UA20101U - Method for processing magnetocardiographic signals - Google Patents

Abstract

The method for processing the magnetocardiographic signals comprises the elimination of the undesired signals. In addition, the magnetocardiographic signals are synchronized within all the magnetocardiographic canals of the magnetocardiograph using the reference electrocardiogram. The magnetocardiographic signals are divided into the cardiocycles. The cardiocycles with the undesired non-harmonic signals are eliminated. The cardiocycles synchronous with the specific electrocardiographic patterns are selected. The selected cardiocycles are summed up.

Description

Description of the invention

The useful model refers to medicine, namely to cardiologists and can be used in 2 instrumental diagnosis of cardiac diseases using the non-invasive method of magnetocardiography (MKG), for population monitoring in order to determine the risk of cardiac diseases, as well as for the study of electrophysiological processes in the human heart in scientific purposes.

In (patent M/003026346) a method of processing magnetocardiographic (MKG) signals by reducing their noise based on wavelet transformation using reference channel signals is considered. In fact, this approach 70 can be considered a method of adaptive interference compensation (ACC). But it considers only one stage of pre-processing - ACP, and does not reveal the entire way of processing signals before building magnetic maps.

I(patent UMO02177821 also considered a method of processing ICG signals by reducing their noise, but at the expense of using an averaging smoothing filter. This approach is a type of digital filtering, but it also does not cover all stages of processing ICG signals before constructing magnetic maps. However, 72 since it uses signals obtained by means of a 7-channel device in unshielded conditions, it is the closest analogue, because this patent also uses signals with a similar level of magnetic interference.

The technical task of the useful model is to develop a method of processing magnetocardiographic signals, which allows to increase the quality (signal/noise ratio level) of ICG signals after processing the data registered without the use of a magnetically shielded camera (MEK) to a level sufficient for the construction of magnetic maps and their subsequent medical analysis

The task set in the method of processing magnetocardiographic signals, which includes cleaning magnetocardiographic signals from interference, is achieved by the fact that it additionally synchronizes magnetocardiographic signals in all magnetocardiographic channels of the magnetocardiograph using a reference electrocardiogram and breaks down magnetocardiographic signals into cardiocycles, extraction of cardiocycles that include undeleted inharmonic interference, selection of cardiocycles synchronous with certain electrocardiographic patterns, and summation of all selected on different groups of cardiocycles.

The technical result is achieved due to the fact that after processing the ICG signal in the presence of a high level of man-made magnetic interference, the signal/noise ratio will be sufficient for conducting medical diagnostics. Av

In Fig. 1 - a view of the ICG signal registered in unshielded conditions;

Fig. 2 - a view of the ICG signal after digital filtering to automatic transmission; --

Fig. 3 - a view of ICG signals, for which it is necessary to set up software filters in manual mode and repeat Ha») filtering and/or carry out additional smoothing; 325 Fig. 4 - a view of the ICG signal after automatic transmission; with

Fig. 5 - a view of ICG signals for which it is necessary to manually eliminate inharmonic interference;

Fig. 6 - a view of the ICG of cardiocycles that need to be deleted in manual mode;

Fig. 7 - view of sinus rhythm on ECG; "

Fig. 8 - a view of cardiocycles that need to be included in manual mode; 50 Fig. 9 - the result of automatic selection according to ECG patterns; From c Fig. 10 - view of the averaged ICG and ECG signals corresponding to extrasystole.

I» The proposed method is based on several consecutive actions that allow to clean the signal from interference, control the quality of processing and build high-quality ICG maps. A feature of the useful model is that it is specially oriented to the input signal, which, along with the useful magnetic field of the heart, has significant magnetic interference, mainly a strong interference of the industrial frequency (Fig. 1). Such a signal is registered when ICG examinations are carried out without magnetic shielding. Another aspect is that the useful model av! focused on the use of a cheap 7-channel magnetocardiograph, which reduces the cost of the examination.

