produced over the recording period,the time series of measured values for detecting an occurring movement of the recorded examination object, which is described by movement information, is evaluated in an evaluation step, andthe movement information is used for controlling the recording process and / or during a reconstruction of an image data set from recorded magnetic resonance data. In addition, the invention relates to a magnetic resonance device, a computer program and an electronically readable data carrier.Magnetic resonance imaging is a diagnostic and monitoring tool that has now been frequently used in medical applications. Due to the longer recording duration of conventional recording protocols for magnetic resonance data, movement is an important issue with regard to improving the image quality, since movements during the recording duration, even during correction, can lead to losses in the image quality and thus also in the usability of the magnetic resonance data and image datasets reconstructed therefrom. During the recording of a patient, movement relates to cyclic movement processes in the living examination object as well as other, wanted and unwanted, externally and internally triggered movement processes of the recording region.It has proven to be particularly relevant to avoid or take into account movement in the case of TSE sequences (turbo spin echo sequences, also known as HASTE sequences or FSE sequences). In a TSE sequence, a radio-frequency excitation pulse is followed by a so-called echo train in which respective refocusing radio-frequency pulses are each followed by readout periods which all benefit from the same excitation radio-frequency pulse. The recording durations are several minutes in several echo trains. Nevertheless, the TSE sequence is used as a "working horse" of medical imaging because of its merits.In an article by D. Buikman et al., "The RF coil as sensitive motion detector for magnetic resonance imaging", magnetic resonance imaging 6 (1988), pages 218-289, it is proposed to evaluate the reverse power reflected back by coil elements of a radio-frequency coil arrangement used for emitting radio-frequency pulses in order to detect movements of a patient. Cyclic movements, i.e. breathing movements and heart movement, are detected in this case. Non-periodic movements, for example swallows or coughs, can also be detected, so that magnetic resonance data deteriorated by movement can be discarded, which can then be recorded again during non-moving time periods.However, the procedure described therein permits only a rough assessment which, for example, permits the discarding of data affected by strong movement, such as occurs, for example, with the heartbeat, or triggering. A more accurate evaluation, for example with regard to non-periodic and / or weaker movements, is not possible due to further influences which can likewise lead to differences in measured values.US 20200375463 A1 discloses a method for extracting a motion of a subject, wherein S-matrix measurements are performed based on a reflected power of a plurality of RF pulses.The object of the invention is therefore to specify a possibility for improved, in particular more reliable, more robust and high-quality, detection of movements using a high-frequency coil arrangement.This object is achieved according to the invention by a computer-implemented method, a magnetic resonance device, a computer program and an electronically readable data carrier according to the independent claims. Advantageous further developments are evident from the dependent claims.In a method of the type mentioned at the beginning, it is provided according to the invention that before the evaluation step, in a preprocessing step, the time series is used for at least partial cleaning of the time series by the property using at least one piece of background information which describes an expected property of the time series which is not movement-related.The present invention also uses the recording of a time series of measurement values of the back-reflected reverse power of used coil elements of the high-frequency coil arrangement, wherein a directional coupler, for example, can be used for the measurement. If the coil element is also used in any case for receiving magnetic resonance signals, the receiving electronics can also be activated during the transmission of the radio-frequency pulse for which a measured value is to be recorded. Otherwise, dedicated measurement electronics can be provided. Since such measurements are already known in other fields of application, for example for long-term stabilization of the radio-frequency pulse voltage in the case of demanding diffusion sequences, it is a general advantage that no modification is necessary on the magnetic resonance device, in particular no additional hardware or even a modification of the time sequence in the recording protocol, in particular of the sequence timing. Thus, the application of the method described here does not lead in particular to an increase in the recording duration or to changes in the sequence of the recording program, since only the radio-frequency pulses need to be provided with a marking and the recording is controlled by the control device, that is to say by the framework.In this case, for example, measurement values can be recorded during a relatively long portion or even during the entire period in which the radio-frequency pulse is output. This means that the measurement period then corresponds to the duration of the output of the radio-frequency pulse. Specifically, it can be provided that a plurality of measurements are carried out in each measurement period, wherein a measurement value representative of the measurement period is determined by statistical summary, in particular averaging, of the measurement results. A plurality of representative measured values can also be determined in this way. It is also conceivable, though less preferred, to maintain each of the measurement values. After an analog-to-digital converter (ADC) of the measurement electronics is usually switched "open" for measurement in magnetic resonance devices, the measurement period can also be referred to as "TX ADC" (in contrast to an "RX ADC" for the acquisition period).It is now proposed to introduce, as background information, additional prior knowledge which improves the evaluations of the time series of measurement values for finding movement. For this purpose, a preprocessing step is carried out before the evaluation step, in which preprocessing step the background information is used in order to at least partially already remove differences in the measurement values which are not attributable to movements. In other words, the background information thus describes an expected change in the measured values over time in the time series as an expected property of the time series. By incorporating this prior knowledge, the identifiability of movement can be greatly improved.The method can be used in a plurality of types of magnetic resonance sequences, wherein the magnetic resonance sequence is particularly advantageously a TSE sequence in which, in particular within echo trains, individual recording periods can be identified as being influenced by movement. However, the fact that individual recording periods and thus also individual k-space trajectory segments, for example k-space lines, can be identified as being affected by movement also allows applicability to other sequence types, for example to gradient echo sequences (GRE sequences), multi-echo sequences and the like.The measured values can be understood as raw B1 reflection data, which can be corrected in the preprocessing step with prior knowledge about the expected profile. The proposed procedure allows a multiplicity of possible applications, which will be discussed in more detail below. The movement information can be determined reliably, robust and with high quality. It should also be pointed out here that, if a distinction is made below between "movement" and "no movement", this is understood with regard to the relevance of the movement. Relevant movement can be given in particular when a relevance condition for the movement described by the movement information, in particular its strength, is fulfilled.In the recording protocol, radio-frequency pulses can be marked, for which measured values of the back-reflected reverse power are to be recorded. A sequence unit of a control device of the magnetic resonance device, which controls the acquisition mode, can also control the acquisition of the measured values for correspondingly identified radio-frequency pulses. In this case, an expedient development of the present invention provides that at least one measured value is recorded at least for each radio-frequency pulse (immediately) preceding a recording period. Thus, at all recording periods, measured values are present which can be used accordingly for movement detection, which is in particular even specific for recording periods. Exemplary embodiments in which measured values are recorded for all radio-frequency pulses of the magnetic resonance sequence and thus of the recording protocol are also expedient, since movement effects can then be tracked over the entire period of time and the like. In particular, it is therefore possible to identify individual k-space trajectory segments, in particular k-space lines, scanned in a recording period as being affected by movement.A selection of specific radio-frequency pulses is expedient in particular when a variant of the method according to the invention is selected in which radio-frequency pulses which are as similar or even identical as possible are