Minimizing common mode interference in physiological measurement devices

Abstract

The invention relates to a physiological measuring device (11). In order to reduce common mode interference and / or to improve the characteristics of a digital input filter (31) of the measuring device (11) without impairing the common mode rejection capability of the input filter, it is proposed to calculate a set of filter coefficients for at least one digital input filter (31) associated with a specific input channel of the measuring device (11) by means of an optimization based on samples (c i ) of a plurality of input signals (s i ) of the measuring device (11) and at least one definition vector (v j ) describing a linear combination of samples (c i ) of at least two input channels, the samples corresponding to test signals (TS) applied to the plurality of input channels of the measuring device (11).

Description

Technical Field

[0001] This invention relates to physiological measurement devices, particularly electrophysiological measurement devices, such as electrocardiogram (ECG) devices or electroencephalogram (EEG) devices. Background Technology

[0002] Physiological measurement devices typically present measurement results to their users (e.g., medical personnel) in the form of differential values ​​(often referred to as "vectors," or more specifically, "leads" in the context of ECG or EEG). Differential values ​​are derived from the differences in physical quantities measured at different points on a patient's body. The measurement device is designed to suppress common-mode components of signals measured at different points as well as possible.

[0003] WO 2018 / 162365 describes a physiological measurement device comprising digital filters arranged at the outputs of different analog-to-digital converters (ADCs) associated with different channels of the measurement device. The filters are calibrated to reduce common-mode interference. The corresponding calibration method includes determining a reference input channel and minimizing the time-domain difference between the reference input channel and each input channel other than the reference input channel. Summary of the Invention

[0004] The purpose of this invention is to further reduce common-mode interference and / or improve the characteristics of digital filters without compromising the common-mode rejection capability of the input filter.

[0005] In a first aspect of the invention, a method for minimizing common-mode interference in a physiological measurement device is proposed, the method comprising: receiving samples of a plurality of input signals of the measurement device, the samples corresponding to test signals applied to a plurality of input channels of the measurement device; inputting at least one defined vector describing a linear combination of samples of at least two input channels; optimizing a computed metric based on at least one vector signal, the vector signal being based on the samples of at least two of the received input signals and the defined vector; and obtaining at least one set of filter coefficients for at least one digital input filter associated with a particular input channel based on the optimization. By optimizing the metric not only based on samples of the input signals but also based on the defined vector, the selection of a reference signal can be eliminated, thereby generating multiple sets of filter coefficients that are well-suited to multiple individual digital input filters. This allows for good mitigation of common-mode interference and has a particularly low adverse effect on the quality of the desired (e.g., differential) components of the vector signal.

[0006] This method can be a calibration method for calibrating measuring equipment. It can be run during the final production testing of the measuring equipment. However, the method is not limited to a single use case. For example, it can be used to perform self-tests or self-calibrations on the measuring equipment. It can also be used in routine maintenance tasks, etc.

[0007] In typical applications, the measurement device to be calibrated has multiple leads corresponding to multiple vector signals. Therefore, multiple defined vectors can be input, and at least one set of filter coefficients can be obtained for a specific vector. In other words, different sets of filter coefficients can be used for the same input channel but for different vector signals. In an embodiment, a dedicated set of filter coefficients, relating to all filters associated with the channel used by a particular vector signal, can be assigned to that vector signal. Using different sets of filter coefficients for different vector signals not only allows for improved common-mode interference reduction but also simplifies optimization in terms of computational complexity.

[0008] Another approach to maintaining low optimization computational complexity involves breaking down the optimization into multiple optimization steps. Specifically, optimizing the metric may include multiple optimization steps, with a set of filter coefficients obtained in one step remaining unchanged for subsequent optimization steps.

[0009] Since the optimization is based on both samples of the input signal and the defined vector, constraints can be defined to control the characteristics of at least one input filter that are not related to or directly related to common-mode interference reduction. For example, the optimization can be subject to at least one linear equality constraint with respect to the filter coefficients. Such a linear equality constraint can be applied to obtain a certain DC gain or a certain gain at the Nyquist frequency.

[0010] The method may include: inputting the DC gain value of at least one input channel and determining a corresponding linear equality constraint based on the DC gain value. When the DC gain of one, more, or all digital input filters is set based on the input DC gain value, differences between channels can be compensated. In an embodiment, the DC gain value of one input channel is input, and the corresponding linear equality constraint can be determined based on this single DC gain value. The DC gain value can be determined by a separate measurement process performed before or in parallel with this method. Applying at least one linear equality constraint, for example, based on the DC gain value, prevents the optimization operation from reaching a trivial solution that minimizes the objective function by simply setting all filter coefficients to zero.