The proposed method of processing magnetocardiographic signals consists of 5 stages: - Y. Cleaning of magnetocardiographic signals from interference; aw! 20 PL. Synchronization of magnetocardiographic signals in all magnetocardiographic channels of the magnetocardiograph using a reference electrocardiogram and splitting of magnetocardiographic signals

With cardiocycles;

I. Removal of cardiocycles, including unremoved inharmonic interference;

IM. Selection of cardiocycles synchronous with certain electrocardiographic patterns; 29 M. Summation of all cardiocycles selected on different groups. c Y. Cleaning magnetocardiographic signals from interference. This stage consists in sequential implementation of: - digital filtering, - adaptive interference compensation (ACC).

Filtering consists in the use of software filters of low (LF), high (HF) frequencies and 60 narrow-band filters. In this implementation, by default, a low-pass filter with a cut-off frequency of 50 Hz is used, and a high-frequency filter with a cut-off frequency of 00.5 Hz, the orders of which can be selected by the operator. As a result, interference whose frequencies lie outside the spectrum of the useful MCG signal is weakened, and the useful signal is not distorted (in Fig. 2 - it can be seen that after filtering the amplitude of the industrial frequency has significantly decreased).

However, the specified filters are effective only in relation to harmonic interferences that lie below (device zero drift, quasi-stationary fluctuations of the Earth's magnetic field, quasi-stationary currents in the human body and on its surface) and above (industrial interference and its harmonics) the specified cutoff frequencies. Therefore, after LF and HF filtering, they remain; 1) interferences lying in the useful signal band (low-frequency mechanical resonances in the patient's bed and industrial frequency subharmonics); 2) inharmonic disturbances - jumps and single pulses.

Obstacles of type 1) remain, because they lie in the transmission band of low frequency and PV. To remove them, narrowband filters are provided in this implementation, the parameters of which (central frequencies and suppression bands) can be selected by the user within wide limits. However, they should be used with care, because they can 7/0 distort the useful signal. Therefore, in this useful model, it is proposed to first carry out processing in automatic mode, and if after averaging harmonic interference remains, then study the signal spectrum, and with its help set the parameters of narrow-band filters, and then perform filtering again.

For example, Fig. 3 shows the ICG signals after automatic processing, where uncompensated harmonic interference is visible at most spatial points, the spectral analysis of which shows that it is an industrial subharmonic /5 with a frequency of about 12 Hz.

Another way to reduce harmonic interference is to use additional smoothing. Therefore, in this processing method, if the above software filters did not bring the expected effect or their use is impractical, it is suggested to carry out additional smoothing of the ICG curves over time in manual mode.

This can happen, for example, when the harmonic interference lies in a wide frequency band.

However, software filters cannot completely suppress harmonic interference in the useful signal band, because they have a limited accuracy of setting coefficients, cut-off frequencies and bandwidths, and their effectiveness depends on the order of the filter. Of course, increasing the order of the filter, widening the suppression bands, lowering the cut-off frequency of the FNHF and increasing it in the FHF will lead to a decrease in the level of interference. But these measures have limited application due to the possible distortion of the useful signal carrying diagnostic information. Therefore, it is recommended to use the filter parameters recorded in the program by default, and to make changes to filter parameters aimed at stronger interference suppression only after checking that such changes will not lead to distortion of diagnostic information.

