Physiological measurement device and method

Abstract

The present invention relates to a physiological measurement device having a plurality of input channels, at least two of which include a digital section configured to process digital signals representative of analog signals present at electrodes assigned to the respective channel, and a signal combiner configured to calculate at least one vector signal from the at least two processed digital signals. In order to achieve an improved common mode interference mitigation, it is proposed that the digital section of at least one input channel comprises a digital filter configured to filter the digital signal based on a set of filter coefficients, and a processor configured to calculate the set of filter coefficients based on the vector signal.

Description

[Technical field]

[0001] The present invention relates to a physiological measurement device and a respective method. Such a device has a plurality of input channels, at least one of which includes a digital part configured to process at least one digital signal representative of an analog signal present at an electrode assigned to the respective channel, and a signal combiner configured to calculate at least one vector signal from at least two processed digital signals. [Background technology]

[0002] A combiner of the above mentioned apparatus combines at least the processed digital signals into a vector signal such that common mode interference present in the processed digital signals is mitigated. Summary of the Invention [Problem to be solved by the invention]

[0003] A known problem when making electrophysiological measurements, such as electrocardiography (ECG), electroencephalography (EEG), and electromyography (EMG), is that the signal paths of each electrode have slightly different transfer functions due to component tolerances, different external cables and electrical properties of the skin-electrode interface, etc. As a result, common-mode signals are only partially canceled.

[0004] The published patent application DE 10 2014 214 994 A1 describes a differential voltage measurement system with an analog variable impedance element, such as an RC circuit, which is controlled for the purpose of impedance matching of the two inputs of the measurement system. Furthermore, a combination of this impedance matching method with Right Leg Drive (RLD) technology is proposed to further reduce common mode interference.

[0005] EP 3372148 A1 discloses a system, method and device for reducing interference in signals transmitted by a physiological measurement device, such as an electrocardiogram device. The physiological measurement device can use a filter that uses coefficients to reduce time domain differences between response signals of the physiological measurement device. The coefficients can be derived during a calibration process in which a test signal is applied to each channel of the physiological measurement device to identify the channel with the slowest or most delayed response. When a monitor signal is then compiled from the response signals filtered with the coefficients, the timing differences between the response signals do not distort the monitor signal, making it more accurate for measurement purposes. [Means for solving the problem]

[0006] It is an object of the present invention to provide a physiological measurement device, a physiological measurement method and respective computer program products that allow an improved common-mode interference mitigation. Furthermore, it should allow a cost-effective implementation of the signal acquisition front-end without compromising the effectiveness of the common-mode interference mitigation.

[0007] In a first aspect of the present invention, a plurality of input channels, at least two of the input channels including a digital portion configured to process digital signals representative of analog signals present at the electrodes assigned to the respective channel; a signal combiner configured to calculate at least one vector signal from the at least two processed digital signals, the digital part of at least one input channel, preferably each input channel of the plurality of input channels, having a digital filter configured to filter the digital signal based on a set of filter coefficients; and a processor configured to calculate a set of filter coefficients based on the vector signal; A physiological measurement device is presented having:

[0008] In a further aspect of the invention there is provided a corresponding method, a computer program having program code means which, when executed on a computer, causes the computer to perform the steps of the methods disclosed herein, and a non-transitory computer readable recording medium storing a computer program which, when executed by a processor, causes the processor to perform the methods disclosed herein.

[0009] Preferred embodiments of the invention are defined in the dependent claims. It is to be understood that the claimed apparatus, method, computer program and medium have similar and / or identical preferred embodiments as the claimed system, particularly as defined in the dependent claims and disclosed herein.

[0010] The invention is based on the idea of ​​considering the characteristics of a vector signal in order to adapt the digital filters of the channels of the measuring device, rather than focusing on the intrinsic characteristics of the channel, such as the impedance. By adapting a set of coefficients of at least one digital filter of a channel based on the characteristics of the vector signal, common mode interference can be mitigated particularly well. As a result, the medical personnel operating the measuring device see a cleaner measurement curve (e.g. EEG or ECG curve) with less noise.