[0011] Furthermore, the optimization may be subject to at least one nonlinear inequality constraint used to limit the gain of an input filter at a given frequency. Such constraints (one or more) allow for at least some control over the frequency response of the digital input filter. These constraints can be defined such that unwanted frequency responses, for example, in parts of the analog front-end, can be at least partially compensated.

[0012] The optimization can also be subject to at least one boundary constraint on the filter coefficients. Such constraints limiting the range of filter coefficients can prevent numerical overflow when they are stored in a specific digital format (e.g., a floating-point format or a fixed-point format). Such constraints can also limit the gain of a particular input filter at a certain frequency to, for example, avoid amplification of unimportant frequency ranges.

[0013] The samples can be received at a calibration sampling frequency increased compared to the operating sampling frequency applied by the measuring device to perform the physiological measurements, and / or the calibration resolution of the received samples can be different from (preferably less than) the measurement resolution of the samples received to perform the physiological measurements. Such adjusted sampling frequencies and / or resolutions are well-suited for calibration because higher frequencies can be accounted for more accurately, while lower resolutions are acceptable during calibration because the amplitude of the test signal can be better controlled than the amplitude of the measurement signal captured during normal operation of the measuring device.

[0014] Optimizing the metric may include minimizing a penalty function for output samples corresponding to at least one vector signal calculated from the vector, based on the input signal filtered using the at least one input filter. Vector signals corresponding to differential measurements are often also referred to as "vectors". Furthermore, the test signal applied during calibration can be the same across all input channels. Therefore, a purely common-mode signal is applied to the channels of the measurement device. Minimizing the norm of at least one vector signal then corresponds to maximizing common-mode reduction. Output samples may correspond to a single, multiple, or all vector signals supported by the measurement device. When multiple sets of dedicated filter coefficients are used for an individual vector signal or for a predefined group of multiple vector signals, the output samples may correspond to the respective vector signal or the respective group of vector signals.

[0015] Optimization can include simultaneously minimizing multiple different penalty functions of the output samples or minimizing a scalar objective function of different penalty functions. In other words, optimization can be multi-objective optimization.

[0016] At least one penalty function may be a convex or quasi-convex penalty function, preferably a norm. For example, the l-infinity (maximum) norm and another norm may be maximized simultaneously or based on a scalar objective function, preferably the l2 (Euclidean) norm. At least one penalty function may include the Huber loss function.

[0017] Optimization can include minimizing scalar objective functions or other penalty functions of different norms to reduce the computational complexity of the optimization. Composing an objective function with several different norms or penalty functions allows for the achievement of multiple objectives or trade-offs between different objectives. For example, using a weighted sum of the l-infinity norm and the l-2 norm will minimize not only the maximum deviation (quantized by the l-infinity norm) but also the energy content of the error signal (quantized by the l2 norm), given a balance of trade-offs given by the weights.

[0018] In another aspect of the invention, a computer program is proposed. The computer program may include program code programmed to cause a computer (e.g., a processing unit of a device) to perform the method. The computer program may be part of a computer program product (such as a computer-readable non-transient recording medium including the program). The computer may include a processing unit of a calibration device. The computer (particularly the processing unit of a calibration device) may include I / O circuitry configured to perform input / output (I / O) operations to access data present in the measuring device and / or control the measuring device to perform the methods described herein.

[0019] In another aspect of the invention, a calibration device for minimizing common-mode interference in a physiological measurement device is proposed. The calibration device includes a processing unit configured to: receive samples of a plurality of input signals from the measurement device, the samples corresponding to test signals applied to a plurality of input channels of the measurement device; input at least one defined vector describing a linear combination of samples from at least two input channels; optimize a calculated metric based on at least one vector signal, the vector signal being based on the samples of at least two of the received input signals and the defined vector; and obtain at least one set of filter coefficients for at least one digital input filter associated with a particular input channel based on the optimization. The calibration device can be implemented as a separate device with respect to the physiological measurement device. Such a separate measurement device can be used to produce final tests. Alternatively, the calibration device can be adapted to be integrated into the physiological measurement device to provide self-testing and / or self-calibration capabilities to the physiological measurement device. In any case, the calibration device (particularly its processing unit) can be configured to perform the methods described above.

[0020] In another aspect of the invention, a physiological measurement device is proposed, the device comprising: multiple input channels for processing multiple (typically analog) input signals; a memory device configured to store multiple sets of filter coefficients, at least one set of filter coefficients being associated with at least one digital input filter of the measurement device, the input filter being configured to filter a sample of an input signal to obtain a filtered sample of the input signal; and a processing unit configured to compute multiple vector signals, one of the multiple vector signals being computed based on a defined vector from the multiple filtered input signals; wherein the memory device is configured to store different sets of filter coefficients associated with one (in other words, the same) input channel, and the processing unit is configured to select one set of filter coefficients from the different sets of filter coefficients for computing a specific vector signal. In an embodiment, multiple vector signals may be defined, and the memory device is configured to store different sets of filter coefficients for one input channel, wherein at least some (preferably each set) of the stored multiple sets of filter coefficients are associated with different vector signals among the multiple vector signals.