Interferences of type 2) - inharmonic impulse interferences, remain because they are poorly filtered, as their frequency spectrum lies in a wide frequency band. Another reason for the appearance of jumps is internal - if the amplitude "- zo and/or the phase of the external harmonic interference change during the recording quite quickly, including jump-like, then after filtering these changes generate slow variations (trend) and jumps of the zero line o

ICG channels (as well as in their synchronous variations of reference signals). This is a parasitic effect, the reason of which "- lies in the fact that the specified software filters have constant coefficients, that is, they are stationary, and therefore - they cannot adequately filter non-stationary signals. at

Therefore, in this implementation, the ACP procedure is used to remove unfiltered interference of both types specified above. Its essence is known and consists in calculating the weighting coefficients (ACC coefficients) of the reference channels and subtracting the weighted sum of the reference signals from the signals in the ICG channels. IKG channels register a useful signal from the heart and interference, and reference channels - only interference. As a result, there is a subtraction (compensation) of obstacles. In this implementation, 4 ICG and Z reference channels are used, therefore, 12 ACP coefficients are calculated. The adaptability of the procedure consists in the fact that the specified coefficients are calculated at each spatial point, and therefore change over time, adjusting (adapting) to a specific noise situation. )" automatic transmission suppresses obstacles quite effectively. This can be seen from the comparison of the signal before (Fig. 2) and after (Fig. 4) ACP.

Thus, after automatic transmission, there is no impulse interference (jump) and significantly compensated industrial interference, up to the level of the next largest interference 150 Hz, which is already visually visible in Fig. 4. However, the accuracy of the ACP procedure

GI is limited by: - computational accuracy of calculation of ACP coefficients; o - the integral nature of the specified coefficients, obtained, as a rule, as a scalar product of the signals of two (ICG and reference) channels; - limited speed of adaptation, which is limited by the length of signal sampling, the minimum duration of which in this implementation is ЗО0s; as - the non-identity of interference in the ICG and reference channels (mainly due to the fact that their antennas are located at different points in space and the penetration of gradient interference into the antennas of the ICG channels).

As a result, these obstacles cannot be completely destroyed. Therefore, the following procedures are used for their reduction: 1) removal of unwanted cardiocycles - to remove inharmonic interference (stage 3); c 2) averaging - to suppress harmonic interference (stage 5).

I. Synchronization and division of signals into cardiocycles. This stage consists in the sequential implementation of: - synchronization of signals in all ICG channels using a reference electrocardiogram (ECG); 60 - division of signals into cardiocycles.

Before removing unwanted cardiocycles, two auxiliary operations are performed - synchronization of signals in all of them

ICG channels using a reference ECG and splitting signals into cardiocycles. Synchronization consists in time-aligned signals in 4 MKG channels with a synchronous ECG channel. First of all, the strongest, as a rule, are K-waves, which are used to divide it into separate cardiocycles. 65 After that, the time markers corresponding to the K-waves of the ECG are transferred to the IKG channels. The division into separate cardiocycles consists in finding their boundaries and marking them with markers. In this implementation, when setting the boundaries, the ratio of the duration of the cardiocycle before and after the K wave is taken as 4095/6095. However, the exact value of this ratio has no diagnostic value, because the limits of the cardiocycle fall on diastole, where there is practically no electrical activity of the myocardium.

Such a division in this implementation is first performed for the ECG, and then applied to the ICG of the channels. As a result, in 4 magnetocardiograms (each sample with a duration of 30s), synchronizing markers corresponding to K-teeth of the ECG and markers of cardiocycle boundaries corresponding to one heartbeat period are automatically placed. At the same time, extreme cardiocycles are cut off, because they are usually incomplete and may have artifacts associated with transient processes in the recording equipment. This stage is performed automatically and does not 7/0 require user intervention.

I. Removal of unwanted cardiocycles. This stage consists in sequentially performing: - automatic removal of unwanted cardiocycles, - manual review of cardiocycles.

Unwanted cardiocycles are those that include unremoved inharmonic interference. Their removal is a 7/5 Mark, which removes them from the time sample. In this useful model, it is proposed to remove the indicated cardiocycles in two stages. The first is automatic deletion that does not require user intervention. In this mode, it is possible to remove obstacles, the description of which can be easily formalized in the form of a mathematical expression. These include interference whose amplitudes exceed the amplitude of the useful signal. Then they are easily recognized if you set a check on the condition - whether the 2o bignal in this cardiocycle exceeds the specified range. If this condition is met, then such a cardiocycle is deleted.