measured in themselves in order already to bring about in this way an excellent comparison of the measured values or of measured value groups, in particular partial time series, of the time series.Thus, an expedient development of the present invention provides that all radio-frequency pulses for which at least one measured value is recorded have at least one identical characteristic variable. In this variant, the selection or labeling of radio-frequency pulses can thus be carried out in such a way that radio-frequency pulses are selected which resemble one another in at least one characteristic variable. This simplifies a correction in the preprocessing step, since the radio-frequency pulses are already basically comparable. In general terms, characteristic variables can comprise a flip angle described by the high-frequency pulse and / or a pulse amplitude and / or a pulse phase and / or a pulse frequency and / or a pulse shape.In particular, in a first variant, as already mentioned, it is conceivable to use radio-frequency pulses which are similar in such a way that a pulse-specific cleaning measure is not necessary, but rather fewer correction measures requiring parameters in the background information can be used, as will be explained in more detail below. For example, it may then be sufficient as background information that, with respect to specific sections or generally of partial time series of the time series, specific profiles will result which can be eliminated, for example by detrenning. In this context, the radio-frequency pulses for which at least one measured value is recorded can particularly advantageously correspond at least in terms of their flip angle, in particular additionally also in terms of their pulse shape. In an example of a TSE sequence as magnetic resonance sequence, refocusing radio-frequency pulses in the echo trains, which relate to an identical flip angle, can be identified for recording at least one measured value.However, since varying flip angles or other varying characteristic variables can frequently occur within magnetic resonance sequences and thus also acquisition protocols, in particular with regard to the flip angle, a practical development of the invention can also provide a pulse-specific correction in a second variant. In particular, it can be provided that the background information describes a characteristic variable that differs between different radio-frequency pulses for which a measured value is recorded. The idea in this embodiment is therefore to correct measured values with prior knowledge about the radio-frequency pulse respectively played out simultaneously. Suitable characteristic variables in which the radio-frequency pulses can differ include, in particular, the pulse frequency, the pulse phase, the flip angle (pulse amplitude or complex pulse integral values over the course of the radio-frequency pulse) and / or the pulse shape.In this case, the at least one characteristic variable in which the radio-frequency pulses differ can be determined, in particular, from control information for generating the radio-frequency pulse. If the transmission chain is controlled, for example, on the basis of data objects which are provided as control information, the at least one characteristic variable can be derived from these data objects in each case for the corresponding high-frequency pulse, in particular already during its output. In other words, ordered real-time events can be supplied to the measuring device and / or the measuring electronics and / or the control device, in particular a measuring unit and / or a preprocessing unit, before or during the execution, preferably a few milliseconds before, for example by means of a so-called real-time event listener (RT event listener). For example, as control information, besides data objects relating to the pulse shape of the radio-frequency pulse and characterizing a real-time event, so-called pulse-frequency pulse-phase events can be processed as further data objects, since the radio-frequency pulses that have been played out and thus their characteristic variables are generated from the combination of these real-time events. The complex-valued envelope of the radio-frequency pulse and the carrier signal parameterized with pulse frequency and pulse phase result in the overall radio-frequency pulse, so that any characteristic variables can be derived therefrom.In particular, it is also generally possible and also expedient in the method according to the invention if at least the portion of the preprocessing step relating to the at least one characteristic variable takes place for each measured value, in particular immediately, after the recording thereof. Consequently, a real-time correction is ultimately possible, which can take place, for example, already before the storage of characteristic values of the time series for later evaluation. This real-time preprocessing, which can therefore take place with respect to each individual radio-frequency pulse, that is to say is pulse-specific, can be realized with little outlay both by software and by hardware, or by a combination of software and hardware. For example, the control information, in particular data objects characterizing real-time events, can be directly queried and used for correction with respect to the at least one characteristic variable in which radio-frequency pulses differ.In this case, it can be provided, for example, that the measured values are normalized in the preprocessing step with respect to at least one of the at least one measured variable, in particular a measured variable describing the flip angle. For example, measured values for radio-frequency pulses with different pulse amplitudes or generally flip angles and / or pulse phases can be made comparable in this way, so that, in general terms, such a correction in the preprocessing step enables a comparison of radio-frequency pulses of different properties. If, for example, the flip angle varies across the radio-frequency pulses, the measured value can be scaled in such a way that a normalization to a specific flip angle value takes place. This enables the reliable and robust determination of movement information even for recording protocols with many variable flip angles, for example in the case of TSE sequences also for echo trains in which the flip angle varies. This is the case, for example, with SPACE (sampling perfect with application optimized contrast using different flip angle evolution) and hyper-echo sequences. If the pulse shapes, for example as refocusing radio-frequency pulses (and possibly other characteristic variables), are the same, it may already be sufficient to normalize the recorded measured values to the different flip angle, in particular by scaling, in order to significantly improve the identification of movement. The same applies correspondingly to other characteristic variables. However, embodiments are naturally also conceivable in which normalization is carried out for a plurality of characteristic variables.Additionally or alternatively to this second variant, an expedient development can provide that characteristic information which assigns characteristic variables to correction values which describe deviations occurring due to different values of the at least one characteristic variable on the basis of the characteristics of the transmission chain used for outputting the radio-frequency pulses is used, wherein the correction value is used for correcting the measured values with respect to the deviations. In this way, characteristics of the transmission chain, for example frequency-dependent attenuation behavior, can also be incorporated into the correction in the preprocessing step. If the behavior of the transmission chain with respect to one or more characteristic variables is known, the measured values can be corrected in accordance with the at least one characteristic variable in order to improve the miscibility or, in general, the evaluability. Specifically, the characteristics can comprise a transmission behavior, in particular a frequency-dependent attenuation behavior of the transmission chain. For example, the characteristic information can therefore comprise a characteristic field and / or a look-up table and / or a mathematical relationship and / or a learned relationship, in particular in the form of a trained assignment function, in order to determine correction values for specific values of the at least one characteristic variable. The characteristics, in particular the transmission behavior, can have been determined beforehand, for example by measurements or the like. In particular, the characteristic information can therefore also be determined empirically.Particularly frequently, in transmission chains in magnetic resonance devices, the frequency-dependent attenuation behavior already mentioned occurs, so that at different pulse frequencies, for example, different attenuation occurs. This is relevant, for example, when the recording protocol relates to recording in a plurality of slices, since a different pulse frequency is required for each of these slices for slice-selective excitation in the presence of a gradient in the slice selection direction. The then occurring different attenuation effect can be at least partially corrected by a corresponding characteristic information which assigns suitable correction values to the pulse frequencies, in particular correction factors for scaling.In summary, in this second variant, by correcting the measured values as B1 reflection raw data with prior knowledge about the respectively output radio frequency pulse, influences such as slice position, flip angle and pulse shape on