WO2012038168A1

WO2012038168A1 - Improving communication with patients in an mri scanner

Info

Publication number: WO2012038168A1
Application number: PCT/EP2011/064313
Authority: WO
Inventors: Mario Bechtold; Peter-Christian Eccardt; Marco Friedrich; Stefan Nunninger
Original assignee: Siemens Aktiengesellschaft
Priority date: 2010-09-21
Filing date: 2011-08-19
Publication date: 2012-03-29
Also published as: DE102010041146A1

Abstract

The invention aims to improve disturbed speech signals for communicating with a patient in a magnetic resonance imaging (MRI) scanner. Such disturbance results from the known effect of the required gradient currents causing audible vibrations and emissions of background noise such that an acoustic signal, picked up by a microphone in the examination chamber, comprises these portions of background noise as well as the speech signal emanating from a patient. In order to reduce the portion of background noise, the gradient currents are measured and, based on a model, corresponding background noise portions are estimated individually for each gradient current. The acoustic signal that has been picked up is corrected on the basis of the individual background noise portions. Alternatively, and likewise based on a model and using the gradient currents, the vibration of the covering surface of the MRI scanner is simulated and the portion of background noise is determined therefrom.

Description

description

Improving patient communication in an MRI The invention relates to the improvement of disturbed speech signals for patient communication on a magnetic resonance imaging (MRI) scanner.

During an MRI scan, the patient is inside the tube of the MRI while the operator of the MRI is scanning the MRI

Device, such as a doctor or an MTA, located in an adjacent control room in which the control console of the MRI is installed. Communication between operator and patient is not possible directly because the MRI is located in a closed, soundproofed examination room. Therefore, as a rule, a communication system with a microphone in the examination room and a loudspeaker in the control room is provided so that voice utterances of the patient can be picked up by the microphone and transmitted to the loudspeaker. An MRI scan typically consists of one or a whole series of MRI scans, in which the data required for imaging is captured by the known sequence of RF and gradient fields, and examination pauses. This device allows a sufficient understanding of the patient at least during the examination breaks, i. in the period between two measurements.

During a measurement, however, the MRI device basically generates a strong background noise, which is also recorded by the microphone. This significantly interferes with or prevents communication from the patient to the operator during an exam. The characteristic of the noise depends on the type of examination selected, in particular on the set measurement sequence and the measurement parameters. The power of the background noise compared to the performance of the patient speech sound at the microphone depends on the type of examination and the microphone design. At typical In contrast to standard MRI equipment, the power of signal components of the microphone signal due to recorded background noise is often significantly greater than the power of signal components due to the recorded speech sound, which greatly reduces the intelligibility of the patient's speech via the recorded microphone signal.

The different causes of the noise are explained in more detail in connection with FIG.

Various methods for improving disturbed microphone signals for patient communication have been proposed.

In US 4696030 A and US 2005283068 Al, a reduction of the noise is achieved on the basis of a recorded in the past microphone signal, which allows an estimate of the current noise component in the microphone. However, these solutions are highly dependent on the characteristics of the noise. Depending on the type of examination selected, it is possible to achieve a reduction of the interference noise and sufficient speech intelligibility at the output of the communication system, or considerable residual interference noise results despite the use of the method. A multichannel process is described in DE 19524847 C1. The patient language is recorded and enhanced using two microphones, with both microphones recording both speech and noise. Other multi-channel methods, such as, for example, in DE 69032637 T2 and in DE 10151033 Al, use one or more reference microphones which record the noise without speech components in the examination room and generate an estimate of the noise component in the patient microphone using various methods. By subtracting the estimated noise component, in some cases noise reduction and improved speech intelligibility can be achieved. The results, however, are highly dependent on the placement of the Microphones and of the chosen methods for reproducing the noise components.

In EP 0655730 Al a reference signal for the noise in the immediate vicinity of the gradient coils of the MRI is used to generate a replica of the noise in the microphone signal. This replica is used to reduce the amount of noise in the microphone signal by subtracting the microphone signal. Although such a realization can in some cases lead to a reduction of the noise components in the microphone signal, it is limited to noise components which are caused exclusively by the gradient coils, so that other noise generating mechanisms can not be taken into account.

A two-stage signal processing approach, as a combination of model-based and single-channel methods, has been described in "Integrated Speech Enhancement For Functional MRI Environment" by Pathak, N. et al. in Engineering in Medicine and Biology Society, 2009, EMBC 2009, Annual International Conference of the IEEE, 2009, p. 2474 - 2477 published. The solution discussed there uses a microphone signal as a reference for the noise.

It is therefore the object of the present invention to provide a way to improve the communication between a patient in an MRI and the operator of the MRI. This object is achieved by the inventions specified in the independent claims. Advantageous embodiments emerge from the dependent claims.