[0021] In another aspect of the invention, a physiological measurement method is proposed, the method comprising: processing a plurality of (preferably analog) input signals via a plurality of input channels; storing a plurality of sets of filter coefficients in a memory device, at least one set of filter coefficients being associated with at least one digital input filter of the measurement device, the input filter being configured to filter a sample of an input signal to obtain a filtered sample of the input signal; and calculating a plurality of vector signals, one of the vector signals being calculated based on a defined vector from the plurality of filtered input signals; wherein different sets of filter coefficients associated with an input channel are stored in the memory device, and a set of filter coefficients from the different sets of filter coefficients is selected for calculating a specific vector signal.

[0022] In yet another aspect of the invention, a computer program corresponding to a measurement method is provided, and a non-transient computer-readable recording medium in which a computer program product is stored, the computer program including program code units for causing the computer to perform the steps of the method disclosed herein when the computer program is executed on a computer, the computer program product causing the execution of the method disclosed herein when run by a processor.

[0023] It should be understood that the claimed methods, systems, computer programs, and media have similar and / or identical preferred embodiments to the claimed systems and, as particularly, those defined in the dependent claims and disclosed herein. Attached Figure Description

[0024] These and other aspects of the invention will become apparent and elucidated with reference to one or more embodiments described below. In the following figures: Figure 1 A block diagram of the physiological measurement and calibration equipment is shown; Figure 2 An exemplary measuring device is shown; Figure 3 The signal flow within the measuring device according to an embodiment is shown; Figure 4 A flowchart of the calibration method is shown; and Figure 5 Additional signal flows within the measuring device according to an embodiment are shown. Detailed Implementation

[0025] Figure 1 A physiological measurement device 11 connected to calibration device 13 is shown. The physiological measurement device 11 has multiple channels. For simplicity, in... Figure 1 Only three channels are shown, but typically there can be N channels. Each channel has a function to input the analog input signal s. i Analog input section 15 (i = 1, ..., N). The input section 15 of the individual channel is connected to the analog front end 17. The analog front end 17 includes analog input circuitry 19 associated with the individual channel. The analog input circuitry 19 may include one or more amplifiers, impedance converters, analog filters, etc. The input circuitry of the individual channel is connected to a multi-channel analog-to-digital converter 21. The multi-channel analog-to-digital converter 21 is configured to process the individual input signal s preprocessed by the individual analog input circuitry 19. i Converted into an input signal s representing an individual channel. i Sample c i(i = 1, ..., N). In the illustrated embodiment, the multi-channel analog-to-digital converter 21 includes a plurality of analog-to-digital converters (ADCs 23), wherein each of these ADCs is associated with a specific channel. In another embodiment, a single ADC 23 may be associated with multiple channels, and the multi-channel ADC 21 may include a multiplexer to selectively connect a channel to a corresponding ADC. When using a multiplexer, a single ADC 33 can be used to convert the input signal of each channel one after another. By using a multiplexer, it is possible to sequentially sample several input channels using a single ADC. The calibration method described herein can compensate for timing skew that may exist in such a multiplexing setup. In a possible implementation (e.g., an ECG device), two ADCs may be present, each capable of sequentially sampling, for example, five ECG input channels (ten input channels in total).

[0026] The digital interface of the multi-channel analog-to-digital converter 21 is connected to the processing unit 25 of the measuring device 11, enabling individual input signals s to be processed. i Sample c i It is forwarded to processing unit 25.

[0027] The processing unit 25 may include a first processor 27 and a first memory device 29 constituting a computing device capable of operating to perform various processing steps required for measurements performed by the physiological measurement device 11. The processing unit 25 includes a plurality of digital input filters 31. The number of input filters may correspond to the number of channels N, and one digital input filter 31 may be assigned to a specific channel. Individual input filters 31 may be implemented having a column vector x i (i 1, ..., N) represents the finite impulse response (FIR) filter coefficients. Individual vector x i The length K corresponds to the number of filter coefficients in each digital filter 31. Digital filters 31 are also called input channel specific digital filters (ICSFs) because they are applied to input signals s... i Sample c i The differential output signal is applied to these samples before being derived.

[0028] Note that this disclosure is not limited to a specific filter topology. Other filters, such as infinite impulse response (IIR) filters, can be applied instead of FIR filters. Filter 31 can be implemented in software, i.e., a computer program stored in memory device 29 can be programmed to cause the first processor 27 to base its operation on the filter coefficients x. i To run the filter algorithm. Filter coefficients x iIt can be stored in a programmable and non-volatile segment of the first memory device 29. The calibration device 13 can access the processing unit 25, enabling the calibration device 13 to read sample x. i And for the filter coefficients x i To do programming.