the time series can be reduced, which facilitates the data analysis in the determination of the movement information and, for example, allows the application to echo trains with variable flip angles and / or different preparation modules.Regardless of whether radio-frequency pulses comparable with respect to at least one characteristic variable are used without a pulse-specific correction (first variant) or whether a pulse-specific correction is carried out on the basis of at least one characteristic variable (second variant), a correction relating to a plurality of measured values and at least partially removing an expected property of the time series can also be carried out in the preprocessing step, which correction, in the case of the first variant or an ignored influence of a characteristic variable or uncovered effects, in the case of the second variant, allows an improvement in the evaluation ability by removing non-movement-related properties of the time series. In other words, in the first variant, a good removal of disruptive influences in the time series can nevertheless be effected by means of a correction with fewer parameters relating to a plurality of measured values, while in the second variant effects remaining after the pulse-specific correction can be corrected.In concrete terms, corresponding embodiments of the method according to the invention can provide that, in the preprocessing step, the time series is initially divided on the basis of at least one piece of time structure information of the at least one piece of background information into partial time series having a property expected for the respective partial time series, after which the cleaning is carried out in a partial time series-specific manner. It is therefore proposed to split the time series retrospectively, i.e. after the recording of a plurality of measured values, into partial time series which show a specific profile property, i.e. in particular a specific expected behavior, which is independent of any movement which may have occurred and is therefore intended to be removed. For this purpose, the measured values are grouped into associated sections, wherein it can be provided in concrete terms that the at least one item of time structure information is selected from the group comprisinga division of the recording protocol into repetitions,a division of the recording protocol into echo trains,dividing the recording protocol into sections having layer groups whose layers are recorded immediately adjacent to one another, anda division of the recording protocol into recording processes for at least one layer.In this way, effects which are exhibited in subgroups, and therefore partial time series, of the time series can therefore be treated in a targeted and robust manner in the preprocessing step. This relates, for example, to properties of the time series which are shown in layer-specific measured values, echo train-specific measured values, repetition-specific measured values and / or section-specific measured values. The measured values can therefore be sorted according to measured values for the slices, measured values for echo trains and the like in order to form partial time series.This will be explained in more detail using an example. If, for example, magnetic resonance data from twenty-seven slices are recorded in eleven repetitions with an echo train length of sixteen using a TSE sequence, measured values can be recorded for all refocusing radio-frequency pulses, such that, for example, a representative measured value can be determined for each recording period. The time series may then include, for example, 4752 representative measurements. If part time series are defined according to repetitions, there are thus eleven part time series each with 297 (representative) measured values, the definition is carried out according to slices, and twenty-seven part time series each with 176 measured values are produced. These can be at least partially corrected by the partial time series if an expected course is known according to the background information. It should be pointed out here that partial time series can also be defined according to different criteria ("dimensions") and can each be preprocessed separately. For example, preprocessing can be effected not only along a repetition, but also additionally along slices and / or echo trains. However, preprocessing only in one division dimension or further splitting into partial time series, for example along echoes per echo train, slices and repetitions, is also conceivable.A division according to layer groups is advantageous in particular when interleaved recording of layer groups takes place. In particular, in the recording protocol, for example to avoid layer crosstalk, a division into layer groups to be recorded successively and each having layers offset from one another can be provided, wherein the layers within each layer group are recorded according to an order, in particular corresponding to a spatial direction. In one embodiment, the slices can be numbered in an order along the slice selection direction, wherein one slice group, in particular to be recorded first, is formed by the slices with odd numbers and another slice group is formed by the slices with even numbers. In such an offset ("interleaved") recording with two layer groups, it has been shown that the number of similar temporal profile sections in the time series can also double, for example a repetition with a similar profile property is present for each repetition which relates to the layer groups one after the other. In the above-mentioned example of eleven repetitions, characteristic sections can therefore result in such a case 22. A division according to these sections can then take place.The time structure information can preferably comprise a set-point progression information, which is in particular the same for each partial time series, wherein the partial time series are at least partially corrected by a progression property described by the set-point progression information. In this case, the setpoint profile information item can in particular specify a functional profile which the measured values are expected to take within the time series in accordance with the profile property. For example, the setpoint profile information can thus indicate that a linear, quadratic or other measured value profile is expected. At least one detrending can be carried out with particular advantage for the correction of the running property. In this way, a trend in data, here the partial time series, can be removed using known methods. Fundamentally known procedures can be used, for example, the detrenning by difference formation, the detrenning by fitting a model and the like. Experiments have shown that linear trends can frequently occur, as can be described by the setpoint profile information, so that linear detrenning can be used. However, linear detrenning has already been found to be useful for other cases.Specifically, as already indicated above, it can be provided that the detection is carried out for a plurality of different separations in partial time series, in particular at least with respect to individual echo trains or layer groups and with respect to individual layers and / or echoes within an echo train and / or repetitions. In other words, an, in particular linear, detection can be effected, for example, both along echo trains and along slices, for which respective corresponding partial time series are defined.Thus, effects relating to different partial time series can be treated and the purification and thus the evaluability is improved.With regard to the evaluation, and therefore the determination of the movement information, it can be provided that the movement information is determined at least partially by comparing measured value groups of the time series with one another and / or with reference information. In this case, measurement value groups can also correspond to partial time series, so that a division from the preprocessing step can advantageously be adopted. For example, a comparison can therefore be made for echo trains, for k-space trajectory sections for slices, sections and the like. The correction performed in the preprocessing step produces excellent comparativeability, so that movements can be identified reliably and in a robust manner even when compared with one another and corresponding movement information can be determined. Reference information can also be used, since a type of "normalization" is carried out by the preprocessing step. Reference information can correspond to a predefined expected value or expected profile without movement, for example as a threshold value. Such a threshold value, for example, indicating the occurrence of movement, can also be obtained from the time series or a measured value group itself, in particular on account of the preprocessing. In a specific embodiment, the threshold value can be determined, for example, as 1.5 to 4 times an average value of the measured values, for example a median. However, the reference information can also be derived specifically for a recording process from sections in which there was certainly no movement.In particular with regard to controlling the recording process itself, it is expedient if at least one comparison already takes place during the recording with the recording protocol. For example, a comparison can always be made when comparable measurement value groups, in particular partial time series, are present. In particular and in a preferred embodiment, it can be provided that a sliding time window, ending at the current point in time, acquiring a plurality of measured values is