[0011] The use of digital filtering allows to compensate for time skew between digital signals of different channels. Such time skew is introduced by a cost-effective analog-to-digital converter (ADC) architecture, using a single ADC to subsequently digitize multiple analog signals related to different channels. In other words, the claimed invention allows to implement a cost-effective signal acquisition front-end without compromising the effectiveness of common mode mitigation. The claimed apparatus, method and program operate effectively so that the RLD technique or similar methods are not required to preserve good measurement results.

[0012] In an example implementation, calculating the set of filter coefficients may be based on the unfiltered digital signal or samples of that signal.

[0013] Furthermore, calculating a set of filter coefficients can be based on a current set of filter coefficients. When calculating one or more new sets of filter coefficients for one or more digital filters based on the unfiltered digital signal and the current set of filter coefficients, an improvement step is performed to adjust currently used filter coefficients. This improvement step may be performed during normal operation of the measurement device, for example during acquisition of an ECG, EEG or EMG curve of a patient. No separate calibration or use of a special measurement signal is required.

[0014] Preferably, the calculation of the set of filter coefficients is performed repeatedly for each newly available sample of the digital signal or for each block of multiple subsequent samples of the digital signal, so that during operation of the measurement device, the filter coefficients are continuously adjusted, while the common mode rejection behavior is gradually improved.

[0015] Said improving step can for example comprise performing an optimization. In other words, calculating a set of filter coefficients comprises performing an optimization step, the measures constituting a target function and the set of filter coefficients constituting input variables of this target function. Although the examples described herein mainly use a cost function to be minimized, the target function can also be a cost function and other kinds of target functions can also be used, for example a utility function that needs to be maximized.

[0016] The inventors have realized that measures, in particular statistical measure values, can represent characteristics of a vector signal that allow for distinguishing between desired physiological measurement signals and common-mode interference. To facilitate distinguishing between desired signals and common-mode interference from each other, the target function can comprise a weighted sum of multiple measures.

[0017] In order to exclude undesirable digital filter settings from the optimization solution space, in some embodiments, performing the optimization step may be subject to at least one constraint related to a set of filter coefficients, such constraints may include at least one equality constraint and / or at least one inequality constraint.

[0018] For example, to avoid a solution in which common-mode interference is filtered by a digital filter rather than removed by a signal combiner, the optimization step may be subject to at least one constraint related to a similarity measure characterizing the similarity of two different sets of filter coefficients.

[0019] Instead of, or in addition to, constraints relating to a similarity measure, the target function may have a similarity measure that characterizes the similarity of two different sets of filter coefficients.

[0020] Preferably, the processor is configured to calculate a plurality of vector signals from the filtered digital signals, the digital part of one channel including at most one digital filter, each digital filter therefore being specific to one channel but not necessarily to a particular vector signal.

[0021] In an exemplary implementation, the processor is configured to receive indicators from a backend and determine at least one measure and / or target function based on the received indicators. This backend may be part of the measurement device or may be included in a separate device. The indicators enable the measurement device to adapt to a special measurement situation that is automatically detected by the backend and / or determined by the backend based on human user input. The adaptation may include selecting a target function and / or configuring the target function based on a number of measures that are set. In particular, weights of a weighted sum of a number of measures are determined based on the indicators.

[0022] For example, the signal combiner is configured to calculate a vector signal from a further vector signal (x3). In an ECG application, the further vector signal corresponds to a Wilson Central Terminal (WCT). At least one vector signal calculated from the further vector signals corresponds to a lead associated with a chest electrode.

[0023] The measurement device may include a signal acquisition front-end including an analog input circuit with an analog filter, individual input channels with ADCs and digital filters, a signal combiner and a processor. Optionally, a back-end may be included in the measurement device. The measurement device may be an ECG, electroencephalography (EEG) or electromyography (EMG) device for medical (diagnostics, patient monitoring, etc.) or non-medical (fitness tracker, smart watch, etc.) applications.