[0029] Processing unit 25 can operate on samples c of individual input signals filtered by individual digital input filter 31 i To calculate one or more output signals y j (j = 1, ..., M). This can be achieved by forming filtered samples with multiple channels. A linear combination of at least two input signals s i Perform differential measurements to calculate the output signal y j (Also known as a vector signal or simply a "vector"). This linear combination can be generated by a specific output signal y. j Definition vector v j Specify, where, And 1 represents a column vector of 1, that is, 1 T = [1, ..., 1], and " " represents the scalar product. The output signal is..." Differential measurements are specified by linear combinations of defined vectors. Differential measurements can be based on two or more channels. In some applications such as ECG or EEG, vector signals are often also referred to as "leads".

[0030] In physiology, voltage measurements are almost always differential. Other physical quantities, such as pressure or temperature, are typically measured as absolute values ​​(e.g., temperature of 37 °C or arterial pressure of 100 mmHg), but in some cases, differential measurements may be of interest (cardiac output can be measured by thermal dilution, which requires measuring the temperature difference between two points in the blood vessel). Therefore, the techniques described herein can also be applied to differential measurements that are not voltage measurements.

[0031] For the sake of simplicity, Figure 1 Only one digital filter 31 is shown for each input channel. However, multiple digital filters 31 can be provided for one input channel i, where each digital filter processes sample c. i Different digital filters 31 can be associated with different output signals. Therefore, the filtered signal... Specific variants can be used to calculate a certain output signal y j .

[0032] The measuring device 11 may include an output device 33 coupled to the processing unit 25, such that the output device 33 can output an output signal y. jFor example, output device 33 may include a device for visualizing the output signal y. j Components, such as a display or printer, may be present for outputting at least one output signal y by output device 33. j At least one output signal y was processed previously. j One or more post-processing steps.

[0033] The physiological measurement device 11 can be an electrophysiological measurement device, such as an ECG or EEG. In particular, various configurations regarding the number of channels N and the number of output signals M are possible.

[0034] A relatively simple example is an ECG or EEG device with a single output signal (single vector) and two electrodes connected to two input sections 15 (two channels). The only defined vector required for such a measurement device 11 could be v1 = [1, -1].

[0035] according to Figure 2 Another example shown provides a 12-lead ECG with nine measuring electrodes connected to N=9 input sections 15 of the measuring device 11. Thus, such a measuring device has nine channels. The electrodes may include three limb electrodes (right arm RA, left arm LA, left leg LL) and six chest electrodes (V1-V6), which are connected to the nine different input sections 15 of the physiological measuring device 11 via measuring cables 35.

[0036] Output vector signals y1, ..., y 12 It can be defined based on the twelve leads visualized to medical personnel using output device 33. The leads and corresponding vector signals include lead I (LA - RA), lead II (LL - RA), lead III (LL - LA), ... , aVF (LL - (RA + LA)) and (Vn - WCT); n = 1, ..., 6. WCT corresponds to the Wilson center terminal voltage, which is approximately (RA + LA + LL) / 3.

[0037] This disclosure is not limited to the two exemplary measurement devices described above. For example, 15-lead or 18-lead ECGs can also be provided based on this disclosure. High-density EEGs with 32 or more channels can also be provided based on this disclosure.

[0038] Figure 3An example is shown in which an output signal y1 is generated by subtracting another filtered input signal s2 from a filtered input signal s1 (defined as vector [1,-1]). Therefore, the output signal y1 corresponds to the signal obtained by differential measurement based on the two input signals s1 and s2. When performing physiological measurements, the input signals s1 and s2 are subject to common-mode interference. Ideally, subtracting the two signals s1 and s2 from each other would completely eliminate the common-mode component. However, in a practical embodiment of the measurement device 11, the signal y1 includes a common-mode component to some extent due to differences in the electrical characteristics of components 19, 21, and 23 of the signal path associated with the individual channel.

[0039] In electrophysiology, the voltage generated between points on a patient's surface due to the activity of muscle or nerve cells is measured. The signal of interest is typically in the range of microvolts (EEG) to millivolts (ECG), while the common-mode component of the signal can be tens or hundreds of millivolts. In a medical context, there are many possible sources of interference (including common-mode interference), spanning other medical instrument components, electronic communication or networking devices, and power line noise. To make the output signal y... j For diagnostic and therapeutic purposes, the electrophysiological measurement device 11 needs to have a high common-mode rejection ratio (CMRR). While there are standardized test procedures for determining CMRR in the 50 Hz or 60 Hz power line frequency range, common-mode interference should also be mitigated at other frequencies. Frequencies where common-mode interference should be mitigated may include frequencies outside the ECG / EEG band, particularly if the recording device performs some form of out-of-band processing (e.g., for electrode impedance measurements or ECG pacemaker pulse detection).