used. However, it is also possible to compare individual measured values with one another and / or a reference information item, in particular a threshold value. However, it is expedient if the movement information item is determined fundamentally on the basis of a plurality of measured values in order not to overvaluate individual measurement errors / outliers. If measured values representative of measurement periods are used, averaging is already involved and it may be expedient also to determine for individual such representative measured values whether or not movement is present. In any case, extremely rapid detection of movement is possible.A particularly expedient development of the invention provides that, when recording in a plurality of layers, the movement information is determined layer-specifically. The measurement value groups are therefore formed as all measurement values relating to a specific layer. For example, it is possible to determine layer-specific, in particular therefore per layer k-space, portions to be discarded, in particular individual k-space lines. Generally speaking, it can thus be provided, as already discussed, that individual k-space segments, in particular k-space lines, are identified as being affected by movement by the movement information.As already mentioned, the present invention, i.e. the improvement of the basis for the evaluation in the preprocessing step, opens up a multiplicity of possible applications.For example, it can be provided that for at least one radio-frequency pulse without a subsequent recording period, at least one measured value is recorded in a preparation period, in particular for establishing a steady state, wherein the recording of magnetic resonance data is triggered if the movement information does not indicate a current movement. Thus, radio-frequency pulses can be output, for example, in order to maintain a state of equilibrium, wherein the recording of actual magnetic resonance data only takes place when no movement is detected any longer. Such a procedure can prove to be advantageous, for example, in the case of GRE sequences as a magnetic resonance sequence.Additionally or alternatively, it can be provided that, during the recording, magnetic resonance data of a time interval still lasting into the future to be discarded due to a movement, the pulse amplitude of further radio-frequency pulses, in particular the transmission voltage used for generating them, is reduced in the time interval. If, for example, movement is detected during a TSE echo train and it is clear that the magnetic resonance data of the echo train must be discarded, the transmission voltage for subsequent radio-frequency pulses can be regulated downward in the rest of the echo train in order to keep the radio-frequency radiation into the patient as low as possible. It should be noted that the adaptation of the pulse amplitude is expediently taken into account in further preprocessing steps for the measurement values of the radio-frequency pulses for which the adaptation was carried out. This means that, on the basis of the preprocessing step already provided, further measured values can nevertheless be recorded and interpreted in order, for example, to determine whether and when the movement has ended. A further expedient development of the invention can provide in this context that, in the case of magnetic resonance data still to be used in the reduction of the pulse amplitude due to the omission of the movement, said magnetic resonance data are corrected in accordance with the reduction. Once the measure is known, it may also be possible, at least in some cases, to correct recorded magnetic resonance data and still use it.In a particularly expedient development of the present invention, it can be provided that the movement information is used for excluding magnetic resonance data which have been recorded in at least one time segment affected by movement during the reconstruction and / or for the renewed recording thereof. It is therefore possible, for example, to record magnetic resonance data corrupted by movement again. In the example of a TSE sequence, a reacquisition echo train can then be added, which can contain, for example, different k-space trajectory sections that cannot be used due to movement (instead of those that previously belong to an echo train). If no magnetic resonance data were affected in individual segments or all others were already recorded, a practical development of the present invention can provide that the k-space center is redundantly scanned once more in the reacquisition echo train. In other words, the reacquisition echo train can be filled up by further scans of k-space regions to be redundantly recorded. In this way, the signal-to-noise ratio can be increased.However, at least for a portion of magnetic resonance data affected by movement, it may be provided to dispense with a re-acquisition and to carry out a reconstruction based on incomplete magnetic resonance data. In this context, an expedient development of the invention provides that a reconstruction function, in particular a trained reconstruction function, compensating missing magnetic resonance data is used for the reconstruction, in particular a deep resolve boost function based on an unrolled network architecture. In the prior art, techniques based on machine learning have already been proposed in order to achieve a high-quality reconstruction even in the case of magnetic resonance data that are not completely available. These can also be expediently used within the scope of the present invention in order to at least partially avoid a renewed recording of magnetic resonance data which are affected by movement.In further, particularly preferred exemplary embodiments, in particular in the case of dedicatedly determined movement information, it can be provided that the movement information and / or the measured values are transferred to a trained reconstruction function as additional input data. In this way, a reconstruction function can thus experience which magnetic resonance data could be influenced by movement, in particular to what extent, and use this knowledge in the reconstruction of a magnetic resonance image dataset. It is also conceivable that the movement information is used for preconditioning it for at least one correction function which is used in the reconstruction and / or preparing it. Thus, a motion correction can be improved by using the motion information that has been derived from the preprocessed time series in a robust, reliable and highly quantitative manner. Possible such retrospective correction methods include SAMER (Scuut accelerated motion estimation and reduction), cf. D. Polak et al., MRM 87 (2022), pages 163-178, and TAMER (Targeted Motion Estimation and Reduction), cf. M. W. Haskell et al., IEEE Trans Med Imaging 37 (2018), pages 1253-1265. For example, the movement information can be used to select k-space trajectory segments relevant for the correction, in particular k-space lines, or to make a decision as to whether correction at all must be made.Furthermore, it can be provided that the movement information is evaluated during the recording process using the recording protocol for outputting at least one information item to an operator and / or for changing the recording protocol. In this way, an operator can be informed about the occurring movement in particular promptly and react accordingly, for example by counter-steering during the recording process and / or acting on the patient, in particular on the basis of a corresponding warning as information. It is also conceivable to automatically react by automatically adapting the recording protocol during the recording process, for example in the case of regular, small movements to a recording scheme that is more robust with respect to movements and / or a magnetic resonance sequence that is more suitable for occurring movements.In particular, it can be provided in this context that the movement information and / or the measured values of the time series are used as additional input data for a trained image quality function, for example a neural network, that evaluates the image quality, in particular monitors it.In general, a trained function maps cognitive functions that associate humans with other human brains. By training based on training data (machine learning), the trained function is able to adapt to new circumstances and detect and extrapolate patterns.Generally speaking, parameters of a trained function can be adapted by training. In particular, supervised learning, semi-supervised learning, unsupervised learning, reinforcement learning, and / or active learning may be used. In addition, representation learning (also known as "feature learning") can also be used. The parameters of the trained function can be adjusted iteratively by a plurality of training steps.A trained function may include, for example, a neural network, a support vector machine (SVM), a decision tree and / or a Bayesian network, and / or the trained function may be based on k-means clustering, Q-learning, genetic algorithms and / or assignment rules. In particular, a neural network may be a deep neural network, a convolutional neural network (CNN), or a deep CNN. Moreover, the neural network may be an Advanced Network, a Deep Advanced Network, and / or a Generative Advanced Network (GAN).In addition to the method, the present invention also relates to a magnetic resonance device, having a main magnet unit with a main magnet for generating a main magnetic field, a gradient coil arrangement and a radio-frequency coil arrangement, which is designed to generate a magnetic field.a transmission chain