[0024] Input signals present at the electrodes of the different channels are typically sensed (acquired or measured) from the patient during use of the device, i.e. a digital part is assigned to each channel and configured to process digital signals representative of the analog signals present at the electrodes sensed from the patient during use of the device. Thus, the measurement device is configured to acquire (receive or retrieve) input signals from the electrodes or sense input signals via the electrodes of the patient. [Brief description of the drawings]

[0025] These and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter. [Figure 1] FIG. 1 shows a block diagram of a physiological measurement device according to a first example. [Diagram 2] FIG. 2 shows a flow chart of a method of operating the apparatus of FIG. [Diagram 3] FIG. 3 shows an ECG measurement device according to a second example, and electrodes connected to this device. [Figure 4] FIG. 4 is a block diagram of the ECG signal acquisition content of the measurement device shown in FIG. [Diagram 5] FIG. 5 shows a block diagram of an ECG signal acquisition front-end according to a third example. [Figure 6] FIG. 6 shows a diagram of the signal in the measurement device over time. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0026] While the invention has been illustrated and described in detail in the drawings and foregoing description, such illustration and description is to be considered as illustrative or exemplary and not restrictive, i.e., the invention is not limited to the disclosed embodiments. Other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure, and the appended claims.

[0027] In the claims, the word "comprising" does not exclude other elements or steps, nor does it exclude a plurality of them. A single element or other unit may fulfill the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

[0028] The computer program may be stored / distributed on a suitable non-transitory medium, such as an optical storage medium or a solid-state medium, supplied together with or as part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless communication systems.

[0029] Any reference signs in the claims should not be construed as limiting the scope.

[0030] Electrophysiological measurements, such as electrocardiography, electroencephalography, and electromyography, work by measuring the voltages that appear between specific points on a patient due to physiological cellular activity. The signals of interest (wanted or desired signals) are usually small (in the millivolt to microvolt range) and typically involve only the potential difference between the locations of the electrodes. However, factors such as electromagnetic interference (e.g., from power lines or electrical devices in close proximity to or in contact with the patient) cause the patient's potential to change over time, causing each electrode to pick up a mixture of the differential mode physiological signal of interest and non-physiological, undesired common mode signals.

[0031] The difference between the positions of two electrodes, or more generally, any linear combination of the electrode potentials whose weighted coefficients sum to zero, is usually called a vector signal, or simply a "vector". In the ideal case, when the potential difference between the two electrodes is formed by calculating this vector signal, the common mode signal completely disappears.

[0032] However, in practical implementations, each electrode has a slightly different transfer function dictated by the quality of the patient / electrode interface, the characteristics of the cable between the electrode and the measurement device, component tolerances in the input circuitry of the measurement device, and the characteristics of the analog-to-digital conversion such as time skew (in the case of sequential sampling), and multiple analog-to-digital converters (ADCs) if multiple ADCs are used. This difference in transfer functions can cause a significant portion of the undesired common mode signal to appear in the vector as so-called common mode interference.

[0033] 1 shows a block diagram of a simple electrophysiological measurement device 11 with two electrodes 13. The measurement device 11 comprises a signal acquisition front end 15 and an analysis back end 17. The front end 15 is arranged to process analog signals E1, E2 input to the front end 15 via the electrodes 13 and to generate therefrom a vector signal x.

[0034] The backend 17 is operable to receive the vector signals x from the frontend 15 and to perform further operations from these vector signals, such as, for example, further signal analysis and / or signal classification, in particular in the case of an ECG device, rhythm classification. Furthermore, the backend 17 may include user interface elements, such as, for example, a screen and / or a keyboard, to display the vector signals x in their original or post-processed form or to display the results of the signal analysis or classification to a user of the device 11. The analysis backend 17 comprises a backend processor 21, which may include computer programs for performing the above-mentioned operations of the backend 17. Besides the illustrated example, a measurement device may be provided that does not include a backend 17. For example, the backend may be remote with respect to the measurement device.

[0035] The front end 15 has input channels 23 dedicated to each electrode 13. The exemplary device shown in Figure 1 has two input channels 23 connected to the electrodes 13. The disclosure of this patent application is not limited to this relatively simple example. It can be used with measurement devices having many more input channels 23, and thus accommodating many more connected electrodes 13.

