RU2636905C1 - Method for analogue-digital parameter measurement at automatic fragmentation of electrocardiosignals - Google Patents

Abstract

FIELD: medicine.

SUBSTANCE: method for analogue-digital measurement of parameters for automatic fragmentation of electrocardiograms (ECS) can also be used in electrophysiology when measuring parameters and fragments of electrophysiological indices and in measuring devices for measuring parameters and fragments of both random and deterministic signals under interference. ECS recording, its preliminary analogue filtering and analogue-to-digital conversion are performed. An a priori established sliding symmetric transformation (SST) of the obtained discrete samples is performed within the local region with respect to its central digital reference (DR). A fragment of the signal containing an extremum is isolated, and the extremum is determined within the fragment. Analogue-to-digital conversion is performed at an increased frequency of time sampling. In order to extract a signal fragment containing the extremum, the first and second sliding asymmetric separate transformation (SAST) for the DR of the same local area are additionally performed during SST. The obtained values of the first and second SASTs are brought to one positive sign and compared with the a priori signatures of the a priori established first and second thresholds, respectively. The nonlinear correlation (NC) of the SAST values is determined only when the SAST values of the corresponding thresholds are simultaneously exceeded at each clock by a pairwise multiplication. For other cases of comparison results, its values are equal to zero. The NC value is compared with the third a priori established threshold and an ECS fragment containing only its first extremum is identified by exceeding the NC of the third a priori established threshold.

EFFECT: increased accuracy and reliability of measurements in real-time parameters with automatic fragmentation of electrocardiosignals and reliability of diagnostics by cardiological indicators under the influence of interference and reduced complexity of method implementation.

2 cl, 12 dwg

Description

The invention relates to medicine, in particular to electrocardiography. The invention can be used in electrophysiology in the measurement of parameters and fragments of electrophysiological indicators under the influence of interference, as well as in the measuring technique for measuring parameters and fragments, both random and deterministic signals.

There are various methods and algorithms for determining or measuring the parameters of fragments in the automatic fragmentation of electrocardiosignals (EX). The extremums of QRS complexes corresponding to Q, R, or S teeth, as well as the time intervals between teeth or reference points of other fragments of EX, are referred to the parameters of ECS. As analysis [1] and [2] showed, in the known methods for measuring fragment parameters such as time intervals between extrema and their amplitudes, during automatic fragmentation of ECS, almost the entire arsenal of signal processing methods developed to date in information technology is used [2] (p. 73 and 76). Known methods for detecting QRS complexes and measuring parameters such as RR intervals, which are based on the so-called wave and frequency-time transformations, including those based on probabilistic characteristics, such as a correlation function or the average number of zeros [2]. Methods are also used based on neural network algorithms, various signal transformations, such as Fourier, Wavelet or Hilbert, empirical mode decomposition (EMD or EMD), requiring a significant amount of numerical operations due to the use of multiple multiplication and addition of the results of multiplication or other mathematical operations and operations. In these methods, all kinds of options are also used, as a rule linear, preliminary filtering of the EKS with the subsequent determination of the amplitude and derivatives of the EKS, including the principle of threshold detection without focusing on the conditions for choosing the frequency of time sampling, since it is assumed by default its standard value, established by the sampling theorem . Based on essentially redundant transformations, the selection of the desired EX fragments during automatic fragmentation is carried out as a result of detection of the corresponding signs of identifiable EX fragments according to the specified criteria using the convolution operation with a set of model, usually orthogonal functions, after preliminary linear filtering of the EX in the conditions of permissible distortions (artifacts) . Other methods are known for automatically detecting EX fragments and measuring their parameters. The most widespread, due to its high reliability, was the method of detecting the Pan-Tompkins EX-QRS complex [1], according to which the following actions and transformations are performed: band pass filtering, differentiation, squaring, followed by integration on the basis of a sliding window and adaptive threshold procedure. All known methods, and especially the Pan-Tompkins method, are aimed at increasing the reliability and reliability of detecting the QRS complex and, as follows from the above actions, are based on a rather complicated sequence of transformations of the initial EX, including its preliminary filtering [2]. Thus, the known methods combine linear filtering of the EX from interference with the procedure for automatic detection of fragments and measurement of their parameters. There are also known methods that offer a separate filtering procedure from interference of the decomposition of the EX on the frequency components, followed by automatic fragmentation, restored the EX on these filtered components. This approach, with the indicated separate solution to the filtering problem and the subsequent automatic fragmentation of the EX, is also not acceptable due to the rather complicated nonlinear filtering procedure by dividing the EX into frequency components [3]. The essence of the method is that after recording and decomposing the ECS into frequency components, followed by filtering each frequency component, the ECS is restored with the release of the cardiocycle and the isoline section on the TP segment (the time interval between the teeth of the EX T and P) of the restored EX by building energy surfaces the density of the selected cardiocycle and the isoline section on the TP segment based on the Hilbert-Huang transform in the energy-frequency-time coordinate system. The threshold surface is formed from the energy density of the isoline section and the energy density surfaces of the selected cardiocycle of the electrocardiogram are compared with the formed threshold surface. This method essentially eliminates the possibility of automatic fragmentation of ECS in real time and ultimately leads to an increase in the time and amount of calculations to obtain the desired result, which is unacceptable in the conditions of screening and monitoring.