The mechanisms which are explained in more detail in connection with FIG. 1 and which are based on the production of the noise are due to the currents in the gradient coils required for the MR measurement, ie, to the gradient currents. The invention therefore provides a method for reduction a noise component in an acoustic signal recorded during an RT measurement, wherein at least one gradient current is generated at least temporarily during the MRT measurement, wherein in the method

at least one parameter which represents a measure of the gradient current, as a rule time-variable parameter, is determined,

- The parameter is used in a first stage of a signal processing to a caused due to the gradient current noise in the recorded acoustic

Estimate signal in particular model-based,

- the estimated amount of noise is used to correct the recorded acoustic signal. For example. can be subtracted from the recorded acoustic signal to correct the estimated noise component.

Typically, during MRI measurement, not only a single gradient current but a plurality of gradient currents is generated at least intermittently. Therefore, for each of the gradient currents, a parameter which represents a measure of the respective gradient current, as a rule time-variable parameter, is determined individually. The respective parameter is used in the first stage of the signal processing in order to estimate an individual noise component due to the respective gradient current in the recorded acoustic signal in particular model-based. The estimated individual noise components are then used to correct the recorded acoustic signal. For example. For this purpose, the individual estimated noise components can be subtracted from the recorded acoustic signal.

In an alternative approach, the first stage comprises a first sub-stage and a second sub-class, wherein

in the first sub-step the parameter determines a vibration value which is a measure of a gradient-related vibration of a surface; represents the MRT device, in particular the cladding surface, wherein the vibration is at least a part of the cause of the noise component, and

- the estimation of the noise caused by the gradient current in the recorded acoustic signal in the second sub-stage is based on the determined vibration value.

As mentioned above, during an MRI measurement, a multiplicity of gradient currents is generally generated at least temporarily. For each of the gradient currents, a parameter representing the respective gradient current is determined individually, wherein in the first sub-stage the respective parameter is used to calculate an individual vibration value, wherein the individual vibration values are each a measure of one with the respective gradient current represent coherent vibration of the surface of the MRI device. Furthermore, in the first sub-stage, the vibration value is calculated from the individual vibration values, for which purpose the individual vibration values can, for example, be summed up.

In the first sub-level, linear and non-linear effects can be taken into account. In the second stage, only linear effects are considered.

The respective gradient current can be measured in a first embodiment, so that the parameter representing the gradient current is the corresponding measurement signal.

Alternatively, the parameter representing the respective gradient current can be calculated on the basis of the measurement method underlying the MRT measurement and the selected measurement parameters.

In an alternative approach of a method for reducing a noise component in an acoustic signal recorded during an MRI measurement, during the MRI measurement with at least one sensor determines a vibration value, which is a measure of a vibration of a surface of the MRI apparatus, in particular the cladding surface, wherein the vibration represents at least part of the cause of the noise component. The vibration value is used in a first stage of a signal processing in order to estimate the noise component in the recorded acoustic signal in particular model-based. The estimated amount of noise is used to correct the recorded acoustic signal. For this purpose, the estimated noise component can be subtracted from the recorded acoustic signal.

The estimation of the noise component is preferably model-based. The estimated amount of noise or estimated noise is subtracted from the recorded acoustic signal.

The corrected in the first stage acoustic signal is fed to a second stage of signal processing in which the corrected acoustic signal is reduced using a one-channel algorithm to residual noise and / or independent of the gradient currents noise. Suitable time domain algorithms are described, for example, in the above-mentioned US Pat. No. 4,691,030 A and US Pat. No. 2005283068 A1.

A signal processing device according to the invention for reducing a noise component in an acoustic signal recorded during an MRT measurement on an MRI apparatus, in particular for a communication system of an MRI apparatus, wherein during the MRI measurement at least one gradient current (14} is generated at least temporarily points The arithmetic unit is designed to estimate a noise component due to the gradient current in the recorded acoustic signal, in particular based on a model Noise is used to record the recorded acoustic Correct signal. For example. For this purpose, the estimated noise component can be subtracted from the recorded acoustic signal. Typically, the MRI apparatus has a plurality of gradient coils, and during the MR measurement, a corresponding plurality of gradient currents are generated at least temporarily. The arithmetic unit can therefore have a number of first filters corresponding to the number of gradient coils, wherein the parameter signal comprises a number of individual parameter signals corresponding to the number of gradient currents, which respectively represent a measure of the respective gradient current and which can be fed to the first filters. The first filters are designed to estimate an individual noise component due to the respective gradient current in the recorded acoustic signal, in particular model-based. The first stage has a computing element in which the recorded acoustic signal can be corrected on the basis of the estimated individual noise components. For example. For this purpose, the individual estimated noise components can be subtracted from the recorded acoustic signal.