for outputting radio-frequency pulses during a recording process according to a recording protocol, which uses at least one magnetic resonance sequence in a recording duration and comprises radio-frequency pulses preceding recording periods, anda measuring device for measuring a reverse power reflected back from a coil element of the radio-frequency coil arrangement, which is used for outputting the radio-frequency pulses, are assigned, wherein the magnetic resonance device further comprises a control device, which has:a sequence unit for controlling the recording process according to the recording protocol and for controlling the measuring device for recording, for at least a portion of the radio-frequency pulses, in each case at least one measured value of the back-reflected reverse power in a measurement period, so that a time series of measured values is produced over the recording period,an evaluation unit for evaluating the time series of measured values for detecting an occurring movement of the recorded examination object, which is described by movement information, in an evaluation step, wherein the sequence unit is designed to use the movement information for controlling the recording process and / or a reconstruction unit is provided, which is designed to use the movement information in a reconstruction of an image data set from recorded magnetic resonance data, anda preprocessing unit which is designed, before the evaluation step, in a preprocessing step, to at least partially clean the property of the time series using at least one piece of background information which describes an expected property of the time series which is not movement-related.All embodiments with respect to the method according to the invention can be transferred analogously to the magnetic resonance device according to the invention and vice versa, so that the same advantages can be obtained.The control device may comprise at least one processor and at least one storage means. Functional units for carrying out steps of the method according to the invention are formed by hardware and / or software, in the present case at least one sequence unit, an evaluation unit and a preprocessing unit, wherein a reconstruction unit is also known in principle for control devices of magnetic resonance devices and is accordingly also provided in a meaningful manner according to the invention. Further functional units can of course also be provided, in particular with regard to the various proposed applications. For example, a control unit for controlling further components of the magnetic resonance device may be provided. The sequence unit corresponds in principle to known sequence units for control devices of magnetic resonance devices that control the acquisition operation, and in the present case also has the possibility of controlling the measuring device for acquiring at least one measured value of the time series, in particular for correspondingly characterized radio frequency pulses. The measuring device can be formed by fundamentally present components, for example, comprise a directional coupler and / or a measuring electronics.A computer program according to the invention can be loaded directly into a storage means of a control device of a magnetic resonance device and has program means such that, when the computer program is executed in the control device, the latter is caused to carry out the steps of a method according to the invention. The computer program can be stored on an electronically readable data carrier according to the invention, which therefore has control information stored thereon, which comprises at least one computer program according to the invention and are configured such that, when the data carrier is used in a control device of a magnetic resonance device, said data carrier is configured to carry out a method according to the invention. The data carrier is in particular a non-transitory data carrier, for example a CD-ROM.Further advantages and details of the present invention are evident from the exemplary embodiments described below and on the basis of the drawings. The following are shown: FIG. 1 shows a general flow chart of exemplary embodiments of the method according to the invention, FIG. 2 shows an exemplary section from an echo train of a TSE sequence, FIG. 3 shows exemplary profiles of measurement values of time series of different recording processes, FIG. 4 shows a first course after cleaning in a preprocessing step, FIG. 5 shows a second curve after cleaning in the preprocessing step, FIG. 6 shows a third curve after cleaning in the preprocessing step, FIG. 7 shows a fourth course after cleaning in the preprocessing step, FIG. 8 shows a comparison of measurement values with a scaled difference (offset) of a pulse frequency from a nominal Larmor frequency of a magnetic resonance device, FIG. 9 shows a flow chart of a first possible sequence of undershoots of the preprocessing step, FIG. 10 shows a flow chart of a second possible sequence of undershoots of the preprocessing step FIG. 11 shows a layer-specific profile of measurement values of the fourth profile for a first layer, FIG. 12 shows a layer-specific profile of measurement values of the fourth profile for a second layer, FIG. 13 shows a schematic diagram of a magnetic resonance device according to the invention, and FIG. 14 shows the functional structure of the magnetic resonance device of FIG. 13.FIG. 1 shows a general flow chart of exemplary embodiments of the method according to the invention. In the following, as a concrete example, the application is used in a recording process with a recording protocol in which a TSE sequence is used. Twenty-seven slices are recorded in eleven repetitions, wherein in each repetition firstly echo trains of an echo train length of sixteen are used for the slices with odd slice numbers of an order along the slice selection direction (slices one, three,..., twenty-seven) and then for the slices with even slice numbers (slices two, four,..., twenty-six) in order to record magnetic resonance data, in the present case a k-spatial line in each recording period following a refocusing radio-frequency pulse. These parameters and also the use of a TSE sequence are to be understood purely as examples.According to a step S 1, measured values are recorded for at least some of the radio-frequency pulses of the recording protocol in a measurement period during the output of the radio-frequency pulse, which describes the back-reflected reverse power from the at least one used coil element of the radio-frequency coil arrangement of the magnetic resonance device that is used. For example, a directional coupler can be provided in the feed line to the coil element, by means of which directional coupler the back-reflected reverse power can be measured on the basis of the reverse voltage by means of a measuring device, which can also comprise measurement electronics. It should be noted that a measuring device can be used which can also measure the forward voltage, for example for SAR considerations.In the present case, measured values are recorded for each refocusing radio-frequency pulse that immediately precedes a recording period. Exemplary embodiments are also conceivable in which measured values are recorded for each radio-frequency pulse or for other selected radio-frequency pulses. The corresponding radio-frequency pulses are identified in the recording protocol, so that a sequence unit can actuate the measuring device for recording measured values. During each measurement period, a plurality of measurements are taken, for example 128 measurements which are statistically processed to produce at least one representative measurement value for the measurement period and thus acquisition period.FIG. 2 shows a section from the sequence diagram of an echo train 1, namely its beginning. As can be seen, in the present specific example, no measured value is recorded for the excitation radio-frequency pulse 2, but apart from the recording periods 3 which follow the refocusing radio-frequency pulses 4, a measurement period 5 is also provided for each refocusing radio-frequency pulse 4 for recording the at least one measured value. While the acquisition period 3 can also be referred to as RX ADC due to the open ADC, the measurement period 5 then corresponds to a TX ADC.FIG. 3 shows possible time series 6, 7, 8, 9, and therefore curves, of measured values for the specific example. Each of these time series 6, 7, 8, 9 is based on a different movement pattern. All measurements clearly show that they consist of the eleven repetitions, wherein the twenty-seven layers were recorded in an interleaved manner, so that twenty-two characteristic sections result. If no movement takes place, compare the first time series 6, the measured value profile for each characteristic section which is defined by a layer group is similar. In the second course, second time series 7, a movement toward the end is present, in the third course, third time series 8, a movement in the middle of the recording duration. In the fourth time series 9, two movements which are short over time, specifically head kicks, finally occur.While the movements lasting longer may obviously also be recognizable as deviations for the human observer in the time series 6, 7, 8, 9 of FIG. 3, which can be referred to as B1 reflection raw data, this is already more difficult to perceive for the time series 9 (short movements). This is due to a large number of further properties of the curves of the measured values which do not correlate with the movement and which can also make automatic evaluation more difficult. It is therefore proposed to carry out a preprocessing