Thus, as the analysis shows (see also [1] and [2]), the development of known methods went along the path of applying sophisticated processing methods for ECS, leading to a significant increase in the volume of calculations, due to which a slight increase in the reliability of automatic detection is achieved desired fragments and their parameters. Moreover, the growth usually amounts to about tenths, or even hundredths of a percent. At the same time, a theoretical analysis of noise immunity when comparing methods, as a rule, is not given in detail, as well as estimates of the probability of errors of the first and second kind with an inevitable threshold decision making in order to automatically detect the corresponding fragment of an EX. Introduction to known methods of automatic fragmentation of adaptation algorithms for tracking the threshold level depending on changes in the ECS from patient to patient and ECG removal conditions leads to a certain positive effect when detecting the desired fragments and evaluating or measuring their parameters [1], [2]. The resulting positive effect is manifested in an increase in the robustness property, i.e. independence or weak dependence of the corresponding automatic fragmentation method on variations of nosological changes in EX fragments in the diagnosis of identical pathologies in different patients. However, with the introduction of adaptation, as practice shows, an increase in the complexity of the method, due to the need to use additional calculations and time-consuming processing of ECS to implement adaptation, is usually not sufficiently compensated by the growth of the resulting positive effect. The effectiveness of the new proposed methods, judging by the published data, is estimated mainly only by the results of modeling on electrocardiograms from an annotated data bank of electrocardiograms, for example, the MIT-BIH database [4]. Therefore, the overall picture of the comparative effectiveness of the proposed methods is revealed according to the results of their modeling, given by the authors, on the basis of the generated ECS data bank, obtained without taking into account the variety of interference that can occur when an ECG is taken under various conditions by electrocardiographs manufactured by different manufacturers, i.e. on different technologies and implementation schemes. Therefore, as a rule, in the known methods, there is practically no theoretical justification for the growth dynamics of the robustness property of the measurement results of the desired parameters to change the conditions for picking up the EX and to the types of artifacts affecting the EX depending on the growth of complexity and, thereby, the increase in the amount of calculations in the proposed methods. Whereas when solving tasks such as screening and monitoring, due to the mass of studies, it is necessary to extremely simplify the requirements for the conditions and complexity of operation of the corresponding technical equipment while reducing the cost and maintaining at an acceptable level of reliability and reliability of obtaining diagnostic results under conditions of exposure of various kinds artifacts.

Thus, the method for measuring the parameters of fragments during automatic fragmentation of ECS should have the property of increased robustness to nosological changes in ECS for different patients, artifacts and various conditions for the use of diagnostic tools with minimal requirements for the complexity and cost of their operation.

Since the most common methods for detecting and identifying QRS complexes with determining the RR time intervals between the selected characteristic or reference points of QRS complexes of the ECS, in particular between maxima of the R teeth, providing the ability to measure rhythmograms, methods for complex solution are of practical interest tasks, i.e. methods combining noise immunity with the selection of the required reference points with the measurement of the desired parameters with automatic fragmentation of the ECS and the relative simplicity of the method using electronic analog-to-digital devices. As reference points, extremes of R teeth are usually chosen, corresponding to either the maxima or minima of the R teeth when they are positive or, at a much less frequent occurrence, negative values. Therefore, the time RR-intervals are measured between the extrema of the R waves of the EX. In this regard, for the analogue of the present invention, it is advisable to choose a noise-resistant method for measuring the position and amplitude of the extrema of random signals, which is implemented by the device proposed in [5]. According to this method, in order to increase the noise immunity when determining the reference point of the extrema, an odd odd number of the number of comparators (in the amount of three) is introduced and the sign of the increment of the input random signal at the adaptively set time cycle with a majority decision is made about the presence of the desired extremum, i.e. in most coinciding cases, a change in the sign of the increment of the input signal. The disadvantages of this method include, firstly, that when implementing this method, circuit redundancy is introduced and, secondly, the found extrema are not identified by additional signs with an extremum belonging to the R wave, for the case when an EX is fed to the input of the device .

Closest to the proposed invention can be attributed to the method [6]. In the method [6], without additional transformations of the EX, introduced in the above methods, in addition to the standard linear filtering, the power of the EX is calculated as the sum of the absolute values of the derivative of the EX in the region around this ECG reference. Further, in the prototype, for preliminary rough determination of the moment of occurrence of the QRS complex by essentially sliding summation of the absolute values of the derivatives of the EX, determine its power in an a priori specified local area or window of the current digital readout (CO) of the EX with a sequential comparison of the obtained summation results with an a priori set threshold value. The window width is selected a priori approximately equal to the width of a typical QRS complex, and the specific threshold value is also selected a priori approximately equal to half of the maximum calculated EX power. By the moment of exceeding the determined power of the ECS of the set threshold, a preliminary rough estimate of the duration of the interval of the QRS complex of the EX is found, with subsequent detection and identification of the first derivative of the extremum within the previously found duration of the interval of the QRS complex. Based on the found extremum, the moment of occurrence of the QRS-complex of the ECS and, therefore, its parameters is specified. It is assumed that the type of extremum, i.e. maximum or minimum is pre-selected depending on the further use of the obtained measurement results. In order to increase the accuracy of finding the extrema of the EX derivative according to the results of analog-to-digital conversion, an approximation by a parabola is performed within the set sampling interval and the moment of appearance of the QRS complex is calculated as the moment of the extremum of a numerically defined parabola. This allows, according to the proposed method, to reduce the measurement error of the extremum in the conditions of standard time discretization of the EX according to the sampling theorem. To attenuate the influence of high-frequency noise, the determination of the derivative of the ECS in the prototype is proposed to be performed according to the results of moving summation of a finite number of weighted finite differences that are symmetric with respect to the current central center of the ECS [7], based on a moving digital filter, which can be considered as a special case of a generalized operator of a moving symmetric transformation (BSS). In the prototype, due to the use of a digital noise-resistant filter in the form of a special case of SSP based on the use of weighted finite differences (hereinafter simply in the form of SSP), additional operations to increase noise immunity, which are introduced in other known methods for detecting QRS complexes, are excluded [8]. In addition, the digital noise-tolerant filtering, which is implemented by the SSP, allows, in contrast to the analogue [5], to find the noise-resistant change of the sign of the derivative, not in parallel but sequentially in time, without introducing hardware redundancy into the implementation scheme. The advantages of the method proposed in the prototype when measuring the RR-interval should include its certain insensitivity to slow changes in the signal or to the drift of the contour due to its exclusion from the original EX due to the use of MSS.