[0040] To mitigate common-mode interference, the digital filter 31 can be designed to make the corresponding output signal y j Minimize the common-mode portion. For this purpose, it is possible to apply... Figure 1 The calibration device 13 shown determines the filter coefficients x. i This causes the individual output signal y to be... j The common-mode component is minimized. The calibration device 13 includes an additional processing unit 37, which includes a second processor 39 and a second memory device 41. The additional processing unit 37 includes input / output (I / O) circuitry configured to couple the additional processing unit 37 to the test signal generator 33 of the calibration device 13. The I / O circuitry is configured to access the first memory device 29, specifically its non-volatile portion, enabling the calibration device 13 to read sample x. i And for the filter coefficients x iProgramming is performed. The I / O circuitry is configured to perform I / O operations to access data present in the measuring device 11 and / or control the measuring device to perform the calibration method 45 described herein. An additional processing unit 37 is coupled to the test signal generator 33 of the calibration device 13. The processing unit 37 can control the test signal generator 43 via the I / O circuitry to generate a test signal TS fed to the individual input section 15 of the calibration device during the execution of the calibration method. A computer including the additional processing unit 37 can be programmed to perform the method described herein for calibrating the physiological measuring device 13. The corresponding program can be stored, for example, in the first memory device 13.

[0041] Figure 4 A method for calibrating the physiological measurement device 11 is illustrated. Method 45 can be performed by the calibration device 13. Therefore, an additional processing unit 37 can be configured to run method 45. In particular, the second memory device 41 may include a computer program programmed to run method 45 when run on a computer (e.g., the additional processing unit 37). Alternatively, at least some operations of method 45 can be performed by the measurement device 11 (specifically, the processing unit 25).

[0042] Method 45 can be run as part of the final production test related to measuring device 11. During the final production test or typically during calibration, calibration device 13 is connected to measuring device 11. An electrical connection configured to apply the test signal TS to input unit 15 may exist between test signal generator 43 and input unit 15, and a connection configured to read sample c may exist between processing unit 37 of calibration device and processing unit 25 of measuring device. i And used to write back the filter coefficients x calculated by calibration device 13 as described below. i Communication connection.

[0043] Following the start 47 of method 45, a test signal TS is applied to the input unit 15 in step 49. For this purpose, a separate processing unit 37 can control the test signal generator 43 to output the test signal TS. The test signal should cover the frequency range of interest of the measuring device, and the amplitude should be large enough to cover a significant portion of the input range of the measuring device ADC 23 in order to minimize errors caused by quantization. Different types of signals can then be used, such as chirped or sweep signals, pulse train signals, periodic signals with sharp transitions (such as square waves or sawtooth waves), or even wide or band-limited noise.

[0044] In step 51, method 45 inputs or otherwise determines one or more definition vectors v1, ..., v M This is one or more defined vectors v1, ..., vM This can correspond to the leads supported by the measuring device 11. For example... Figure 4 As shown, the defined vectors can be arranged in the matrix V=[v1, ..., v M ]middle.

[0045] Step 53 of method 45: Input at least one DC gain value DC for a specific channel i. i The DC gain value can be determined in a separate measurement procedure (which can be performed during final production testing). i And describe the gain at frequency 0 within a specific channel (specifically within the analog circuit 19 of that channel).

[0046] In step 55 of method 45, calibration device 13 configures the sampling frequency f of ADC 23. SAMP And resolution res. Compared with the sampling frequency used during normal operation of the measuring device 11, the sampling frequency f can be increased. SAMP The resolution res used during calibration method 45 can be reduced compared to the resolution used during normal operation. For example, the sampling frequency during calibration can be at least 2000 Hz, where the ECG recorder typically operates at a sampling frequency between 500 Hz and 2000 Hz when performing ECG measurements. To configure a special sampling frequency and resolution during calibration, the measuring equipment (specifically the multi-channel analog-to-digital converter 21 and / or processing unit 25) can be switched to a special operating mode.

[0047] When the test signal TS is applied, the measurement device 11 records the unfiltered ADC samples c of all input channels. i In step 57, method 45 receives samples c1, ..., c of the test signal TS, which are associated with individual channels and represent distortions caused by the analog front end 17. N .like Figure 4 As shown, it can be represented by an L×N matrix C = [c1, ..., c N [] represents a sample. L is the number of samples recorded.

[0048] In step 59, the recorded samples C and the defined vector V can be used to establish an optimization problem that minimizes the objective function TF. The objective function TF may depend on at least some (preferably all) filter coefficients x. i The output vector X. The filter coefficients can be represented by the column vector X = [x1] T , ..., x N T ] T It means that it has The length of each element, K, corresponds to the number of filter coefficients in each channel. Therefore, all filters 31 in the illustrated embodiment have the same number of coefficients. However, individual filters 31 may also differ in the number of coefficients, and the calibration method 45 may be adjusted accordingly.