based on background knowledge in a preprocessing step S 2 first of all before the evaluation in the evaluation step S 3 of FIG. 1 in order to significantly improve the identifiability of movement. In preprocessing step S 2, background information describing a non-movement-related, expected property of time series 6, 7, 8, 9 is used in concrete terms to at least partially clean time series 6, 7, 8, 9 about this property.FIGS. 4 to 7 show exemplary results of preprocessing step S 2 for time series 6, 7, 8 and 9, the eleven repetitions over rising ramps 10 being indicated in each case for improved orientation, and the double of the median value of the (corrected) measured values (as a potential threshold value for identifying relevant movements) being shown by a line 11. FIG. 4 shows the corrected time series 12 to the first time series 6, where it can now be clearly seen that no greater deflections are present (no movement) and the measured values remain at least substantially constant.In the corrected time series 13 to the second time series 7, the movement at the end of the recording period can be clearly seen, just as in the corrected time series 14 to the third time series 8 the movement in the middle. In the corrected time series 15 to the fourth time series 9, the short pitching movements can be recognized particularly well as distinct excursions, since they were barely recognizable in FIG. 3, time series 9.The movement patterns can thus be clearly recognized and identified in all cases. In comparison with the original course of the time series 6, 7, 8 and 9, in particular the points in time of the movement taking place emerge more sharply in the corrected time series 12, 13, 14 and 15. It should be noted here that it is also conceivable to configure preprocessing parameters of preprocessing step S 2 to be adaptable in order to emphasize more greatly various aspects of the movement that are to be detected, and therefore to ascertain corrected time series 12, 13, 14 and 15 in a targeted manner with regard to the evaluation taking place. Since the points in time at which movement is present are relevant in the present case, the optimum distance of further, non-movement-related properties of the time series 6, 7, 8 and 9 is preferred and is used further in the following.There are various possibilities for the specific implementation of step S 2. It is conceivable, on the one hand, if the radio-frequency pulses correspond in one or preferably even a plurality of characteristic variables, to correct or clean only a property of the profile of the measured values or of partial time series that can be described by a plurality of parameters. This is useful, for example, when the refocusing radio-frequency pulses 4 all have the same pulse shape and relate to the same flip angle. However, a pulse-specific correction based on at least one characteristic variable of the radio-frequency pulse, in particular its flip angle and / or its pulse frequency and / or pulse phase, is also possible and preferred. Such a pulse-specific correction can then nevertheless also be followed by a profile correction over partial time series, and therefore a plurality of measurement values, in order to remove remaining effects.FIG. 8 shows in this respect, as a section of any desired time series, the profile 16 of the real part of the measurement values in comparison with the profile 17 of the frequency offset from the Larmor frequency for the respective radio-frequency pulses, wherein the frequency offset has been scaled appropriately. A high correlation is seen, which can be utilized to remove the dependence of the measured values on the slice position. Not shown is the pulse phase, which likewise varies depending on the slice position and can likewise be corrected.It can also be shown that the value profile of the measured values during a measurement period 5 reflects the profile of the radio-frequency pulse, i.e. both as regards the pulse frequency and the pulse phase. Both characteristic variables (and also further characteristic variables of the radio-frequency pulse) can easily be derived from control information to the transmission chain, in particular data objects describing real-time events. This also applies in particular to flip angles caused by radio-frequency pulses.As the flow chart of FIG. 9 shows, the pulse-specific correction can be implemented, for example, in concrete terms in such a way that, first, in a step S 2 a, the at least one characteristic variable to which the correction is to be made is determined from control information in the corresponding measurement periods 5. For example, these can be the three characteristic variables flip angle, pulse phase and pulse frequency mentioned.In a step S 2 b, as a first correction step, in particular immediately after a measured value has been recorded and before it is stored in the time series, a normalization with respect to at least one of the at least one characteristic variable determined in step S 2 atakes place. This has proven to be expedient, for example, with regard to at least the flip angle. Corresponding normalization factors can be determined empirically, for example, but also in another way. For example, scaling can be effected with respect to the flip angle. The normalization information usable for this purpose forms part of the background information.If normalization is carried out with respect to the flip angle in this way, it is possible, for example, to still allow reasonable evaluation in the case of magnetic resonance sequences with many different flip angles.In a step S 2 c, characteristics of the transmission chain can also be included, in the present case specifically frequency-dependent attenuation behavior. For this purpose, the background information includes characteristic information which assigns correction values to characteristic variables which describe deviations which occur due to different values of the at least one characteristic variable on the basis of the characteristics of the transmission chain used for outputting the radio-frequency pulses. In the present case, a corresponding correction factor is assigned to the pulse frequency as a characteristic variable in order to correct a frequency-dependent damping behavior. This is applied to the measurement value to improve the comparison. Step S 2 cis also preferably carried out immediately after the measurement, optionally before the measured value is actually entered into the time series. A pulse-specific real-time correction is thus provided.It should be noted here that this can also be implemented without problems and with relatively low outlay on the basis of the control information which is present in any case on the transmission chain, be it in hardware or else in software.FIG. 10 shows a flow chart of falling below step S 2 in order to make corrections on the basis of properties expected for partial time series. The portion of the background information necessary for this is at least one piece of time structure information which also comprises a piece of setpoint profile information.In a step S 2 d, the measured values are grouped into associated partial time series. This grouping can be effected according to repetitions, slices, characteristic sections (slice groups), echo trains and the like. For example, for the characteristic sections of the layer groups with even and odd layer numbers of each repetition, it is evident from FIG. 3 that a roughly linearly decreasing trend is present. If a division were carried out according to repetitions, two such roughly linearly decreasing trends would be present for the layer groups. This expectation of the profile of the measured values within the partial time series is described by the setpoint profile information.It should also be noted here that the division into partial time series does not necessarily have to contain measurement values recorded directly in succession, but can also relate, for example in the case of slices, to measurement values recorded at a time interval. Trends which can be described by means of the desired profile information can also be present for such partial time series, depending on the specific recording protocol or specific magnetic resonance sequence.In step S 2 e, the setpoint profile information is then used to at least partially clean the partial time series from the profile property expected for this. To this end, detrending can be carried out. In the specific example to which the graphs shown also relate, a linear detection for the characteristic sections can be set, for example. Linear detection is preferably carried out for a plurality of splitting dimensions, for example both along the echo trains and along slices or shots. Also conceivable is detection for only one type of partial time series, but division according to several division dimensions, for example echoes per echo train, slices and repetitions.It should be noted here once again that the specific time structure information can easily be determined empirically and / or analytically for specific application cases and can correspondingly differ from application case to application case, in particular magnetic resonance sequence to magnetic resonance sequence or recording protocol to recording protocol.The correction according to steps S 2 dand S 2 emay also be implemented before the recording protocol is completed, i.e. before the recording duration has elapsed, for example always as soon as a partial time series is complete or contains sufficient measured values.Returning to FIG. 1, in an evaluation step S 3, an evaluation of the respective currently recorded, cleaned time series 12, 13, 14, 15 is carried out in