Since the prototype implies, by default, the use of standard time sampling, established by the sampling theorem, which is not optimal for determining the derivative with respect to finite differences, therefore, to compensate for the influence of the sampling interval on the accuracy of measuring the time position of the QRS complex, an additional parabolic approximation of the EX is performed. At the same time, when using digital noise-resistant differentiating filters in the form of MSS, an important role in obtaining the results of differentiation under the influence of interference, including network interference, is played by the coordinated choice of the sampling interval of the ECS not only with the upper cut-off frequency of the spectral function of the EX, but also with the interference repetition rate , including network interference [9]. Fulfillment of the last condition leads to the need to increase the sampling frequency of the EX when it is converted to analog-to-digital, or, in other words, to measure the EX when the time resolution is increased.

The disadvantages of the method proposed in the prototype.

- The main disadvantage of the prototype, for the elimination of which were proposed various methods listed in [2] and noted above, should be attributed to the low reliability of determining the duration of the time interval of the QRS complex containing the maximum R of the EX wave, calculated by exceeding the local power of the EX calculated according to the first derivative based on the MTP in the area of the current central heating center, a priori set threshold. This disadvantage is due to the fact that in the prototype local power significantly depends and, therefore, varies not only from the quality and conditions of information retrieval when exposed to interference, but also from patient to patient, both healthy patients and their nosological state, which affects the reliability of the a priori choice of the threshold. The result of this drawback is the deterioration of the robustness of the method, as a result of which the reliability of the rough determination of the ECS region is significantly reduced, i.e. The QRS of the EX fragment containing the desired extremum R of the tooth, due to an increase in the probability of errors of the first and second kind, leading either to false detection or to skip it due to interference amplified by the error in calculating the first derivative of the EX at the non-optimal sampling frequency.

- The measurement of the moment of appearance of the R wave of the QRS wave of the ECS complex is not direct, but indirect, due to the introduction of additional operations related to the approximation of the derivative within the specified ECG sampling interval under the same conditions of residual interference after preliminary filtering. It is known that indirect measurement reduces the accuracy of measurements in comparison with direct measurements, which are achieved with an increase in the frequency of temporal sampling or temporal resolution of the ECS compared to the standard frequency established by the sampling theorem, and excludes the possibility of high-precision measurement of the parameters of the ECS under the influence of residual interference.

Thus, the analogue and prototype, as well as the well-known methods further noted above, based on sophisticated information processing technologies, including such transformations as Fourier, Wavelet or Hilbert, EMD, as well as correlation, neural networks and adaptation algorithms, do not provide real-time combination of robust automatic fragmentation and measurement of EX parameters with a decrease in requirements for the volume and complexity of processing EX while maintaining the established reliability and accuracy measured th in situations of intense exposure to noise and simplified conditions of technical equipment implementing the method. On the other hand, the desire to simplify the prototype method and, therefore, the technology of automatic fragmentation and measurement of ECS parameters in real time by reducing the amount of computation, cannot provide an effective solution to the problem, for example, screening and prophylaxis of the population by cardiology and other indicators, with due reliability and robustness while maintaining the required accuracy and reliability of measurements in situations of intense exposure to interference with reduced requirements for operating conditions and complexity technical means for performing a method.

The technical result of the method consists in increasing the accuracy and reliability of real-time measurements of parameters during automatic fragmentation of ECS and the reliability of diagnostics by cardiological indicators under the influence of interference, reducing the amount of computation and the complexity of hardware and software hardware that implements the method.

The essence of the invention lies in the fact that to achieve a technical result, as in the prototype, an ECS is removed, its preliminary analog filtering, analog-to-digital conversion of EX, perform a priori established MSS of the obtained discrete samples within the local area, relative to its central CO. A signal fragment containing an extremum is isolated and an extremum is determined within the fragment. The difference is that the analog-to-digital conversion is carried out at a higher time resolution. To select an EXS fragment containing an extremum, when performing MSS, the first and second sliding asymmetric transformations (SARPs) are additionally performed for the center of the same local region. The obtained values of the first and second ARPDs lead to one positive sign and are compared with the first and second thresholds of the same name a priori set respectively. Only with simultaneous exceeding of the SARP values of the corresponding thresholds at each cycle by pairwise multiplication, the nonlinear correlation (NK) of the SARP values is determined and its value is equal to zero for the remaining cases of the comparison results. The NK value is compared with the third a priori established threshold and an EX fragment is selected that contains only its first extremum by exceeding the NK value of the third a priori established threshold.