But there are situations that make it impossible to correctly remove obstacles in automatic mode. The program has two types of errors (they are known as errors of the 1st and 2nd kind) - either a real obstacle is not removed, or a signal related to (possibly pathological) heart activity is mistaken for an obstacle. In this implementation, these errors are as follows: 1) when the interference has amplitudes that do not exceed the useful ICG signal. Such obstacles are difficult to formalize and remove automatically. Figure 5 shows 36 MKG signals after processing, from which it can be seen that the obstacle in the form of a jump of relatively small amplitude was not automatically removed at the spatial point 4-14. "- zo 2) When interferences have amplitudes that exceed the useful IKG signal, but there is no certainty that this impulse or signal jump is not related to the activity of the heart. at

The presence of the situations described above is not known a priori, therefore, in this useful model, it is suggested to always "- use the review of cardiocycles in manual mode. At the same time, the user must make a decision himself, which, accordingly, consists of: o 1) additional removal of synchronous cardiocycles in all 4 ICG channels with such interference, if they simultaneously (synchronously) take place in at least one of the ICG channels and at least in one of the reference channels.

For example, in Fig. b it is necessary to remove two middle ICG cardiocycles, because they include jumps (shown by arrows), which are absent on the ECG. 2) additional inclusion of synchronous cardiocycles in all 4 IKG channels with unrecognized signals (pulses or jumps), if they simultaneously (synchronously) take place in at least one of the IKG channels and the ECG channel.

The meaning of removing the cardiocycle in case 1) is that if some signal is correlated in the ICG and )» reference channels, it means that it is generated by external interference sources and not by the heart. The point of leaving in the sample a cardiocycle that has some unrecognized signal (that was falsely recognized as interference by the program) is that if some signal is correlated in the ECG and ICG channels, it means that it is generated

GI with the heart, not external sources of obstacles.

IM. Selection of cardiocycles. As it was noted above, even after ACP, uncompensated harmonic interference may remain in the MCG signal of the channel (along with pulse signals) (Fig. 4). For their further reduction, an averaging operation was used. But before the actual averaging, it is necessary to correctly select cardiocycles, 5p which belong to certain types of rhythm. This stage consists in: o - accumulation of representative ECG complexes reflecting a certain type of heart rhythm (so-called heart patterns); - classification of ECG cardiocycles according to the specified patterns, - selection of IKG cardiocycles synchronous with a certain ECG pattern (the so-called selection of IKG cardiocycles).

The necessity of this stage is due to the well-known fact that different pathologies of the heart are manifested in different ways on the ECG. These manifestations have a characteristic shape of the ECG curve (type of rhythm), which differs from the norm (sinus rhythm c - Fig. 7). At the same time, in the patient, as a rule, the pathological rhythm can coexist with the sinus rhythm. To analyze the pathological sources in the myocardium that generate each of the specified pathological rhythms, it is necessary to have a set of appropriate templates - representative (standardized) ECG complexes that reflect a certain type of heart rhythm. Therefore, the first stage of averaging is the accumulation of such standardized ECG complexes. In this implementation, this is implemented as a learning mode, when the user himself selects an arbitrary complex of a certain type of rhythm of the reference ECG as a template for the corresponding rhythm. In this useful model in order to take into account different types

ICG complexes provide for the selection of several ECG templates.

When such a set of templates is collected, it becomes possible to proceed to the next stage - classification. It consists 65. in the division of all ECG complexes of this sample into groups related to a certain type of rhythm. In this implementation, this is done by comparing the degree of closeness of a certain ECG cardiac cycle with a set of templates.

At the same time, the cardiocycle belongs to the group corresponding to the template, the degree of closeness to which is maximum.