[0036] Each input channel 23 includes an analog section having circuits for receiving an analog signal e1, e2 from a respective electrode 13 and for carrying out basic signal pre-processing of the respective analog signal. To this end, the analog section has a transfer function H i (s), where i denotes the channel number. The analog filter 25 may be designed as an anti-aliasing filter to prevent aliasing effects when digitizing the analog signal, and / or may filter each signal e i The analog filter 25 may be designed to adapt the voltage or energy range of the analog signal e output by the analog filter 25 to the input range of an analog-to-digital converter (ADC 27). hi , the corresponding digital signal d i , thereby forming an interface between the analog portion of each channel 23 and the digital portion of that channel 23. The digital portion of the channel 23 converts the digital signal d i The filtered digital signal d is fi The digital filter 29 includes a filter coefficient G i The filter may include a finite impulse response (FIR) filter defined by a set of

[0037] 1, one ADC 27 is provided for each channel 23. However, it is also possible to implement fewer ADCs 27 in the front end 15 than there are channels 23. In other words, at least one ADC may be shared between several channels 23. In a practical embodiment, one ADC may include an analog input multiplexer, the inputs of which are connected to the outputs of the analog filters 25 of the different channels 23, and the individual filtered analog signals e hi These signals e are sequentially sampled and digitized by a shared ADC. hi The sequential sampling and digitization is usually performed by sequentially transferring the analog signals e i Since the digital signals d i This creates a time skew between the

[0038] The front end 15 outputs at least two filtered digital signals d fi and a combiner 31 configured to combine the first filtered digital signal df2 from the second filtered digital signal df1, thereby obtaining a vector signal x1. In the example shown, the combiner 31 is adapted to subtract the second filtered digital signal df2 from the first filtered digital signal df1. Thus, the weights used to calculate the vector x1 are 1 and -1, respectively.

[0039] The signal processing front end 15 further comprises a front end processor 33. This processor 33 may include a programmable digital computer 35 including a memory device 37. A computer program is stored in the memory device 37 and programmed to cause the computer 35 to perform the processing operations described herein. Such a computer program may be stored in any storage device (flash storage, magnetic storage, etc.) that is not part of the processor 33. The program may be made available for downloading, for example, via a communication network such as the Internet. It is also noted that in some embodiments, signal processing operations performed by blocks in FIG. 1 shown outside the processor 33, in particular the operations performed by the digital parts of the input channel 23, may be performed by the computer 35 of the processor 33.

[0040] The front-end processor 33 is configured, e.g., programmed, to receive the vector signal x1 and to calculate, based on this vector signal, a set of filter coefficients G1, G2 of the digital filter 29. Furthermore, the processor 33 can calculate the set of filter coefficients based on the samples d1, d2 and the unfiltered digital signals output by the ADCs 27 of the respective channels 23. Moreover, the processor 33 is configured to iteratively calculate the set of coefficients of the filter 29, i.e., the calculation of the filter coefficients is further based on the current set of filter coefficients.

[0041] 2 illustrates a method 39 performed by the front-end processor 33. In particular, the program stored in the memory device 37 is programmed to perform the method 39.

[0042] After the start 41 of the method, the current set of filter coefficients G1,...,G N where N indicates the number of channels 23. In the example shown in Figure 1, the number of channels is two, so N=2.

[0043] Step 43 is followed by step 45, which reads a sequence of samples of the unfiltered digital signal, where d i is the sequence of samples of digital signal i, for i=1,...,N. The unfiltered digital signal d i is the digital signal present in each channel 23 that has not yet been processed by the digital filter 29. For example, the unfiltered digital signal corresponds to the signal generated by each ADC 27. Step 45 can take exactly one sample per channel 23 (online mode). Alternatively, step 45 can collect a sequence of samples for each channel 23 (batch mode). For simplicity, we will refer to one sample of a channel 23 or a sequence of samples of a channel as d i Therefore, in step 45, the values ​​d1,...,d N is obtained.

[0044] In step 47 of method 39, subsequent samples x of one or more output vectors are j are collected. The example of FIG. 1 has only one vector signal x1. However, in the general case, the method 39 collects M vector signals x1,...,x M We can consider the individual vector signals x j The number of samples collected for each sequence x j Preferably, the statistical measures are selected such that they are calculated in a rational manner.

[0045] Step 49 of method 39 selects a cost function CF based on indicators BI received from backend processor 17. Indicators BI include information determined by said backend processor 17, e.g. resulting from a long-term, high-level signal analysis, such as rhythm classification, or an automated analysis or classification performed by backend processor 17. Thus, indicators BI include information usable by processor 33 to appropriately select and / or configure the cost function CF.