Not only symmetric and asymmetric finite differences with respect to the central center, but also other operators of sliding symmetric and asymmetric transformations that are compatible with each other and can be selected as SSP and SARP, allowing the formation of SARP from SPS by a priori choice of the corresponding parameters. In the proposed method, as well as in the prototype, out of the many possible variants of the SSP, SSPs are used to determine the moment of the signal extreme occurrence, in the form of symmetrical moving weighted, i.e. multiplied by an a priori established numerical coefficient, the finite differences of the EX [7], [8]. Moreover, the SSP is performed according to the a priori chosen symmetric final number of the color centers relative to the central color center with the discretization of the EKS at an increased temporal resolution. The numerical coefficients of the SSP are determined a priori on the basis of the principle of least squares, which ensures the least influence of noise on the result of determining the derivative with respect to central finite differences at a given point, as proposed in [7].

According to the author, the main distinguishing feature of the proposed method, to which the requested amount of legal protection extends, is that, with an increased time resolution, in addition to the known BSC, ARPAs are introduced with the final asymmetric number of COs established relative to the central CO in the local area, consistent with the local area of the SSP, with the subsequent determination of the tax code between the separately executed ARPAs reduced to the same sign. Moreover, NDs are determined by the results of separate comparisons of ARPA, but at coinciding time points (by numbers or addresses of the CO), with a priori given first and second thresholds of the same sign, pairwise multiplication of the results of the comparison only if the corresponding first and second thresholds are simultaneously exceeded and equated to zero NK values for all other possible cases of interaction of ARPA with thresholds. A fragment is selected with the signal extremum belonging to this fragment according to the coincident events of exceeding the NK of the a priori established positive sign of the third threshold and the noise-resistant moment of only the first detection of the extremum by SSP.

It is possible, for example, to select the BSC and ARPA of the color center signal based on the wavelet transform with the corresponding asymmetric setting of the shift parameter relative to the central color center within the set local area and a fixed scale parameter.

For the parallel measurement of the parameters of EX, belonging to fragments of different durations, for example, EX fragments such as T or P teeth, it is allowed to use parallel transformations and actions similar to those described above, but with different local areas of the SSP and time intervals between discrete color centers.

To generalize and illustrate the description of the SSP, we will present its mathematical form of recording, according to [7] and [8], in the form

where y (jτ _d ) is the SSP result in accordance with the well-known [7] algorithm for determining the weighted central finite differences of a signal, for example, EX at discrete time instants jτ _d (J = 1, 2, ...) with an interval of time sampling τ _{d of the} initial signal x (t) (this interval can be selected as a multiple of the analog-to-digital conversion time);

n is the set number of signal centers that determines the local area relative to the current central center jτ _d , within which the MSS is located;

m is a parameter that determines the interval between the color centers when determining the final difference, which is set depending on the time of the analog-to-digital conversion;

- априорно заданная весовая функция ССП.

- a priori given weight function of the SSP.

In accordance with formula (1), for the formation of each ARPA, asymmetrically relative to the current central center, jτ _d, set the center to zero a priori chosen values of the indices i. In the simplest embodiment, ARPAs are presented in accordance with the formulas

and

- соответствующие весовые функции САРП и

.Where

- corresponding weighting functions of ARPA and

.

To bring the ARPA obtained in accordance with formulas (2) and (3) to one positive sign depending on the sign of the EX tooth, the negative sign in the transformations (2) or (3) is corrected in the positive sign by a standard technique, for example, by multiplying by minus one.

The simplification and reduction of the amount of computation in the proposed method is achieved by the fact that the BSC and ARPA are performed in real time in accordance with formulas (1) - (3), moreover, the established number n of the EC EX should not usually exceed four, and the values of the corresponding weight functions are determined a priori and entered into the storage device of the corresponding technical means that implements the method. Moreover, the values of the terms in formulas (2) and (3) coincide with the similar terms in the transformations performed for the SSP in accordance with formula (1), the difference is only in the number of terms used. In essence, we obtain a significantly simplified analogue of the fast Fourier transform due to the fact that the number of terms in the sum and the number of multiplication operations performed are substantially less than in the Fourier transforms and, therefore, in the Hilbert transforms, not to mention the wavelet transforms proposed in the well-known methods [2] for automatic fragmentation of EX.

To find the values of NK SARP, one selects and at each time step only those results of the SARP EX that are multiplied in pairs that exceed the a priori positive values of the first and second thresholds at the same time step. In this case, the NK values are equal to zero for all other possible cases of separate interaction of the values of the ARPA with the established levels of the first and second thresholds. Since the ARPA, for example, of such interference as a superposition of network and random interference, leads to a comparison with the first and second thresholds of the results of the conversion of the ARPA, shifted in phase close to 90 °, the probability of coincidence in sign of the results of exceeding the first and second thresholds by the interference tend to to zero. As a result, the desired NC values are obtained, which for the time interval of the EX fragment containing R teeth are many times higher than the amplitude R of the EX teeth, while for other time intervals of the EX fragment, the NC values are either zero or many times lower than its maximum value. Therefore, with a proper choice of the SARP asymmetry, the value of the NK on fragments of ECS that do not contain an R wave decreases significantly, up to zero, due to a decrease in the likelihood of a coincidence for them of exceeding the SARP levels of the first and second thresholds. This, as well as the fact that only a smaller part of the values of the SARP EX is multiplied to determine the NK, and the achieved positive technical effect of the proposed method consists in comparison with the known ones, in which, taking into account the sign, all the results of the corresponding transformations of the TS EX are taken into account.