Alternatively, the arithmetic unit has a first and a second sub-stage, wherein

the first sub-stage has a filter device to which the parameter signal can be fed,

the filter device is designed to determine a vibration value based on the parameter signal, which represents a measure of a gradient current related vibration of a surface of the MRI device, in particular the cladding surface, wherein the vibration represents at least part of the cause of the noise component, and

- The second sub-stage is designed to estimate the caused due to the gradient current noise in the recorded acoustic signal based on the determined vibration value in particular model-based. If the MRI apparatus has a multiplicity of gradient coils and a corresponding multiplicity of gradient currents is generated at least temporarily during the MRT measurement, the filter device of the first sub-stage has a number of second filters corresponding to the number of gradient coils, in which

the parameter signal comprises a number of individual parameter signals corresponding to the number of gradient currents, each of which represents a measure of the respective gradient current and which can be fed to the second filters,

the second filters are designed to respectively estimate an individual vibration value, in particular model-based, wherein the individual vibration values each represent a measure of a vibration of the surface of the MRI apparatus which is related to the respective gradient current, and

- The first sub-stage comprises a computing element for calculating the vibration value from the individual vibration values.

An alternative embodiment of the signal processing device for reducing a noise component in an acoustic signal recorded on an MRI apparatus during an MRI measurement, in particular for a communication system of an MRI apparatus, comprises:

at least one sensor, which can be attached to a surface of the MRI apparatus, in particular on the cladding surface, for determining a vibration value that represents a measure of a vibration of the surface of the MRI apparatus, wherein the vibration represents at least a part of the cause of the noise component .

a first stage with an arithmetic unit, wherein the vibra- tion value can be fed to the arithmetic unit and the arithmetic unit is designed to estimate, based on the vibration value, the noise component in the recorded acoustic signal, in particular model-based, and - A computer to which the recorded acoustic signal and the estimated noise component can be supplied and in which the recorded acoustic signal is correctable based on the estimated noise component.

The signal processing device may have a second stage, to which the acoustic signal corrected in the first stage can be supplied, the second stage being designed to reduce the corrected acoustic signal using a single-channel algorithm for residual interference noise and / or noise free from the gradient currents ,

In summary, according to the invention, the interference noises in the input signal of the communication system are modeled by parallel arrangement of a plurality of adaptive filters, ideally in each case one filter for each measurement signal of the gradient currents or for each gradient coil. In the alternative embodiment, the vibrations of the cladding surface of the MRI system are modeled in a two-stage, model-based approach. From this conclusions about the noise in the input signal of the communication system are drawn. In this embodiment, all relevant noise-generating mechanisms of the MRI system are advantageously detected, and not just the dominant source of interference, ie the vibrations of the gradient coil structure. The introduction of non-linear signal processing approaches, for example the realization of a Hammerstein model, for example with a power filter structure, results in the fact that noise components which can not be reproduced by linear models can also be taken into account. This leads to a further reduction of the noise and an improved adaptation of time-invariant filters. The invention realizes an integration of several signal processing algorithms into a total solution, consisting of model-based approaches in the first stage, followed by a single-channel noise reduction in the second stage.

The multistage improvement of the speech signal in the communication channel means that the solution according to the invention simultaneously achieves the high reduction of the noise by algorithms of the first stage as well as the reduction of any residual noise and of further disturbances uncorrelated to the gradient currents in algorithms of the second stage can.

Further advantages, features and details of the invention will become apparent from the embodiment described below and with reference to the drawings.

1 shows an MRI system with a communication system,

Figure 2 shows a first embodiment of the first stage of a

Signal processing means,

Figure 3 shows a second embodiment of the first stage of a

Signal processing device.

In the figures, identical or corresponding areas, components, component groups or method steps are identified by the same reference numerals.

1 shows a schematic diagram of a sectional view of an MRI apparatus 1 with the relevant parts of the invention for explaining the invention. The MRI 1 is accommodated in an examination room 2, which is cable-insulated from a control room 3. In the control room 3 there is an operator 4 of the MRT 1. The patient 5 to be examined lies in the tube 6 of the MRI 1. In order to transmit speech utterances 7 of the patient 5 to the operator 4 in the control room 3 a communication system 21 with a microphone 8 and an electroacoustic transducer 9, for example. A speaker provided. The microphone 8 picks up the voice utterances 7 of the patient 5, so that they can be played back via the loudspeaker 9.