[0046] As will be described in more detail below, the method 39 comprises: j The statistical properties of CF are used to determine whether or to what extent the vector signal is adversely affected by disturbances, such as common mode interference. Because the measurement device is used to process a wide range of normal and pathological waveforms, such as ECG waveforms, adapting the cost function CF to specific pathological patterns, such as tachycardia or ventricular fibrillation, can improve the performance of the method 39.

[0047] For example, in an ECG recorder, the components responsible for signal acquisition and pre-processing (front-end 15) are unaware of this long-term signal analysis. Thanks to the indicators BI received from the front-end processor, the cost function CF can be adapted to best distinguish the signal corresponding to the currently identified heart rhythm from common-mode interference sources. As an alternative to the dynamic cost function selection, adaptation or configuration of step 49, a fixed cost function CF can be used and step 49 can be omitted.

[0048] The cost function CF is a function of the vector signal x jIt is used to measure the statistical similarity of a vector signal to a desired signal, e.g. an ECG signal, or to a common-mode interference signal of other expected disturbances. Components of the cost function can include statistical measures of the (numerically calculated) derivative of said vector signal. This derivative takes into account the time structure, whereas measures such as kurtosis and skewness take into account the histogram, which do not change when the order of the samples is changed. As an example, a measure quantifying the similarity of the distribution of a vector signal to that of a sinusoid shows a high similarity when the input signal is a random permutation of sinusoidal samples, but the difference becomes clear when the derivative of the vector signal is considered.

[0049] In one embodiment, the values ​​of the vector signals used to calculate the statistical measures are first normalized, i.e. shifted so that their mean is zero and scaled so that their standard deviation is 1. Depending on the number of samples available, the normalization may be approximate. Normalization makes the calculated measures (and their derivatives) scale invariant.

[0050] Below, possible statistical measures that may be calculated based on the collected subsequent samples of the vector signal determined in step 47 are described.

[0051] Vector signal x j To measure the density of x j For example, for a vector signal x used in ECG applications, j Typically, except for some pathological patterns, most samples have a sparse density at least close to 0. However, common-mode signals usually have sinusoidal periodic or quasi-periodic patterns, so the signals are dense.

[0052] Vector signal x j To measure the total variation of x over time, we use the differential vector signal x jThe L1 norm of x can be used. The differential vector signal can be calculated by subtracting successive values ​​from each other (by subtracting each from its previous value). For example, the total variation of an ECG vector signal is low, whereas the total variation of a common mode interference signal is high.

[0053] To measure the amplitude, the L2 norm (Euclidean norm) or the L∞ norm may be used. The desired ECG signal has a limited amplitude, typically in the range of less than 10 mVp-p, while the amplitude of the interfering common mode signal is often larger than 10 mVp-p.

[0054] Moreover, the vector signal x j A measure related to the probability density distribution of σ may be used. Using the kurtosis of said probability density distribution as a cost function, vector signals, which usually have pronounced supergaussian / leptokurtic properties, can be distinguished from disturbing common-mode signals, which usually have subgaussian / platykurtic properties.

[0055] As an alternative to, or in addition to, kurtosis, a measure of negentropy (or an approximation thereof) calculated for the probability density distribution can be used as a cost function.

[0056] To measure the symmetry of this probability density distribution, a measure of the skewness of the probability density function can be used as a cost function. x j *exp(-x j 2 / 2) Asymmetric heuristics may be used that are more robust to outliers, such as:

[0057] One of the cost functions mentioned above is to optimize the set of filter coefficients G iAlternatively, the cost function CF can be a weighted sum of multiple cost functions.

number

[0058] Method 39 is a method for determining a set of filter coefficients G i The method further comprises a step 51 for defining one or more constraints to be used in optimizing cost function C. Such constraints, for example in the form of equalities or inequalities, place limits on the solution space, i.e. on the mathematical domain within which an optimization algorithm used in later steps of the method 39 can search for an optimum value of the cost function C.

[0059] For example, the coefficient G iAn equality constraint requiring the sum of x,y ...