The essence of determining NK by pairwise multiplication is clearly revealed by the mathematical record of the determination of NK SARP, which has the form

where R [y ₁ (jτ _d ), y ₂ (jτ _d )] are the values of the TC; y ₁ (jτ _d ) and y ₂ (jτ _d ) are the results of a separate pairwise transformation of the SARP for j = 1, 2, ...; y ₀₁ and y ₀₂ - a priori set values of the levels of the first and second thresholds.

Exceeding the maximum of the NC SARP by several times the maximums of the R of the EX teeth under the influence of interference determines the robustness property and the reliability of measuring the parameters of the EX in the proposed method when exposed to artifacts in comparison with the prototype and other known methods. The appropriate value of the third threshold for NK SARP is recommended to be a priori chosen based on the permissible worst ratio between the level of the EX and interference. In this case, in order to establish the value of the third threshold a priori, one should be guided by the fact that the value of the NC on fragments of the EX, not containing the R wave, tends to zero or, due to the influence of noise significant in level, exceed the zero level with an admissible low probability and low amplitude or level. It is for leveling the last case that in addition to the first and second threshold it is proposed to use the third threshold, since due to the installation of the third threshold, the influence of interference during the automatic determination of a fragment containing an R wave is practically reduced to zero. This statement follows from the fact that a fragment containing an R wave differs from other ECS fragments in terms of a combination of characteristics: maximum and duration with a probability close to unity, even under the influence of random fluctuations. In this case, the influence of the shape variability of the EX fragment containing the R wave on its automatic detection is minimized, which virtually eliminates the need for methods using the Wavelet transform with varying scale and shift parameters to increase the reliability of identification of EX fragments.

The analysis of known methods also shows that the often used operation of squaring or another degree of central ECS in addition to increasing the amplitude R of the tooth also increases the amplitude of random emissions of ECS, which is practically eliminated in the proposed method.

An additional explanation of the essence of the selection of the EX fragment, containing its extremum, by the result of exceeding the value of the NC of the third threshold at each current beat j = 1, 2, 3, ..., reduces to the following. According to the proposed method, the NK value determined in accordance with formula (4) in the form R [y ₁ (jτ _d ), y ₂ (jτ _d )] is compared with the a priori set value of the third threshold y ₀₃ . At the same steps, at which the inequality R [y ₁ (jτ _d ), y ₂ (jτ _d )]> y ₀₃ is realized for the desired fragment, the steps are taken to detect the zero value of the SSP, i.e. y (jτ) = 0 (see formula (1)). If during the time interval (jτ, (j + s) τ _d ) equal to the length of the fragment on which R [y ₁ (jτ _d ), y ₂ (jτ _d )]> y ₀₃ , the event corresponding to the equality y [(j + l) τ] = 0, or rather, the fulfillment of the condition y [(j + l) τ _d ]> 0y [(j + l + 1) τ _d ] <0, then the maximum of the R wave is not fixed. However, with the appropriate choice of threshold levels y ₀₁ , y ₀₂ , y ₀₃ and the permissible level of residual interference after preliminary filtering, the probability of this event is practically zero (see the explanatory figures 1-12 below obtained for real ECS under the influence of artifacts). We denote the event of the detection of the R wave or its non-detection in the time interval (jτ _d , (j + s) τ _d ), which determines the fragment duration, by S _R [(j + l) τ _d ] for l = 1, 2, ..., s. Then, mathematically, the condition for detecting an event, S _R [(j + l) τ _d ], is 1 and corresponds to the presence of an R wave in the interval sτ _d , at which R [y ₁ (jτ _d ), y ₂ (j + s) τ _d )]> y ₀₃ , taking into account the opposite event equal to 0, can be written in the form for l = 1, 2, 3, ..., s

Let (j + l ₀ ) τ _{d be the} moment of appearance of the first extremum for (1 <l ₀ <s). Then, in order to exclude even unlikely cases of repeated determination of the extremum, i.e. the occurrence of the repeated event S _R [(j + l ₀ + l) τ _d ] = 1 for l = 1, 2, ..., l _s are not fixed. Thus, the time interval (0, τ _s ), where τ _s = l _s τ _d is the a priori set delay, determines the time interval for the prohibition of determining the repeated extremum within the detected fragment.

An analysis of the above formulas shows that the higher the temporal resolution during analog-to-digital conversion of the EX, the more accurately the boundaries and, therefore, the duration of the fragments, as well as the temporary position of the EX parameters are determined by reducing the time sampling interval τ _d .

When a priori establishing the values of the levels of the first, second and third thresholds, it is additionally advisable to be guided by the following a priori information. When establishing the levels of the first and second thresholds, they are guided by the fact that, on average, the results of the impact of ARPA on the ECS within the fragment containing the R wave significantly exceed the results of these effects in other cases. This property determines the robustness of the proposed method due to the non-critical nature of the requirements for choosing the values of the first and second thresholds, the values of which should be set equal to (35-45)% of the average maximum values of the SARP. Moreover, on average, the difference between the values of the first and second thresholds, due to some asymmetry of the phase portraits for separate SARPs, does not exceed 4-6% (see the explanatory Fig. 4 below). However, in view of the robustness of the method, the values of these thresholds can be set equal to each other. Moreover, the introduction of an adaptive algorithm for the correction of a priori selected levels of the first and second thresholds, as practice has shown, for various variations of the results of measuring the EX and interference levels does not give advantages.