[0049] Typically, when considering the case where each channel has more than one digital filter 31, the length of the coefficient vector X is 1…imax. The minimum value of imax represents the case where each input channel has only one matched filter, and the maximum value of imax represents the case where each input channel has one matched filter and a possible output vector. Without loss of generality, we will assume imax = N below.

[0050] The objective function TF can include a convex or quasi-convex penalty function, particularly a norm, preferably the L-infinity norm (maximum norm). Other norms (such as the L2 norm (Euclidean norm)) can also be used. Due to the number of samples, the optimization problem to be solved in step 59 is relatively large, but it can be handled using existing solvers for optimization problems.

[0051] The independent variable of the objective function can include unfiltered sample c. i Define vector v j and filter coefficients x i Filter coefficients x i The target variable is determined during optimization. Based on the optimization, the optimal or near-optimal values ​​of the filter coefficients are obtained as the result. Optimization may involve applying a numerical solver to the optimization problem.

[0052] The independent variables of the objective function can include the column vectors of all samples of the vector signal, and can be modeled in the following form: in, To(ci) is a Toeplitz matrix with ci elements and a size of (L-K+1, K). The first column contains the last (L-K+1) elements of ci; the first row contains the first K elements of ci in reverse order. For an example of c = [1 2 3 4 5 6 7 8 9]T, L = 9, K = 4: CA = [To(c1) To(c2) … To(cN)] is an extended input sample matrix used to facilitate the computation of all output samples. CA can also be constructed from a Hankel matrix; this only changes the order of the elements, thus changing the order of the matched filter coefficients in the output.

[0053] It is a Hadamard (element-wise) matrix product. Kronecker product : Multiply the sample matrix of the input channel samples used only in vector i by the appropriate sign and coefficients. Note the use of Hadamard / element-wise product and Kronecker product.

[0054] It is the column vector of the output samples of the vector signal i. Based on the objective function's independent variable, the optimization can be formulated as a multi-objective optimization: minimizing the l-infinity norm of the output sample vector, and also minimizing another norm (or a similar norm's penalty function, such as the Huber loss function, which is robust to outliers of the l1 norm and is also a smoothing function (the l1 norm is not)) (usually the l2 norm of the output sample vector (with lower priority)). Regarding X, minimize Typically, if the subsequent processing of differential signals is sensitive to outliers (e.g., pacing pulse detection in ECG) rather than signal energy, then the optimization problem should minimize the l-infinity norm (i.e., the maximum absolute deviation from zero when a pure common-mode signal is applied to the input). However, it is possible to sacrifice some l-infinity norm optimality to find a solution with a small l2-norm (or a norm other than l-infinity). Depending on the solver used, the solution for l-infinity norm optimization may already be biased towards minimizing other norms. If this is not the case, then the multi-objective optimization problem can be scalarized as: Regarding X, minimize Alternatively, it can be divided into first l-infinity norm optimization, and then minimized for different norms while keeping the l-infinity norm unchanged (possibly allowing some deviation from the value of the first optimization to simplify the work of the optimization algorithm).

[0055] Optimization problems can include constraints. For example, they can be applied in the form of... Linear equality constraints are applied to compensate for differences in DC gain between individual channels. This can be based on the DC gain value obtained in step 53. i To derive value b i Similarly, linear equality constraints can be used to constrain the gain of at least one digital filter 31 at the Nyquist frequency.

[0056] At least one nonlinear equality constraint can be applied to the optimization problem, for example, to constrain the gain of digital filter 31 to be equal to the value b.i This constraint has This type of constraint can also be formulated using a vector of real-valued complex exponents instead of sine and cosine values. In this case, a scalar product can be used to produce a real-valued output.

[0057] Furthermore, nonlinear inequality constraints can be applied to the optimization problem to, for example, constrain the gain of at least one of the digital filters in digital filters 31 at a given frequency to be equal to or less than the value b. i This constraint has This type of constraint can also be formulated using a vector of real-valued complex exponents instead of sine and cosine values. In this case, a scalar product can be used to produce a real-valued output.

[0058] Boundary constraints, which can be applied as special cases of linear inequality constraints, can be used to constrain the values ​​of the coefficients of the individual digital filter 31. These constraints have... These constraints can be used to prevent numerical overflow when storing coefficients and when calculating filter outputs. Furthermore, the high-pass behavior of individual filter 31 may be limited.

[0059] It can be applied based on the coefficient x i Additional constraints related to the norm of the calculation The norm can be the l2 norm of the filter coefficients. This constraint limits the average gain and high-pass behavior of the filter. This constraint corresponds to a nonlinear convex inequality constraint.