[0060] Furthermore, the constraint is that during the search for the optimum point of the cost function CF, one or more digital filters 29 of different channels 23 may be too similar to each other and may cause the input signal e i This can occur in situations where the common mode signals are dominated by a single frequency (e.g., a power line frequency of 50 Hz or 60 Hz). The optimization algorithm can place zeros in the frequency response of each digital filter 29, so that, for example, two digital signals d i Vector signal x is calculated by subtracting j The common mode interference can be removed even before is calculated. Removing the common mode part by band-reject filtering instead of subtraction would not be the purpose of electrode equalization and would have a significant adverse effect on the difference signal of interest. This constraint can include a similarity measure calculated for the set of filter coefficients of each digital filter. Instead of or in addition to this similarity measure, an inequality-type gain constraint is used for frequencies that are in the range of the signal of interest to prevent the operation of the band-reject.

[0061] In addition to or instead of using constraints to prevent the method 39 from selecting similar filter coefficients for different filters 29, a similarity measure related to the similarity of the filters 29 may be included in the cost function C.

[0062] Constraints may be defined to limit the norm of some of the pulses of one or more filters 29 to a given value. Limiting this norm has the effect that the numerical stability of the filters 29 is improved by avoiding numerical overflow or saturation.

[0063] Based on the selected or configured target function C and the constraints defined in step 51, step 53 of the method 39 performs an optimization operation to obtain the unfiltered digital signal d i of samples (online mode) or a sequence of samples (batch mode), and a vector signal x j to the new filter coefficients G i,opt Calculate the set of new filter coefficients G i,opt constitute the variables of the optimization problem to be solved in step 53. The delay introduced by the digital filter 29 is determined by the coefficient G i Since the optimization step 51 can be fairly easily controlled (compared to analog filters) by appropriately selecting the set of i To find a solution that reduces any adverse effect that the above-mentioned time skew between the

[0064] When operating in online mode, known minimization algorithms are applied, such as stochastic gradient descent, in particular least mean squares (LMS) or recursive least squares (RLS). In batch mode, known minimization algorithms can be applied, such as deepest descent, Newton iteration or inner-point methods. However, step 53 is not limited to a particular type of optimization algorithm. Although the optimization problem is formulated here as a cost minimization problem, an equivalent utility maximization problem may be formulated in steps 49 and 51 and solved in step 53.

[0065] After the optimization run is completed in step 53, the obtained new filter coefficients G i,opt can be applied to the individual digital filters 29.

[0066] After step 55, the digital signal d i The method 39 returns to step 43 such that steps 43-55 are performed iteratively either for each sample (online mode) or for a sequence of subsequent samples (batch mode).

[0067] Although the principles of the measuring device 11 and method 39 described herein can be applied to any kind of electrophysiological measuring procedure, such as EEG or EMG, below a more detailed example related to ECG measurement is given. This example includes a free channel ECG measuring device 11, with channels corresponding to three limb electrodes (right arm RA, left arm LA, left leg LL). As shown in Fig. 3, the three electrodes LA, LA, LL correspond to analog electrical signals e1, e2, e3, respectively. This example can be extended by adding further channels and electrodes, for example six chest electrodes (V1,...,V6), as explained below.

[0068] Figure 4 shows a block diagram of the signal acquisition front end 15 of the three channel ECG device shown in Figure 3. This three channel front end 15 has the same structure as the front end 15 shown in Figure 1, but with the addition of an additional channel 23. Due to the higher number of channels, more than one vector signal x iis defined. In the ECG application described herein, the three vector signals x1, x2, x3 correspond to the three leads, referred to as "Lead I", "Lead II" and "WCT", respectively. WCT stands for Wilson Central Terminal, which is a virtual potential calculated as the average of LA, RA and LL. In the embodiment shown, the processor 33 simultaneously optimizes the two leads x1, x2, thereby optimizing the coefficients G of the filters 29 of all channels 23 used to calculate the respective vector signals x1, x2. i Even if at least one channel 23, for example the channel 23 associated with the input signal e2 and the electrode RA, is used to acquire multiple vector signals, only one channel filter 29 is assigned to that channel. In other words, the digital signal df of the same channel 23 is j Different vector signals x determined based on j share the digital filter 23 of that channel.

[0069] As can be seen from FIG. 4, the combiner 31 combines the individual vector signals x j In particular, x1 (Lead I) is calculated by subtracting the digital signal df2 associated with electrode RA from the digital signal df1 associated with electrode LA. Vector signal x2 is calculated by subtracting the digital signal df2 associated with electrode RA from the digital signal df3 associated with electrode LL.