During an MRI measurement, noise 10 can not be avoided for the reasons described below. The microphone signal S _K would thus, if the patient 5 during a measurement speech uttering of himself, consist of a superposition of the noise 10 and the utterances 7 speech. Accordingly, the microphone signal S _M can be represented as the sum of a noise component S _n and a speech signal component S _sp , ie SM = S _n + S _sp . The signal components S _r due to noise can, depending on the type of the selected MRT measurement and microphone implementation, have a considerably higher power than the signal components S _sp by voice utterances, so that the transmission or intelligibility of the voice utterances 7 in the control room 3 is considerably disturbed.

The interference noises 10 are caused by alternating electric currents in the gradient coils 12 of the MRT device 1. An MRT device 1 contains a gradient coil system 12 with typically several gradient coils 12/1, 12/2, 12/3 (not shown in detail here), which gradient currents 14/1, 14/2, 14 generated individually by a power amplifier 13 / 3 are flown through. The currents 14/1, 14/2, 14/3 associated with the different gradient coils 12/1, 12/2, 12/3 can flow simultaneously and / or sequentially depending on the measurement.

These electrical currents 14 induce a typically tubular structure in which the gradient coils 12 are mounted together to vibrate, through interaction with the strong magnetic field of the MRT system 1, which results in acoustic radiation of the noise 10a from the surface of the gradient coils. Structure 12 lead. The vibrations are also transmitted via the attachment 15 of the gradient coil structure 12 to other parts of the MRI system, which also noise from these parts 10b are emitted.

In addition, the gradient currents 14 lead to an eddy current induction in adjacent conductive structures, for example in the vessel 16 for the superconducting coils and in RF antennas 17 on the cladding 18 of the MRT system 1. The induced eddy currents lead by interaction with the basic magnetic field of the MR MR system 1 and with the gradient magnetic fields of the gradient coils 12 to vibration excitations of various parts of the system, which in turn noise emissions are emitted from these parts. The latter interaction is usually a nonlinear effect.

The illustrated parts of the MRI system are typically located together under the panel 18 of the MR system l _r, for example, a flat plastic panel. The panel 18 is excited mechanically and acoustically by the underlying parts of the system to vibrate, resulting in the emission of the noise 10 in the examination room. The recorded by the microphone 8 in üntersuchungsraum Störschall 10 of the MR system 1 is essentially the radiated from this panel 18 sound.

As mentioned, a communication system 21 is provided with a microphone 8 and a speaker 9. In addition to these two components and possibly existing, but not shown here further components such as amplifiers or the like. has the communication system 21 via a signal processing device 22 according to the invention with a first stage 23 and an optional second stage 24. With the aid of the signal processing means 22, a reduction of the im

Microphone signal S _M contained noise or 10 of the noise component S _r. , In the first stage 23, the noise component S _n is estimated in the microphone signal S _M , based on a determination of the currents 14/1, 14/2, 14/3 in the gradient coils 12/1, 12/2, 12/3 of the MRT 1. Since a plurality of individually supplied gradient coils are usually present in an MRI apparatus, the gradient currents describing parameters or parameter signals 25/1, 25/2, 25/3 are determined individually for each of the gradient coils. For this purpose, a device 35, the gradient currents 14/1, 14/2, 14/3 are supplied.

To determine the parameter set 25, comprising the parameters 25/1, 25/2, 25/3, in principle two different ways are available:

On the one hand, it is possible to actually measure the individual currents 14 generated by the power amplifier 13 in the device 35. The parameters 25 are then exactly these measurement signals.

Alternatively or additionally, it is conceivable to merely calculate or model the currents 14 in the device 35, for example a computer, since the currents required to supply the gradient coils depend on the selected measuring method, for example a spin echo Sequence or a FLASH sequence, and depending on the selected or predetermined measurement parameters such as, for example, the resolution. In this case, the parameters 25 would be the calculated currents needed to perform the measurement method. The computer 35 can then also be the control computer 36 to be operated by the operator 4 in the control room 3.

In the following it will be assumed, without loss of generality, that the currents are actually measured, ie the parameters 25 are the above-mentioned measuring signals. In a first embodiment of the first stage 23 of the signal processing device 22, the gradient signals representing the measuring signals 25/1, 25/2, 25/3 are used to individually for each of the gradient coils 12/1, 12/2, 12/3 the From her caused noise component S _n , i, Sn,?, S _n , 3 in the microphone signal S _M estimate.