[0060] This disclosure is not limited to the constraints described above applied to optimization problems. It may also include other types of constraints, such as constraints on at least some coefficient vectors x. i The value of is reduced to have a smaller amplitude, which forces the filter to be closer to a minimum phase (and minimum delay) system.

[0061] It is possible to modify the optimization problem to, for example, reduce its computational complexity. One approach is to break the optimization problem down into subproblems that are generally smaller than the original problem. For example, in ECG applications, calibration can first compute the matched filter coefficients of the vector formed by the limb leads (I, II, III, aVR, aVL, aVF), and then, keeping the first set of coefficients unchanged, compute the coefficients for each chest electrode individually (since vectors with two chest electrodes are not used in practice, it is possible to compute the chest electrode filter coefficients individually for each chest electrode). This results in seven smaller optimization problems.

[0062] Alternatively, the calculation of the Wilson central terminal (WCT) vector signal (RA+LA+LL) / 3 can use its own coefficients. In this case, the filter coefficients of the channels associated with the chest electrodes V1, ..., V6 must be matched simultaneously (i.e., during an optimization run) because they all require the WCT. Calculating the filter coefficients specifically for the WCT vector signal breaks the original problem into two smaller subproblems—an optimization run for the limb lead vector signals RA, LA, LL, and an optimization run for the chest lead vectors V1, ..., V6.

[0063] When multiple matched filters are used for each electrode, the optimization problem can be divided into completely independent subproblems, each of which can be solved independently without adversely affecting the overall optimality of the total solution.

[0064] After one or more optimization runs are completed, step 61 of method 45 obtains the filter coefficients X of the individual digital filter.

[0065] In step 63, the obtained filter coefficients X can be stored in an appropriate storage area of ​​the first memory device 29 of the measuring device.

[0066] In step 65, method 45 terminates.

[0067] In one embodiment, each channel has exactly one digital filter 31, and the filter samples generated by these filters 31 Used to calculate each output signal y corresponding to a specific vector or lead. j .exist Figure 5 In the different embodiments shown, at least one channel has at least two sets of filter coefficients corresponding to different digital filters 31a, 31b. These filters can be used to calculate one or more specific output values. Figure 5 In the example shown, digital filter 31a is used to calculate the output signal y1, and digital filter 31b is used to calculate the output signal y2. The two digital filters 31a and 31b are configured to filter the input signal s2 of one channel. In this embodiment, each output signal y... j It has a set of dedicated digital filters 31. In other words, it is designed for the output signal y. j The corresponding specific vector yields a set of filter coefficients corresponding to the individual digital filter 31.

[0068] In summary, this calibration method obtains the filter coefficients of the digital filter 31 by optimizing an objective function based on samples of the test signals applied to the input units 15 of different channels and defining vectors. This allows for direct determination of the filter coefficients without selecting a reference signal. Furthermore, constraints related to the characteristics of the filter 31, which are unrelated to common-mode interference mitigation, can be easily incorporated into the calibration method. The calibration method 45 can be used in conjunction with the measurement device 11, which performs digital calibration on the sampled input signal s. i Filtering is performed, and then a vector signal is calculated based on the filtered samples corresponding to the filtered input signal. Furthermore, a measuring device 11 comprising one or more digital input filters 31 associated with a particular channel is described. The digital input filters 31 associated with the same channel can be associated with different vector signals.

[0069] Although the invention has been illustrated and described in detail in the accompanying drawings and the foregoing description, such illustrations and descriptions should be considered illustrative or exemplary, and not restrictive; the invention is not limited to the disclosed embodiments. Those skilled in the art, through studying the drawings, the disclosure, and the claims, will be able to understand and implement other variations of the disclosed embodiments in practicing the claimed invention.

[0070] In the claims, the word "comprising" does not exclude other elements or steps, and the words "a" or "an" do not exclude a plurality. A single element or other unit can perform the functions of several items recited in the claims. Although certain measures are recited in different dependent claims, this does not indicate that combinations of these measures cannot be advantageously used.

[0071] Computer programs can be stored / distributed on suitable non-transient media, such as optical storage media or solid-state media supplied together with or as part of other hardware, but can also be distributed in other forms, such as via the Internet or other wired or wireless telecommunications systems.

[0072] No reference numerals in the claims should be construed as limiting the scope.

Claims

1. A method (45) for minimizing common-mode interference in a physiological measurement device (11), the method (45) comprising: Receive (57) multiple input signals (s) from the measuring device (11) i ) samples (c i The sample corresponds to the test signal (TS) applied to multiple input channels of the measuring device (11). Input (51) describes samples (c) from at least two input channels. i At least one defined vector (v) of a linear combination of ) j ); The metric calculated based on at least one vector signal is optimized (59) based on samples (c) of at least two of the received input signals. i ) and the defined vector (v) j );and Based on the optimization (59), at least one set of filter coefficients (x) of at least one digital input filter (31) associated with a specific input channel are obtained. i ).