Based on the foregoing, it follows that in the proposed method, in comparison with the prototype and the other listed methods, the accuracy and reliability of parameter measurement is increased by automatically extracting fragments of ECS under different conditions of its removal in patients with different nosological conditions while reducing the requirements for the volume and complexity of numerical and measuring operations in the method of analog-to-digital diagnostic devices with high time resolution.

In the proposed method there are no requirements for the determinism of the appearance of extrema in the signals under the influence of interference, which are characteristic of ECS, photoplethysmography, electroencephalography or rheography. The invention can also be used to measure the amplitude and repetition period of harmonic, triangular, sawtooth and other periodic signals when exposed to interference.

The measurement of parameters during the automatic extraction of signal fragments under the conditions of intense exposure to various types of interference by analog-digital hardware-software tools is an important diagnostic task, the reliability of assessing the state of, for example, the cardiovascular system and the effectiveness of further treatment depends on the solution of this during monitoring and screening or prophylaxis according to established nosological or other conditions of the diagnosed patient. The essence of the proposed solutions is illustrated by the following drawings.

FIG. 1 - graphs of EX fragments (dash-dotted line) and its symmetric difference (solid line) for n = 4 and τ = 4τ _pr = 800 μS in the region of the R wave, depending on the number of the center of the EX frequency.

FIG. 2 - a fragment of the graph of the real EX after preliminary filtration at τ _pr = 200 μS depending on the number of the center of the EX (abscissa).

FIG. 3 - graphs of the SSP and SARP obtained during the conversion of the ECS in the time domain from 12501τ to 13201τ (τ = 4τ _pr = 800 μS) containing a randomly selected R wave (the SSP graph is highlighted by a lighter wider line).

FIG. 4 - graphs of separate SARP EX for τ = 800 μS, presented in the form of a phase portrait, depending on the change in the result of the SSP for EX.

FIG. 5 - graphs of NK and EX after preliminary filtering at τ = 200 μS, presented in the form of a phase portrait, depending on the change in the result of the SSP EX.

FIG. 6 is a graph of changes in the ND (ordinate axis) for the EX fragment shown in FIG. 2, depending on the current cycle of the CO EX.

FIG. 7 - dynamics of changes in the tax code (chart 1) and an EX fragment (chart 2), within the RR-interval of the dependence depending on the numbers of the central station of the EX.

FIG. 8 - the dynamics of separate changes in the SARP (graphs 1 and 2), as well as the BSC (graphs 3) within the local area of the R wave, in adverse cases of interference with ECS.

FIG. 9 is a comparison of the R wave shape and the NKF for the fragment of the graph shown in FIG. 8.

FIG. 10 is a fragment of the graph of the ECS of another patient, measured under other conditions, after preliminary filtration at τ _pr = 150 μS.

FIG. 11 - graphs of NEC and EX in the form of phase portraits, depending on the change in the SSP for EX, a fragment of which is shown in FIG. 10.

FIG. 12 - fragments of graphs of ECS of different patients: a) under local exposure to an artifact, and b) with significant variation in the amplitude of the R wave and the effect of interference.

Information confirming the possibility of implementing the proposed method. The method can be implemented in analog-to-digital electronic devices, including devices: information retrieval with analog gain; pre-filtering; time sampling and analog-to-digital conversion, as well as the microprocessor and its components.

When implementing the method, the initial information for selecting the parameters of the BSC and the type of SARP in accordance with formulas (1) - (3), in order to perform the established sequence of actions and operations, is:

1) the frequency range Δƒ of the measured signal Δƒ = ƒ _GDV -ƒ _UAH (for EX, for example, ƒ _UAH = 0.05, and ƒ _GDV = 250 Hz);

2) the frequency of the temporal sampling of the signal, which is recommended to choose within the range of (15-20) ƒ _GDV for the implementation of the requirement of high temporal resolution for a given bit depth of the analog-to-digital conversion (usually in the range of 12-14 binary digits);

3) the ratios of the characteristics and parameters of the ECS and interference, acceptable to achieve the specified reliability of the diagnostics, moreover, both energy characteristics and the ratio of the peak and temporary values of the EX and the total interference can be taken as parameters;

4) the values of the parameters BSC and SARP EX, and a priori set the parameters are:

n is the number of signal centers defining the local area (window) relative to the central center;

;ϕ (i) is an a priori given weight function of an integer argument, for example, in the form:

;

τ _d is the time interval of time sampling of the EX, multiple or equal to the time of analog-to-digital conversion;

m is a parameter that determines the interval between the center for determining the final difference;

5) a priori set values of the levels of the first, second and third thresholds y ₀₁ , y ₀₂ and y ₀₃ ;

6) a priori set delay τ _s = l _s τ _d , which should not exceed the maximum time interval (jτ, (j + s) τ _d ) when the conditions R [y ₁ (jτ _d ), y ₂ (jτ _d )]> y ₀₃ (see formula (5)).