FIG. 2 shows a schematic representation of the first stage 23, in which the use of three gradient coils

12/1, 12/2, 12/3 (not shown in detail). Accordingly, a respective filter structure 26/1, 26/2, 26/3 is identified, which determines the deviation between the noise component S _n , i, S _n , 2 _/ S _n , 3 of the respective gradient coil 12/1 estimated in the filter structure , 12/2, 12/3 and the present in the microphone signal S _M real noise component S _n minimized. For this purpose, in each of the filter structures in a statistical analysis based on a correlation between the respective input signal 25/1, 25/2, 25/3 of the filter structure 26/1, 26/2, 26/3 and the microphone signal S _M to the respective Measurement signal 25/1, 25/2, 25/3 associated noise component S _n , i, S _ri , ₂ , S _R , 3 model-based estimated. The noise component S _n in the microphone signal S _K is finally by subtracting the estimated noise components S _c , i _> Sn, 2 _/ S _n , 3 of the respective gradient coil 12/1, 12/2, 12/3 from the microphone signal S _M in a summation 33rd largely reduced, so that at the output of the summation element 33 and the first stage 23, a corrected microphone signal S * 'is present.

Speech signal portions S _{sp of} the microphone signal S _K are not affected by this processing step and are after reduction of the noise S _r. perceptible with significantly improved intelligibility.

The filter structures 26/1, 26/2, 26/3 can each be implemented as a parallel combination of a plurality of adaptive filters whose transmission behavior can be changed so that the deviation between the estimated noise component Sr., i _/ S _n , 2, S _n , 3 and real existing noise S _n in the micro- fonsignal SM is minimized as described above. Such adaptive filters are known per se.

In addition, the system behavior of the filter can be kept temporarily or permanently constant, but this leads to an increase in the deviations and thus the remaining noise in the communication channel with changes in the noise generating system or the characteristic of the noise.

One of several possible embodiments of the error-minimizing filters 26/1, 26/2, 26/3 results when the transmission behavior of the filter corresponds to the transmission behavior of the system path between measuring signal 25 of the gradient current and the associated noise component in the microphone signal to a good approximation.

In an alternative embodiment of the first stage 23, this is subdivided into a plurality of sub-stages 27, 28. A schematic representation of this alternative embodiment is given in FIG. The subdivision is based on the observation that the system properties of parts of the noise-generating system are slowly changing over time, while the system properties of other parts may be subject to rapid temporal changes. Furthermore, this subdivision is based on the observation that noise components S 1 in the microphone signal S _M , which are caused by different parts and transmission paths of the MRI system 1 and by different gradient currents 14, radiate essentially from one common surface 18 into the examination space 2 become. The system properties of this portion of the noise generating system may be considered to be slowly variable or even constant for the application discussed herein. By contrast, the system properties of the acoustic path between the cladding surface 18 of the MRI system 1 and the microphone 8 change significantly and rapidly, for example during a movement of the patient. The subdivision of the first stage 23 of the signal processing 22 therefore takes place in such a way that a first sub-stage 27 contains a parallel arrangement of filters 28/1, 28/2, 28/3 and at least one summation element 29 which, based on the measuring signals 25/1, 25/2, 25/3 of the gradient currents 14/1, 14/2, 14/3 simulate the vibration of the cladding surface 18. In this case, by specifically modeling the vibration of the cladding surface 18, advantageously several of the above-mentioned vibration initiation mechanisms of the cladding surface 18 are considered. The result S _r " _{sum of} the summation in the summation element 29 is an estimate of the vibration excitation S _n , vib of the cladding surface 18, ie S _n , vib ⁼ S _n , stm - the filters 28/1, 28/2, 28/3 of the first Sub-level 27 can be implemented as time-invariant or time-variant filters.

Alternatively or additionally to the use of the filters 28/1, 28/2, 28/3, the vibrations of the panel 18 of the MR system 1, ie the parameter S _n , vib _{/ can} also be detected by means of a sensor 34, for example an acceleration sensor , which is mounted on a suitable location of the panel 18 and outputs a sensor signal S _se n, which is a measure of the vibration excitation of the cladding surface 18. Instead of using the filters 28/1, 28/2, 28/3 estimated signal S _n , vib then the sensor signal S _se n is used to describe the vibrations of the cladding surface 18, ie S _n , vib ⁼ S _sen - This is shown in the figure 3 by the dashed line.

In the event that both the filters 28/1, 28/2, 28/3 and the sensor 34 are to be used for determining the vibrations of the cladding 18, the signal S _R , vib to be estimated can be used as a suitable combination of the Sensor signal S _ser . and the output signal S _n , _{sum of} the summation element 29 are calculated. For example. could be a weighted averaging used here. A second sub-stage 30 contains a further filter 31, which generates an estimate of the noise component S _N in the microphone signal S _M based on the modeled vibration excitation S _n , vi _{b of} the panel 18 of the MRT system 1. The

Noise S _N in the microphone signal S _M are reduced by subtracting the estimated noise components of the microphone signal S _M in a summing 32. The output signal of the summation element 32 or the first stage 23 also corresponds in the embodiment of FIG. 3 to a corrected microphone signal S "'.

Speech signal portions S _{SP of} the microphone signal S _M are not affected by this processing step and are perceptible after reduction of the noise S _N with much improved intelligibility.