order to determine movement information which is then used in a step S 4 in order to influence the control of the recording process and / or the reconstruction of an image dataset from the recorded magnetic resonance data if necessary. Consequently, the evaluation in the evaluation step S 3 can also take place at least partially already during the recording duration. It is expedient here to use a sliding time window that ends at the current measurement value. The evaluation can comprise in particular comparisons of measurement value groups, in particular also partial time series, with one another and / or with reference information, for example the threshold value already addressed.There are a plurality of specific possible uses of the motion information item which can be determined in an improved, in particular robust, reliable and highly quantitative manner or of the corrected time series 12, 13, 14, 15, of which at least a part is to be explained again below.Measurement values are preferably divided into layer-specific measurement value groups if such measurement value groups are not already present in any case as partial time series. It is therefore possible for curves of the measured values of the corrected time series 12, 13, 14, 15 to be formed and evaluated for each layer, for example by corresponding comparisons. This will be explained in more detail again for the corrected time series 15 originating from the fourth time series 9 with reference to FIGS. 11 and 12. In addition to the resulting measurement value groups, i.e. layer-specific curves 18 for a first layer and 19 for a second layer, there is also indicated as stages from which repetition the corresponding plotted measurement values originate. Thus, sixteen measured values (number of recording periods 3 in each echo train 1, i.e. echo train length) are present for each stage in the specific example. It can be seen very clearly for both slices that specific echo trains 1 are affected by the first pitch, i.e. the first brief movement. Whereas the third echo train 1 was affected in the first layer (FIG. 11 ), it is already the fourth echo train 1 in the second layer (FIG. 12 ). In a further refinement, it is even possible to identify individual k-space lines affected by movement. This allows application also in other magnetic resonance sequences, such as GRE sequences.Based on the movement information, recorded magnetic resonance data can be recognized as being corrupted by movement. This magnetic resonance data can then be excluded from a reconstruction, for example, wherein a trained reconstruction function is preferably used for the reconstruction in step S 4, which is designed to compensate missing magnetic resonance data, for example a deep resolve boost function.It is additionally or alternatively also conceivable to discard corrupted magnetic resonance data and, for example, to resume it again at the end or inserted in an additional reacquisition echo train or generally a reacquisition protocol section. In this case, if individual sections of the reacquisition protocol section are not required for re-recording magnetic resonance data, repeated recording of the k-space center can take place in order to increase the signal-to-noise ratio in this way.It is also possible with the procedure described here to achieve a triggered recording by initially outputting radio-frequency pulses without subsequent recording period 3 in order to maintain a steady state (steady state) until the measured values indicate that no movement is detected any longer. Magnetic resonance data are then recorded.If movement is detected during a TSE echo train and it follows therefrom that the magnetic resonance data of the echo train must be discarded, the transmission voltage for the radio-frequency pulses for the rest of the echo train can be regulated downward in order to keep the radio-frequency radiation into the patient and thus the SAR loading as low as possible. Nevertheless, it is possible to record further measured values for the time series 6, 7, 8, 9, since a correction with respect to this reduced transmission voltage (for example, characteristic variable flip angle: scaling / normalization) is possible within the framework of the cleaning.A combination with methods for assessing the image quality can also be carried out in that, for example, the movement information and / or the measured values form additional input data for an image quality function. In general, an evaluation can also be carried out with regard to an automatic control action of the recording process and / or the output of information to an operator. For example, a warning can be output to an operator during movement or a change can be made to a different recording scheme or a different magnetic resonance sequence.It is also possible to use the movement information and / or the measured values of the corrected time series 12, 13, 14, 15 for preconditioning retrospective correction methods such as TAMER or SAME, for example in order to select k-space lines relevant for the correction or to make a decision as to whether a correction has to be carried out.FIG. 13 shows a schematic diagram of a magnetic resonance device 20 according to the invention. The magnetic resonance device 20 comprises a main magnet unit 21, which contains a main magnet that is superconducting here and not shown in detail, and has a cylindrical patient holder 22, into which a patient can be moved for a recording process with a patient table not shown in detail here. In the present case, a gradient coil arrangement 23 and a radio-frequency coil arrangement 24 are provided surrounding the patient holder 22. The high-frequency coil arrangement 24 can also be provided at least partially as a local coil arrangement. The high-frequency coil arrangement 24 is actuated by means of a transmission chain 25 for the output of high-frequency pulses 2, 4. Likewise merely indicated in FIG. 13 is the measuring device 26 for recording measured values which describe the back-reflected reverse power for at least one coil element of the high-frequency coil arrangement 24.The operation of the magnetic resonance device 20 is controlled by means of a control device 27, the functional structure of which is described in more detail by FIG. 14. Accordingly, the control device 27 first comprises a storage means 28, in which various data, in particular also magnetic resonance data, corrected time series 12, 13, 14, 15, movement information and the like can be stored.The control device 27 further comprises a sequence unit 29 which controls the recording operation of the magnetic resonance device 20 and is also designed in the present case to control the measurement device 26 for recording measurement values at least for correspondingly identified radio-frequency pulses of a recording protocol according to step S 1. In a preprocessing unit 30, preprocessing step S 2 is carried out, in particular with the corresponding undershoots S 2 ato S 2 cand / or S 2 dand S 2 e. As already mentioned, the preprocessing unit 30, in particular for immediate pulse-specific correction, steps S 2 ato S 2 c, can also comprise hardware components.In an evaluation unit 31, the movement information can be ascertained according to step S 3. A reconstruction unit 32 serves for reconstructing image data sets from recorded magnetic resonance data and can be configured according to the described possibilities for step S 4. A control unit 33 for other components of the magnetic resonance device 20 can also be designed for at least partially executing step S 4. Finally, in particular with regard to interventions in the sequence of the recording protocol or generally the control of the recording process, the sequence unit 29 can also be designed for at least partially carrying out step S 4, for example for renewed recording of magnetic resonance data corrupted by movement.Regardless of the grammatical sex of a certain term, individuals with male, female or other sex identity are included.

Claims

Computer-implemented method for operating a magnetic resonance device (20) during a recording process according to a recording protocol which uses at least one magnetic resonance sequence in a recording duration and comprises radio-frequency pulses (2, 4) preceding recording periods (3), wherein - for at least some of the radio-frequency pulses (2, 4) in each case at least one measured value which describes the reverse power reflected back by a coil element which is used for outputting the radio-frequency pulses (2, 4) is recorded in a measurement period (5), so that a time series (6, 7, 8, 9) of measured values arises over the recording duration, - the time series (6, 7, 8, 9) of measured values is evaluated in an evaluation step (S3) in order to detect an occurring movement of the recorded examination object which is described by a movement information item, and - the movement information is used for controlling the recording process and / or during a reconstruction of an image data set from recorded magnetic resonance data, wherein before the evaluation step (S3), in a preprocessing step (S2), the time series (6, 7, 8, 9) is used for at least partial cleaning of the time series (6, 7, 8, 9) by the property using at least one piece of background information which describes an expected property of the time series (6, 7, 8, 9) which is not movement-related, characterized in that the background information describes a characteristic variable which differs between different radio frequency pulses for which a measured value is recorded.Method according to Claim 1, characterized in that at least one measured value is recorded at least for each high-frequency pulse (4) preceding