The steps necessary to implement the proposed method are as follows. EKS are removed, amplified in accordance with the range of the used analog-to-digital converter, preliminary analogue filtering is performed to select EKS in the selected frequency range Δƒ and eliminate the isoline drift. After preliminary analog processing, the ECS performs its analog-to-digital conversion and temporal sampling with an increased temporal resolution that is 5-10 times higher than the upper cutoff frequency of the frequency range cutoff Δƒ They perform SSP in accordance with formula (1) with a priori set parameters. An example of determining the BSC under conditions of exposure to residual interference after filtering is shown in FIG. 1 for a fragment of a real ECS shown in FIG. 2.

Further, according to the proposed method, the following sequence of actions is performed. Based on the data that are used to perform the MTSP, within the same window in real time, separate the SARPs are determined (see formula (2) and (3)) with the established parameters, for example, in accordance with the formulas

and

where y ₁ (jτ) and y ₂ (jτ) are the results of a separate conversion of the ARPA.

By changing the negative sign to positive of one of the separately received negative SARPs in accordance with the polarity R of the EX wave, both separate SARP values are brought to a positive sign. The ability to perform within the same window the SSP and the SARP during one time interval of sampling the EX should be ensured by the speed of the selected type of microprocessor. An example of the determination of separate SARPs, in accordance with formulas (6), is shown in FIG. 3 and 4 for an EX, a fragment of which is shown in FIG. 2. Moreover, in FIG. 3 shows for comparison the graphs of BSC and ARPA obtained as a result of transforming an EX in the time domain containing an arbitrarily selected R wave, whereas in FIG. Figure 4 shows graphs in the form of phase portraits linking the dependence of the BSC with the SARP when measuring EX for one minute. In this case, the values of the SSP EX are plotted along the abscissa axis, and the values of the ARPA along the ordinate. Within the same sampling interval, the ECS separately compares the results of the ARPA with the a priori given first and second thresholds of the same name. Based on the results of comparison, multiplication determines the values of NK only by simultaneously exceeding the result of separate comparison of the SARP of the first and second thresholds. For all other possible cases of interaction of the ARPA with the first and second thresholds, the TCs are equal to zero. In FIG. 4, the phase portrait traces the asymmetry of the results of the conversion of the ECS in accordance with the SARP obtained for n = 4 by the formula (6). Due to this asymmetry, the phase of the results of the SARP EX for interference is shifted by a value close to 90 °, which leads to the value of the NC for the interference close to zero. The phase shift of the interference is illustrated in FIG. 3 in the area of changes in the ECS in front of the local area containing the R wave and after it. In FIG. 3 dash-dotted line and a solid thin black line describe the dynamics of the impact of ARPA on ECS. As a result, the desired NC values are obtained, which in the temporal region of the R wave exceed the amplitude R of the EX teeth, while for other time intervals, the NC values are either zero or many times lower than the maximum value. The graphs shown in FIG. 5, and also based on the results of comparing the graphs of FIG. 2, 6, 7 confirm that the NK maximum is several times higher than the maximum R of the EX teeth, moreover, as follows from FIG. 6 and 7 in time intervals that do not coincide with the R wave, regardless of the level of interference, the TC is zero. This is the achieved positive effect of the proposed method due to the difference between the definition of NK from the classical, well-known method for determining cross-correlation, in which the results of all transformations of the DSP signals are summed up taking into account the sign.

The indicated NK property is especially important with a significant level of interference affecting the ECS, under adverse conditions for information retrieval. Exceeding the maximum of the NC ARP by several times the maximums of the R of the EX teeth under the influence of interference, just determines the robustness property of the proposed method with respect to artifacts in comparison with the prototype and other known methods and allows you to a priori establish the corresponding value of the third threshold based on the allowable worst-case ratio between the level EX and interference. For an a priori determination of the value of the third threshold, one should be guided by the fact that the value of the NC in areas other than the local region R of the tooth is either zero (see Fig. 6, phase portraits shown in Fig. 5, as well as Figs. 7-9) , and if it is different from zero, due to the impact of significant interference level, then with a low probability and low level of exceeding the zero level. It is for the latter case that it is proposed to use the third threshold, since due to the installation of the third threshold, the effect of interference is practically reduced to zero. This statement follows from the fact that the R wave in the aggregate of such parameters as maximum and duration differs from similar parameters of random fluctuations of EX and its other fragments with a probability close to unity. Due to this difference, with the appropriate choice of the asymmetry of the SARP and the levels of the first and second thresholds, the value of the NK in areas other than the local time region R of the tooth decreases significantly, up to zero, due to a decrease in the probability of coincidence in sign of the excess of the SARP of the interference of the first and second thresholds . Based on the results of comparing the NC with the third threshold, if the value of the NC exceeds the level of the third threshold, they establish the presence of the SSB value equal to zero by detecting a change in the SSB sign in the current sampling interval, compared with the previous interval, from positive to negative (see Fig. 3). In this case, in the absence of a zero SSP value and an excess of the NK level of the third threshold, the measured time parameter, for example, the RR interval, in the counter or in the corresponding memory cell is increased by one. If it is found that the SSP value is zero when the NK of the third threshold is exceeded, then the operation of storing the obtained value of the measured time parameter with its subsequent zeroing is performed in order to measure the next sought value of the time parameter. With each detection of a zero value of the SSP, within the interval of exceeding the value of the NK of the third threshold, the timer for prohibiting the execution of these operations for repeated, i.e. false due to the influence of interference, detection of a zero value of the SSP. The time interval of the timer τ _s must not exceed the maximum duration of the QRS complex.