An expedient embodiment of the second sub-stage 30 takes place, for example, as an adaptive filter 31, which detects the deviation between the estimated noise S _N and that in FIG

Microphone signal SM real existing noise component minimized. The system behavior of the filters in this embodiment of the first stage of the signal processing can be kept temporarily or permanently constant, but this can lead to an increase in the deviations and thus the remaining noise in the communication channel with changes in the noises generating system or the characteristic of the noise , It has also been observed that the generation of noise components in the microphone signal can not be fully described by linear system behavior. For example, as mentioned above, the interaction between the eddy currents induced by the gradient currents and the gradient currents themselves constitutes a non-linear component. The above-described filters 26, 28, 31 of the first and second embodiments can therefore expressly include both linear and non-linear processing steps. For example. For example, nonlinear filters 28/1, 28/2, 28/3 for the first sub-stage 27 and a linear filter 31 for the second sub-stage 30 may be used. The first sub-stage 27 then forms the vibrations of the cladding surface 18 of the

MRI system 1 for various linear and non-linear excitation mechanisms of the MRI system, while the second stage 30 estimates the recorded background noise 10 after the acoustic path with the aid of a linear filter 31.

With the help of the explained two embodiments of the first stage 23 of the signal processing device 22, the intelligibility of the patient language 7 in the microphone signal S _M can be significantly improved even if the power of the signal components of noise 10 of the MRI is significantly greater than that of the patient language. 7

However, depending on the exact realization of the first stage 23 and the signal properties of the interference noise 10, disturbing residual interference noise S _n , _res t may still be present after the first stage of the signal processing, ie the proportion S _{n of} the noise in the microphone signal S _M sets together with the reducible with the first stage 23 portion, hereinafter denoted by S _n , o, and the residual interference noise corresponding proportion S _n , rest - The corrected microphone signal S _M 'may thus also contain a component S _n , rest , Furthermore, with the first stage 23 further, uncorrelated to the gradient currents 14 noise 10c in the examination room 2, as caused for example by an air conditioner or ventilation 19 or by the helium pump 20 of the MRI system 1, not reduced. These additional interference noises are reflected in the microphone signal S _M as an additional component S _r "div. Summing up all this, the following relationships apply: M s + S _n + n, di i S- _; ⁼ S _n , 0 "t" Sn, rest SM '- SM - S _n , o - S _sp + S _n , _res t + S _n , div

In the optional second stage 24 of the signal processing 22 there is therefore a reduction of the remaining noise S _n , res _{t of} the MRT system 1 and the other, uncorrelated noise S _n , _d iv - For this purpose, known single-channel methods can be used which provide an estimate of the Noise and a reduction of the estimated disturbances in the time domain or frequency range. Possible solution approaches in the time domain are based on an estimate of periodic noise components, as described, for example, by the aforementioned US Pat. No. 4,691,030 A or US 2005283068 A1. Signal enhancements in the frequency domain are achieved, for example, by an individual weighting of frequency bands, so that frequency bands having a dominant noise component are strongly attenuated, while frequency bands having a dominant voice signal component are output almost unattenuated. The output signal S _M * 'of the second stage 24 is then an improved estimate of the undisturbed speech signal S _sp .

The peculiarity of the presented multilevel solution is that after a previous reduction of the noise with the aid of the first stage 23 now also single-channel algorithms can be applied, which were previously not directly applicable due to the low signal-to-noise ratio of the microphone signal S _M. , The single-channel reduction of the residual interference noise and incorrelated interference in the second stage 24 leads to a further improvement in the intelligibility and quality of the patient language 7 in the output signal S _M '' of the communication system 21. The improved signal S _M '' on the speaker rather 9 is output to the operator 4. The operator 4 of the MRI system 1 can understand the examined person 5 during examination breaks as well as during an MRI measurement. The various filters mentioned above in connection with the figures ultimately represent a replica of the entire noise-generating system or the corresponding system path between the interference sound source assigned to the respective filter and the microphone. is a Fehlerminimierende implementation of the filter 26/1 a model of the entire noise generating system, which possibly the amplifier 13, all noise generating mechanisms and transmission paths on the MRI (see above), the emission of sound into the room and the acoustic path taken into account. In contrast, a failure-minimizing implementation of the filter 31 of the second sub-stage 30 simulates the emission of background noise based on an estimate of the vibration of the surface and the acoustic path.

The models used for the respective estimation of the noise components in the various filters for simulating the system sections can be determined, for example, with the aid of corresponding reference measurements, wherein the models can be obtained, for example, from identification measurements. Alternatively or additionally, a physical modeling is basically possible, but due to the complexity and the variance between the systems comparatively expensive.