2. The method (45) according to claim 1, wherein, Input multiple defined vectors (v j ), and for the multiple defined vectors (v) input j A specific defined vector in ) obtains at least one set of filter coefficients (x i ).

3. The method (45) according to claim 1 or 2, wherein, The metric described in optimization (59) includes multiple optimization steps, and a set of filter coefficients obtained in one step remains unchanged for subsequent optimization steps.

4. The method (45) according to claim 1 or 2, wherein, The optimization is constrained by the filter coefficients (x) i ( ) at least one linear equality constraint.

5. The method (45) according to claim 4, wherein, The method (45) includes: inputting (53) the DC gain value of at least one input channel, and based on the DC gain value (DC... i To determine the corresponding linear equality constraints.

6. The method (45) according to claim 1, 2 or 5, wherein, The optimization (59) is subject to at least one nonlinear inequality constraint used to limit the gain of an input filter (31) at a given frequency.

7. The method (45) according to claim 1, 2 or 5, wherein, The optimization is constrained by the filter coefficients (x) i At least one boundary constraint.

8. The method (45) according to claim 1, 2 or 5, wherein, The calibration sampling frequency (f) is increased compared to the operating sampling frequency applied by the measuring device (11) to perform physiological measurements. samp ) Receive the sample (c i ), and / or, where the received sample (c i The calibration resolution (res) of the sample is different from the measurement resolution of the sample received to perform the physiological measurement.

9. The method (45) according to claim 1, 2 or 5, wherein, The optimization (59) metric includes minimizing the penalty function of the output samples, which correspond to the values ​​of the filter coefficients (x). i The input signal is filtered according to the defined vector (v) j ) calculate at least one vector signal.

10. The method (45) according to claim 9, wherein, Optimization (59) includes minimizing multiple different penalty functions of the output sample simultaneously or minimizing the scalar objective function of different penalty functions.

11. The method (45) according to claim 9, wherein, At least one penalty function is a convex or quasi-convex penalty function.

12. The method (45) according to claim 9, wherein, At least one penalty function is a norm.

13. A computer program product comprising program code units, wherein when the computer program product is executed on a computer (37), the program code units are configured to cause the computer (37) to perform the steps of the method (45) according to any one of the preceding claims.

14. A calibration device (13) for minimizing common-mode interference in a physiological measurement device (11), the calibration device (13) comprising a processing unit configured to: Receive (57) multiple input signals (s) from the measuring device (11) i ) samples (c i The sample corresponds to the test signal (TS) applied to multiple input channels of the measuring device (11). Input (51) describes samples (c) from at least two input channels. i At least one defined vector (v) of a linear combination of ) j ); The metric calculated based on at least one vector signal is optimized (59) based on samples (c) of at least two of the received input signals. i ) and the defined vector (v) j );and Based on the optimization (59), at least one set of filter coefficients (x) of at least one digital input filter (31) associated with a specific input channel are obtained. i ).

15. A physiological measurement device (11), comprising: Multiple input channels are used to process multiple input signals (s i ); Memory device (29), which is configured to store multiple sets of filter coefficients (x i The multiple sets of filter coefficients (x) i At least one set of filter coefficients in the measurement device (11) is related to at least one digital input filter (31), which is configured to filter an input signal (s) i ) samples (c i ) Perform filtering to obtain filtered samples of the input signal, and The processing unit (25) is configured to compute a plurality of vector signals, one of which is calculated based on a defined vector associated with the one vector signal according to filtered samples of the input signals associated with the defined vector among the plurality of input signals; The memory device (29) is configured to store different sets of filter coefficients (x) associated with an input channel. i ), and the processing unit (25) is configured to select the different groups of filter coefficients (x i A set of filter coefficients is selected for computing a specific vector signal, and the selected set of filter coefficients for computing the specific vector signal is obtained based on an optimization of the computing metric according to the specific vector signal.

16. A physiological measurement method, comprising: Multiple input signals are processed via multiple input channels (s) i ); Read multiple sets of filter coefficients (x) from memory device (29) i The multiple sets of filter coefficients (x) i At least one set of filter coefficients in the ) are related to at least one digital input filter (31) of the physiological measurement device (11), said at least one digital input filter (31) being configured to filter an input signal (s) i ) samples (c i ) Perform filtering to obtain filtered samples of the input signal, and Calculate multiple vector signals, wherein one of the multiple vector signals is calculated based on a defined vector associated with the one vector signal, using filtered samples of the input signals associated with the defined vector from among multiple input signals; Among them, different sets of filter coefficients (x) related to an input channel are read from the memory device (29). i ), and select the different groups of filter coefficients (x) i A set of filter coefficients is selected for computing a specific vector signal, and the selected set of filter coefficients for computing the specific vector signal is obtained based on an optimization of the computing metric according to the specific vector signal.

Images