a recording period (3).Method according to one of the preceding claims, characterized in that the characteristic variable relates to a flip angle described by the radio-frequency pulse (2, 4) and / or a pulse amplitude and / or a pulse phase and / or a pulse frequency and / or a pulse shape, and / or the characteristic variable is determined from control information for generating the radio-frequency pulse.Method according to one of the preceding claims, characterized in that the measured values are normalized in the preprocessing step (S2) with respect to at least one of the at least one measured variable, in particular a measured variable describing the flip angle, and / or in that characteristic information which assigns characteristic variables to correction values which describe deviations which occur as a result of different values of the at least one characteristic variable on the basis of the characteristics of the transmission chain (25) used for outputting the radio-frequency pulses is used, wherein the correction value is used for correcting the measured values with respect to the deviations.Computer-implemented method for operating a magnetic resonance device (20) during a recording process according to a recording protocol which uses at least one magnetic resonance sequence in a recording duration and comprises radio-frequency pulses (2, 4) preceding recording periods (3), wherein - for at least some of the radio-frequency pulses (2, 4) in each case at least one measured value which describes the reverse power reflected back by a coil element which is used for outputting the radio-frequency pulses (2, 4) is recorded in a measurement period (5), so that a time series (6, 7, 8, 9) of measured values arises over the recording duration, - the time series (6, 7, 8, 9) of measured values is evaluated in an evaluation step (S3) in order to detect an occurring movement of the recorded examination object which is described by a movement information item, and - the movement information is used for controlling the recording process and / or during a reconstruction of an image data set from recorded magnetic resonance data, wherein before the evaluation step (S3), in a preprocessing step (S2), the time series (6, 7, 8, 9) is used for at least partial cleaning of the time series (6, 7, 8, 9) by the property using at least one piece of background information which describes an expected property of the time series (6, 7, 8, 9) which is not movement-related, characterized in that, in the preprocessing step (S2), the time series (6, 7, 8, 9) is first divided on the basis of at least one piece of time structure information of the at least one piece of background information into partial time series having a property expected for the respective partial time series, after which the cleaning is carried out in a partial time series-specific manner.Method according to Claim 5, characterized in that the at least one item of time structure information is selected from the group comprising - a division of the recording protocol into repetitions, - a division of the recording protocol into echo trains (1), - a division of the recording protocol into sections having layer groups, the layers of which are recorded immediately adjacent, and - a division of the recording protocol into recording processes for at least one layer.Method according to Claim 5 or 6, characterized in that the time structure information comprises a set point profile information item which is identical in particular for each partial time series, wherein the partial time series are corrected at least partially by a profile property described by the set point profile information item.Method according to Claim 7, characterized in that at least one detrending, in particular a linear detrending, is carried out in order to correct the running property.Method according to one of the preceding claims, characterized in that, when recording in a plurality of slices, the movement information is determined slice-specifically and / or in that individual k-space sections, in particular k-space lines, are identified as being affected by movement by the movement information.Method according to one of the preceding claims, characterized in that the movement information is used for excluding magnetic resonance data which have been recorded in at least one time segment affected by movement, during the reconstruction and / or for the renewed recording thereof.Method according to Claim 10, characterized in that a reconstruction function, in particular a deep resolve boost function based on an unrolled network architecture, compensating for missing magnetic resonance data is used for the reconstruction.Magnetic resonance device (20), comprising a main magnet unit (21) with a main magnet for generating a main magnetic field, a gradient coil arrangement (23) and a radio-frequency coil arrangement (24), which comprises - a transmission chain (25) for outputting radio-frequency pulses (2, 4) during a recording process according to a recording protocol, which uses at least one magnetic resonance sequence in a recording duration and radio-frequency pulses (2, 4) which precede recording periods (3), and - a measuring device (26) for measuring a reverse power reflected back from a coil element of the radio-frequency coil arrangement (24), which is used for outputting the radio-frequency pulses, wherein the magnetic resonance device (20) further comprises a control device (27), which comprises: - a sequence unit (29) for controlling the recording process according to the recording protocol and for controlling the measuring device (26) for recording, for at least a portion of the radio-frequency pulses (2, 4), in each case at least one measurement value of the back-reflected reverse power in a measurement period, so that a time series (6, 7, 8, 9) of measurement values arises over the recording duration, - an evaluation unit (31) for evaluating the time series (6, 7, 8, 9) of measurement values for detecting an occurring movement of the recorded examination object, which is described by a movement information item, in an evaluation step (S 3), wherein the sequence unit (29) is designed to use the movement information item for controlling the recording process and / or a reconstruction unit (32) is provided, which is designed to use the movement information item in a reconstruction of an image data set from recorded magnetic resonance data, and - a preprocessing unit (30), which is designed to, before the evaluation step (S 3), in a preprocessing step (S 2), the time series (6, 7, 8, 9) using at least one piece of background information which describes an expected property of the time series (6, 7, 8, 9) which is not movement-related and a characteristic variable which differs between different high-frequency pulses at which a measured value is recorded, at least partially in order to clean the property.Magnetic resonance device (20), comprising a main magnet unit (21) with a main magnet for generating a main magnetic field, a gradient coil arrangement (23) and a radio-frequency coil arrangement (24), which comprises - a transmission chain (25) for outputting radio-frequency pulses (2, 4) during a recording process according to a recording protocol, which uses at least one magnetic resonance sequence in a recording duration and radio-frequency pulses (2, 4) which precede recording periods (3), and - a measuring device (26) for measuring a reverse power reflected back from a coil element of the radio-frequency coil arrangement (24), which is used for outputting the radio-frequency pulses, wherein the magnetic resonance device (20) further comprises a control device (27), which comprises: - a sequence unit (29) for controlling the recording process according to the recording protocol and for controlling the measuring device (26) for recording, for at least a portion of the radio-frequency pulses (2, 4), in each case at least one measurement value of the back-reflected reverse power in a measurement period, so that a time series (6, 7, 8, 9) of measurement values arises over the recording duration, - an evaluation unit (31) for evaluating the time series (6, 7, 8, 9) of measurement values for detecting an occurring movement of the recorded examination object, which is described by a movement information item, in an evaluation step (S 3), wherein the sequence unit (29) is designed to use the movement information item for controlling the recording process and / or a reconstruction unit (32) is provided, which is designed to use the movement information item in a reconstruction of an image data set from recorded magnetic resonance data, and - a preprocessing unit (30), which is designed to, before the evaluation step (S 3), in a preprocessing step (S 2), the time series (6, 7, 8, 9) using at least one piece of background information which describes an expected property of the time series (6, 7, 8, 9) which is not movement-related, at least partially in order to clean the property and firstly to divide the time series (6, 7, 8, 9) into partial time series having a property expected for the respective partial time series on the basis of at least one piece of time structure information of the at least one piece of background information, after which the cleaning is carried out in a partial time series-specific manner.Computer program which has program means such that, when the computer program is executed on a control device (27) of a magnetic resonance device (20), said computer program is caused to carry out the steps of a method according to one of Claims 1 to 11.Electronically readable data carrier on which a computer program according to claim 14 is stored.

Images