Claims

claims

1. A method for reducing a noise component (S _n ) in an acoustic signal (S _M ) recorded during an MRT measurement, wherein at least one gradient current (14) is generated at least temporarily during the MRT measurement,

characterized in that

at least one parameter (25) representing a measure of the gradient current (14) is determined,

the parameter (25) is used in a first stage (23) of a signal processor (22) to estimate, in particular model-based, a noise component (S _n ) in the recorded acoustic signal (S _M ) caused by the gradient current (14),

- The estimated noise component (S _n ) is used to correct the recorded acoustic signal (S _M ).

2. The method according to claim 1, characterized in that during the MRT measurement, a plurality of gradient currents (14 / i with 1 = 1,2,3) is generated at least temporarily and for each of the gradient currents (14 / i) individually

a parameter (25 / i with i = 1, 2, 3) representing a measure of the respective gradient current (14 / i) is determined,

the respective parameter (25 / i) in the first stage (23) of the signal processor (22) is used to generate an individual noise component (S _{n /} i) due to the respective gradient current (14 / i) with i = 1, 2, 3) in the recorded acoustic signal (S _M ) in particular model-based estimate,

wherein the estimated individual noise components (S _{n /} i) are subsequently used to correct the recorded acoustic signal (S _M ).

3. The method according to claim 1, characterized in that the first stage (23) comprises a first sub-stage (27) and a second sub-stage (30), wherein

- In the first sub-stage (27) on the basis of the parameter (25) a vibration value (S _{n <V} ib) is determined, which is a measure of a vibration associated with the gradient current (14) of a surface (18) of the MRI apparatus (1), in particular the cladding surface, wherein the vibration represents at least part of the cause of the noise component (S _n ), and

- The estimation of the caused due to the gradient current (14) noise component (S _n ) in the recorded acoustic signal (S _M ) in the second lower stage (30) based on the determined vibration value (S _n , _V ib) in particular model-based.

4. The method according to claim 3, characterized in that during the MRT measurement, a plurality of gradient currents (14 / i with i = l, 2,3) is generated at least temporarily and for each of the gradient currents (14 / i with i = l , 2,3) is determined individually a parameter for the respective gradient current (14 / i with i = l, 2,3) representing parameter (25 / i with i = l, 2, 3), wherein in the first sub-stage ( 27)

- the respective parameter (25 / i) is used to calculate an individual vibration value (S _n , vib, i i = 1, 2, 3), the individual vibration values (S _n , _V i _b , i ) represent in each case a measure of a vibration of the surface (18) of the MRI apparatus (1) associated with the respective gradient current (14 / i), and

- The vibration value (S _n , vib) from which the individual vibration values (S _n , vib, i) is calculated.

5. The method according to claim 3 or 4, characterized in that in the first sub-stage (27) linear and non-linear effects are taken into account.

6. The method according to any one of claims 3 to 5, characterized in that in the second sub-stage (30) only linear effects are taken into account.

7. The method according to any one of the preceding claims, characterized in that the respective gradient current (14 / i) ge and the parameter (25 / i) representing the gradient current (14 / i) is the corresponding measurement signal.

8. The method according to any one of the preceding claims, characterized in that the respective gradient current (14 / i) performing parameters (25 / i) is calculated on the basis of the MRT measurement underlying measurement method and the selected measurement parameters.

9. A method for reducing a noise component (S _n ) in a recorded during an MRI measurement acoustic signal (S _M ), in which

- During the MRI measurement with at least one sensor (34) a vibration value (S _n , _V Ib) is determined, which is a measure of a vibration of a surface (18) of the MRI apparatus (1), in particular the cladding surface, wherein the vibration represents at least part of the cause of the noise component (S _n ),

- The vibration value (S _n , vib) n a first stage (23) of a signal processing (22) is used to estimate the noise component (S _n ) in the recorded acoustic signal (S _M ) in particular model-based, and

10. The method according to any one of the preceding claims, characterized in that in the first stage (23) corrected acoustic signal (S _K ^T ) of a second stage (24) of the signal processing (22) is supplied, in which the corrected acoustic acoustic Signal (SM 'using a single-channel algorithm to generate residual noise (S _n , _re st) and / or noise that is independent of the gradient currents (14 / i)

(S _n , div _/ 10c) is reduced.

11. Signal processing device (22) for reducing a noise component (S _n ) in an acoustic signal (S _M ) recorded during an MRT measurement on an MRT device (1), in particular for a communication system (21) of the MRT device. Device (1), wherein during the MR measurement at least one gradient current (14) is generated at least temporarily, with a first stage (23) having a computing unit (26, 27, 30), wherein

- The calculation unit (26, 27, 30) a parameter signal (25) can be supplied, which is a measure of the gradient current (14), and

- The arithmetic unit (26, 27, 30) is designed to estimate a due to the gradient current (14) caused noise component (S _n ) in the recorded acoustic signal (SM) in particular model-based,

- The estimated noise component (S _n ) is used to correct the recorded acoustic signal (SM) ZU.