This stage occurs automatically without user intervention, and all ECG cardiac cycles are marked with numbers corresponding to the number of a certain ECG pattern. Figure 8 shows an ECG signal with Z cardiocycles related to sinus rhythm, presystolic state and extrasystole. It can be seen that the automatic selection correctly assigned the indicated cycles to the groups, respectively, number 1, 3, and 2. This indicates that the set of ECG patterns includes the indicated ECG patterns under the corresponding numbers.

Next is the proper selection of the ICG of cardiocycles - the selection of the ICG of cardiocycles synchronous with a certain ECG pattern.

This also happens automatically in all ICG channels. As a result, all IKG cardiocycles are divided 7/o into groups, the number of which is marked with a number corresponding to a certain ECG pattern.

However, type 1 and type 2 errors are also possible in automatic selection. When selecting an ECG, these errors are as follows: 1) Type 1 errors are when some ECG cardiocycles are assigned to a "foreign" template. 2) Errors of the 2nd kind are when some "foreign" ECG cardiocycles are assigned to "our" template.

The reason for errors of the 1st kind on the ECG - when the patterns are very similar, can in principle be reduced to an arbitrarily low level due to the improvement of mathematical methods for calculating the similarity of cardiocycles. There are two causes of type 2 errors on the ECG: the first is similar to the cause of type 1 ECG errors. The second reason is that in this sample there is a type of rhythm on the ECG that is absent in the set of ECG templates. This reason cannot be eliminated by software and requires adding a new template manually.

When selecting ICG, these errors are formally analogous to those on the ECG: 1) Errors of the 1st kind - some ICG cardiocycles are assigned to a "foreign" template. 2) Errors of the 2nd kind are when some "foreign" ICG cardiocycles are assigned to "our" template.

The cause of errors of the 1st kind on the ICG are the corresponding errors on the ECG, and therefore they are automatically corrected during the correction of the ECG of the errors of the 1st kind and do not require the intervention of the operator. Fig. 9 shows an example of such an error, where it is VISIBLE THAT the second IKG cardiocycle (upper curve) is mistakenly assigned to the Mo2 pattern, while it belongs to the sinus rhythm (Me1 pattern). The reason lies in the corresponding error in the attribution of the second ECG to the cardiac cycle, which differs from the others by a quasi-constant displacement of the zero line. Such a shift is a parasitic effect in ECG measurements (change in skin resistance, skin-electrode contact, skin-galvanic potentials, etc.) and has nothing to do with heart activity. There are also two reasons for errors of the 2nd type according to the ICG: the first is similar to the errors of the 2nd type of the ECG (see above - two reasons). The second reason for MKG errors of the 2nd kind is that there is a pathology in the myocardium in which there are changes that are reflected on the MKG, but are not reflected on the ECG. This reason is due to the advantage of ICG in comparison with ECG and can be corrected only in manual mode.

The presence of the above-described errors is not known a priori, therefore, in this useful model, it is suggested to always o 5 Use cardiocycle selection in manual mode. At the same time, the user must identify selection errors himself and make a decision, which, accordingly, consists in: 1) additional deletion of ECG cardiocycles mistakenly assigned by the program to "other" templates and including them in groups of "own" templates - in case the templates are different , but their morphology (shape) is similar. In this way, errors of the 1st kind and partially of the 2nd kind will be corrected on the ECG and (automatically) on the ICG. "

Thus, the cardiocycle in Fig. 9 should be removed from group 2 and included in group 1, which describes sinus rhythm. sv c 2) Additional inclusion of a new pattern (s) - if there is an ECG type of rhythm, the morphology (form) of which is significantly different from the pattern to which it was mistakenly assigned. In this way, type 2 errors will be corrected completely on the ECG and partially on the ICG.

C) Additional selection of a separate set (subgroup) of IKG cardiocycles, within a certain ECG template - in case there is an IKG type of rhythm, the morphology (shape) of which is significantly different from the ECG template to which it was mistakenly assigned. In this way, errors of the 2nd kind according to the ICG will be completely corrected.