[0070] Furthermore, for each vector x j It should be noted that it is not necessary to apply the optimization method 39 described above to the vector signal x3(WCT). In the embodiment shown, the vector signal x3(WCT) is excluded from the optimization, in particular the cost function CF is not calculated based on this vector signal x3.

[0071] In another example shown in FIG. 5, the measurement device 11 shown in FIG. 3 and FIG. 4 measures the electrical signals e input from the chest electrodes V1 to V6. i , i=4,...,M, a further vector signal x j , j=4,...,M. The front end 15 thus has at least one additional channel 23 assigned to one chest electrode. In the case of a nine channel ECG device, six additional channels 23 are provided for the individual chest electrodes V1 to V6. Each combining element 61 of the signal combiner 31 combines the filtered digital signal df i 5 shows only one of these additional channels 23 and one respective combining element 61. j is a further variable of the cost function CF and represents each unfiltered digital signal d i are the filter coefficients G i 2. No filtering is applied to the vector signal x3 related to the WCT. The processor 33 performs the optimization method 39 described above and shown in FIG. 2. However, in comparison with the example of FIG. 1, the filter coefficients G i There are many sets of vector signals x j Since the number of ,is large, the optimization problem becomes more complex.

[0072] 4 and 5, the signal combiner 31 may include a plurality of combining elements 61. At least two processed digital signals df i At least one vector signal x j A single combined element 61 may be used (e.g., to calculate vector signals x1, x2, x3 corresponding to lead I, lead II, and WCT) or multiple combined elements 61 may be combined, particularly in a cascade (e.g., to calculate vector signals x1, x2, x3 shown in FIG. 5 associated with the chest leads). j5 and the WCT combining element 61 shown at the bottom of FIG. 4). Combining and / or cascading combining elements 61 is particularly useful when different vector signals need to be calculated by the signal combiner 31 from a common vector signal, e.g. vector signal x3 corresponding to the WCT.

[0073] FIG. 6 shows how the method 39 improves the signal quality when performing an online optimization during the measurement process. More specifically, the diagram of FIG. 6 shows the voltage in millivolts of the vector signal over time (seconds). The vector signal obtained without applying the adaptive filter 29 is shown as a grey curve. The vector signal obtained when applying said filter 29 on the individual channels 23 is shown as a black curve in the diagram. As can be seen, the latter signal is superimposed with noise due to common mode interference at the beginning of the measurement (time t=0). During this measurement process, the method 39 optimizes the filter coefficient G of the digital filter 29. i As the set of s is successively adapted, the proportion of noise is gradually reduced.

[0074] In summary, the method 39 and respective devices 11, 15 described herein allow iteratively adapting the digital filters 29 present in the individual input channels 23 to reduce disturbances such as common mode interference. i is at least one vector signal x j The electrode-specific adaptive filters 29 are iteratively adjusted based on an optimization algorithm that uses a cost function that is constructed from statistical properties of at least one vector signal, e.g., L1 norm, total variation, kurtosis, skewness, or negentropy. The inventors have recognized that good common mode rejection can be obtained by minimizing the statistical similarity of the vector signal with the pattern of a normal common mode signal and / or maximizing the statistical similarity of at least one vector signal with the expected pattern of a desired signal, e.g., the pattern of an expected ECG signal.

[0075] The method disclosed in EP 3 372 148 A1 relies on device properties that are measured while the device is not actively making measurements (i.e. using test signals) and only compensates for effects that are measurable at this time. Effects that only occur when the device is connected to the patient and making measurements, such as the electrical behavior of the skin-electrode interface, cannot be compensated for in this way.

[0076] As explained above, in physiological measurements, the signal path of each electrode has a slightly different transfer function due to factors such as component tolerances, different external cables and electrical properties of the skin-electrode interface. If these different transfer functions are not compensated for, this will result in the common mode signal being only partially cancelled by forming a difference between the two electrode potentials. The part of the common mode signal that is not cancelled will appear in the differential signal and will prevent the analysis of the physiological signal for diagnostic purposes. Part of the asymmetry of the transfer function is specific to each device and can be compensated for by a calibration process (e.g. as disclosed in EP 3372148 A1) before the device is sold to a customer. However, any part of the asymmetry that is external to the device, such as different cable and skin-electrode interface properties, is not known at the time of manufacture and cannot be compensated for by a factory calibration. Furthermore, the asymmetry changes slowly over time as adhesive electrodes attached to a patient dry out or the local conductivity of the patient's skin changes. If too much common-mode signal leaks into the differential-mode vector, the diagnostic value of the resulting waveform is diminished: medical conditions may go undiagnosed or be misdiagnosed, or analysis of the signal (including automated analysis) may become impossible.