12. Signal processing device according to claim 11, wherein the MRT device (1) has a multiplicity of gradient coils (12 / i with i = 1, 2, 3) and during the MRT measurement a corresponding plurality of gradient currents (14 / i with i = 1, 2, 3) is generated at least temporarily,

characterized in that the arithmetic unit (26) has a number of first filters (26 / i with i = 1, 2, 3) corresponding to the number of gradient coils (12 / i), wherein

- The parameter signal (25) comprises a number of gradient currents (14 / i) corresponding number of individual parameter signals (25 / i with i = l, 2,3), each of which represents a measure of the respective gradient current (14 / i) and which are the first filters (26 / i) can be supplied,

- The first filter (26 / i) are formed to a due to the respective gradient current (14 / ii = l, 2,3) caused individual noise component (S _{n /} i with i = l, 2,3) in the recorded acoustic signal (S _M ) in particular model-based estimate,

in which

- The first stage comprises a computing element (33) in which the recorded acoustic signal (S _M ) on the basis of the estimated individual noise components (S _{n /} i) is correctable.

13. Signal processing device according to claim 11, characterized in that the arithmetic unit (27, 30) has a first Lower stage (27) and a second lower stage (30), wherein

- the first sub-stage (27) has a filter device (28) to which the parameter signal (25) can be supplied,

- The filter device (28) is designed to use the parameter signal (25) to determine a vibration value (S _n , _V i _b ) which is a measure of the gradient current (14) associated vibration of a surface (18) of the MRI Device (1), in particular the cladding surface, represents, wherein the vibration is at least part of the cause of the noise component (S _n ), and

- The second stage (30) is designed to estimate the due to the gradient current (14) noise component (S _n ) in the recorded acoustic signal (SM) based on the determined vibration value (S _n , _V ib) in particular model-based.

14. The signal processing device according to claim 13, wherein the MRT device (1) has a multiplicity of gradient coils (12 / i with i = 1, 2, 3) and during the MRT measurement a corresponding plurality of gradient currents (14 / i with i = 1, 2, 3) is generated at least temporarily,

characterized in that the filter device (28) of the first sub-stage (27) has a number of second filters (28 / i with i = 1, 2, 3) corresponding to the number of gradient coils, wherein

- The parameter signal (25) comprises a number of gradient currents (14 / i) corresponding number of individual parameter signals (25 / i with i = l, 2,3), each of which represents a measure of the respective gradient current (14 / i) and which can be fed to the second filters (28 / i with i = 1, 2, 3),

- The second filter (28 / i with i = l, 2,3) are formed to each estimate an individual vibration value (S _n , vi _b , i with i = l, 2,3) in particular model-based, wherein the individual Vibration values (S _n , vib _r i) in each case a measure of one associated with the respective gradient current (14 / i). depict hanging vibration of the surface (18) of the RT device (1), and

- The first sub-stage (30) comprises a computing element (29) for calculating the vibration value (S _n , _V i _b ) from the individual vibration values (S _n , vib, i).

15. Signal processing device (22) for reducing a noise component (S _n ) in an acoustic signal (S _M ) recorded during an MRT measurement on an MRI apparatus (1), in particular for a communication system (21) of the MRI apparatus (1 ), having

at least one sensor (34) attachable to a surface (18) of the MRI apparatus (1), in particular on the cladding surface, for determining a vibration value (S _n , vib) which is a measure of vibration of the surface (18) of the Represents MRT device (1), wherein the vibration represents at least part of the cause of the noise component (S _n ),

- A first stage (23) having a computing unit (30), wherein the vibration value (S _n , vib) of the arithmetic unit (30) can be supplied and the arithmetic unit (30) is designed to be based on the vibration value (S _n , _V ib) estimate the noise component (S _n ) in the recorded acoustic signal (S _M ) in particular model-based, and

- A computing element (32) to which the recorded acoustic signal (S _M ) and the estimated noise component (S _n ) can be supplied and in which the recorded acoustic signal (S _M ) on the basis of the estimated noise component (S _n ) is correctable.

16. Signal processing device according to one of claims 11 to 15, characterized in that a second stage (24) is provided to which in the first stage (23) corrected acoustic signal (S _M ') can be fed, wherein the second te stage (24) is adapted to the corrected acoustic signal (S _M ') using a single-channel algorithm to residual noise (S _n , _res t) and / or from the gradient (14 / i with i = l, 2,3) to reduce independent noises {S _{nf C} üv, 10c).