It should be noted that the analysis of known methods shows [2] that the operation of squaring (or another degree) of the EC EX in addition to increasing the amplitude R of the tooth also increases the amplitudes of random emissions of EX, which is practically impossible in the proposed method. This statement is illustrated in FIG. 5 in the form of phase portraits linking the dynamics of changes in the values of NC SARP and TSO EX, depending on the change (deposited on the abscissa axis) of the effects of SSP on EX. In addition to FIG. 5 in FIG. 6 are shown in the form of a graph of the values of NK of the EX fragment shown in FIG. 2, depending on the current time step j or the serial number of the CO EX. From FIG. 6 and 9 (for FIG. 9, a fragment of the ECS is shown under conditions of intense interference (see also Fig. 8)) it follows that for the case under consideration, the NC is nonzero only in the local region R of the tooth, which allows the probability of false determination to be reduced to zero the reference point, which coincides with the maximum of the R wave, relative to which the RR interval of the ECS is measured. The last statement confirms the robustness property of the proposed method. This fact is particularly illustrated in FIG. 7-12 with a significant distortion of the morphology of the EX in the R region of the tooth (Fig. 12) and when exposed to a significant noise level (Fig. 8-10), in various modes of operation of the electrocardiograph and the conditions for taking information from different patients.

By increasing the sampling frequency due to the speed of modern analog-to-digital converters and microprocessor performance, they technically provide for the direct measurement of not only the maxima of the EX tooth, but also the intervals between them under the influence of noise and drift of the EX isoline contour. Moreover, the accuracy of measuring the temporal parameters of the EX for zero SSP increases with increasing frequency of the temporary sampling of the EX, and the accuracy of measuring the maximum, for example, of the R wave, depends on the bit depth of the analog-to-digital converter.

Information sources

1. Tikhonov EP The conceptual model of the subject area of automatic fragmentation of electrocardiosignals based on the logical scheme of algorithms (Part 1) // Izv. LETI. 2015. No8. S. 85-95.

2. Tikhonov EP The conceptual model of the subject area of automatic fragmentation of electrocardiosignals based on the logical scheme of algorithms (Part 2) // Izv. LETI. 2016. No1. S. 70-83.

3. Pat. No. 2486862. The Russian Federation. Adaptive noise suppression method in electrocardiograms / Bodin O.N., Krivonogov L.Yu., Tychkov A.Yu., Churakov P.P., Volchikhin V.I., Shurygin V.A. Application 2011147031/14, pending. 11/18/2011, publ. 07/10/2013. Bull. No. 19.

4. "MIT / BIH arrhythmia database-Tape directory and format specification," Document BMEC TR00, Mass. Inst. Technol., Cambridge, 1980. Database is available from Bioengineering Division KB-26, Beth-Israel Hospital, 330 Brookline Avenue, Boston, MA 02215.

5. A.c. No. 805358. RU. A device for determining the extreme values of random processes / Zhivilov G.G., Smetanin N.M., Tikhonov E.P. Application 2682708 / 18-24, pending. 10.11. 1978, publ. 02/15/1981. Bull. No. 6.

6. Pat. No. 2122349. The Russian Federation. Method for detecting qrs-complex and determining the moment of its appearance / Mazhara Yu.P., Dmitriev I.N. Application 94037134/14, 09/30/1994, publ. 11/27/1998.

7. Lantsosh K. Practical methods of applied analysis. Spavochny manual / Transl. from English M.Z. Kainer, ed. A.M. Lopshitsa. - M. FIZMAT-LIT. 1961 .-- 524 p., Ill.

8. Tikhonov EP ECG filtering algorithm with high-resolution time sampling // Izv. SPbGETU "LETI". 2012. No.6. S. 89-95.

9. Tikhonov E.P., Tsvetkov E.E. ECG filtering algorithm with high resolution time sampling. Interference Suppression // Izv. SPbGETU "LETI". 2012. No7. S. 97-102.

Claims (2)

1. The method of analog-to-digital measurement of parameters during automatic fragmentation of electrocardiosignals (ECS), which consists in the fact that they perform pick-up of the ECS, its preliminary analog filtering and analog-to-digital conversion, a priori established sliding symmetric conversion (BSC) of the obtained discrete samples within the local area relative to its Central digital reference (CO), select a signal fragment containing an extremum, and determine the extremum within the fragment, characterized in that the analog-to-digital conversion is performed at an increased frequency of temporal sampling, and to extract a signal fragment containing an extremum, when performing the MSS, the first and second sliding asymmetric separate transforms (ARPAs) for the center of the same local region are additionally performed, the obtained values of the first and second ARPAs to one positive sign and compare with the same threshold signs of the first and second thresholds respectively set, only if the ATS values are exceeded Using the corresponding thresholds at each cycle by pairwise multiplication, the nonlinear correlation (NK) of the SARP values is determined and its values are equal to zero for the remaining cases of the comparison results, the NK value is compared with the third a priori set threshold, and an ECS fragment containing only its first extremum exceeding the NK of the third , a priori set threshold. 2. The method according to p. 1, characterized in that from the set of possible options SSP and SARP choose weighted final signal differences with the final asymmetric number of the CO EKS relative to the central CO established for the formation of the SARP.

Images