[0077] According to the present invention, in contrast to known solutions, in particular the solution disclosed in EP 3372148 A1, no test (or reference) signal is used, but the input signal is measured / sensed when the electrodes are actually attached to the patient and the device is actually in use. In this way, the disclosed device and method can automatically compensate for the asymmetry of the transfer function of the electrodes, without using a test (or reference) signal injected into the patient or the electrodes. In some embodiments, the difference in the statistical properties of the physiological signal and the normal common mode interference signal can be used to calculate or adapt the coefficients of the digital filter. The calculation and adaptation of the filter coefficients of the electrode-specific filters aims at minimizing or maximizing a cost function that reflects certain statistical properties of the vectors that allow to distinguish between physiological signals and non-physiological interference. This compensates for the difference in the input transfer functions between the electrodes and increases the extent to which the components of the non-physiological common mode signal are cancelled.

Claims

1. a plurality of input channels, at least two of the input channels including a digital portion configured to process digital signals representative of analog signals present at the electrodes assigned to the respective channel; and a signal combiner configured to calculate at least one vector signal from the at least two processed digital signals; A physiological measurement device having The digital portion of at least one input channel comprises: a digital filter configured to filter the digital signal based on a set of filter coefficients; and a processor configured to calculate the set of filter coefficients based on the vector signal; A physiological measuring device comprising:

2. 2. The apparatus of claim 1, wherein calculating the set of filter coefficients comprises calculating at least one measure, preferably a statistical measure, of blocks of subsequent time domain samples of the vector signal.

3. The apparatus of claim 1 or 2, wherein calculating the set of filter coefficients is further based on the unfiltered digital signal.

4. 4. The apparatus of claim 1, wherein calculating the set of filter coefficients is further based on a current set of filter coefficients, and wherein calculating the set of filter coefficients is repeated for each newly available sample of the digital signal or for each block of multiple subsequent samples of the digital signal.

5. 5. The apparatus according to claim 1, wherein calculating the set of filter coefficients comprises performing an optimization step, the measures constituting a target function and the set of filter coefficients constituting input variables of the target function.

6. The apparatus of claim 5 , wherein the target function comprises a weighted sum of multiple measures.

7. 7. The apparatus according to claim 5 or 6, wherein performing the optimization step is subject to at least one constraint relating to a set of filter coefficients, in particular comprising at least one equality constraint and / or at least one inequality constraint.

8. 8. The apparatus according to claim 5, wherein performing the optimization step is subject to at least one constraint related to a similarity measure characterizing the similarity of two different sets of filter coefficients.

9. 9. The apparatus of claim 5, wherein the target function comprises a similarity measure that characterizes the similarity of two different sets of filter coefficients.

10. 10. The device according to claim 1, wherein the digital part is configured to process digital signals present on electrodes assigned to the respective channels and representative of analog signals sensed from a patient during use of the device.

11. 11. The apparatus of claim 1, wherein the processor is configured to calculate a plurality of vector signals from the filtered digital signal, and the digital part of one channel includes at most one digital filter.

12. 12. The apparatus of claim 1, wherein the processor is configured to receive indicators from a backend and to determine the at least one measure and / or the target function based on the received indicators.

13. 13. Apparatus according to any one of the preceding claims, wherein the signal combiner (120) is arranged to calculate the vector signal from further vector signals.

14. In the physiological measurement method, processing at least two digital signals, each representing an analog signal present at an electrode assigned to a digital portion of a respective input channel of a plurality of input channels of a physiological measurement device, the processing comprising digitally filtering the digital signal of the respective input channel based on a set of filter coefficients; calculating at least one vector signal from the at least two processed digital signals; and calculating the set of filter coefficients based on the vector signal; A physiological measurement method comprising:

15. A computer program comprising program code means for causing a computer to carry out the steps of the method according to claim 14, when said computer program is run on said computer.

Images