M. Averaging. The averaging procedure consists in summing up all cardiocycles selected at the stage of MI on different groups of cardiocycles synchronized on the B-wave of the ECG. Its meaning is based on: - - imperfection of the heart as a generator, the oscillation period of which has a certain physiological corridor; - sources of external harmonic interference are uncorrelated with each other; o - there are many such sources.

Кк As a result, the interference phases from one cycle to another change almost randomly, so the total interference is stochastic in nature. It is known that the amplitude of the total (averaged) cardiocycle increases proportionally to the number of members of the sum, and the amplitude of the stochastic interference - as the square root of the specified number. After such a procedure, the signal-to-noise ratio of the averaged cardiac cycle in the square root of the number of summed cardiac cycles is greater than this ratio in the signal to averaging. This stage consists in: c - summation of cardiocycles; - generation of averaged ICG files corresponding to certain patterns.

The summation of cardiocycles consists in the summation at each spatial point of synchronous ECG and MCG of cardiocycles, because separately in the ECG channel and in each MKG channel, and at each point - separately for a certain type of rhythm (pattern). In this implementation, summation takes into account different cardiocycle durations. As a result, for each pattern that was distinguished on the ECG, we will get 1 ECG and 36 averaged IKG cardiocycles. Based on this set of averaged cardiocycles corresponding to a certain pattern, a file is generated. Thus, in this useful model, for the purpose of separate diagnosis of each of the rhythm types corresponding to certain patterns, the generation of several averaged ICG files is provided. This step is performed automatically and does not require user intervention. Figure 10 shows the averaged ECG and MCG at point 5-4 of cardiocycles, which correspond to extrasystole and demonstrate the required quality of processing, which consists in the absence of magnetic interference.

For convenience in the described implementation, it is recommended to first perform the processing in automatic mode.

Then, in this useful model, it is necessary to perform quality control of the processing by reviewing all 36 ICG signals in the format of 6Xb curves. Then, for each spatial point, he must carry out processing in manual mode by performing the following actions: 1) if the processing quality is visually determined to be sufficient, then sequentially perform the actions provided for in stages I and IM). 2) if the quality is insufficient - in addition, first perform the actions provided for in stage I). 70 The need for certain actions depends on the nature of residual obstacles or errors during the automatic culling of cardiocycles and/or their selection according to the order described in the specific implementation of the indicated processing stages. The number of their repetitions is determined when the residual interference level does not distort the signal shape. For example, Fig. 3 indicates the need to configure a narrow-band filter at 12Hz, and Fig. 5 - the need to manually remove jumps at point 4-1.

If the level of interference is so great that repeated processing in manual mode does not provide the required signal quality, it must be re-registered. Thus, the advantages of this method in comparison with the current state include the fact that it is focused on the processing of ICG signals registered without the use of a magnetically shielded room (MEK). At the same time, the main problems for processing are created by the strong magnetic pollution of modern medical institutions, which take place in large cities. The reliability of the proposed method of processing magnetocardiographic signals in relation to the negative impact of magnetic interference allows the examination to be carried out in the conditions of a regular clinic without the use of an expensive MEK, and thus reduces the cost of the examination, which is especially relevant in the conditions of Ukraine.

A specific implementation of the method in the utility model is described in detail for purposes of illustration. It is clear that in practice, people experienced both in the processing of ICG signals, and in the processing of ECG and signals of other medical nature, as well as in ICG technology in general, can make some changes and modifications to the proposed procedure. However, if these modifications are made without significant deviations from this utility model, they are subject to this utility model.

Claims (1)

The formula of the invention -- "in) The method of processing magnetocardiographic signals, which includes the cleaning of magnetocardiographic signals from "- interference, which is distinguished by the fact that it additionally synchronizes the magnetocardiographic signals in all magnetocardiographic channels of the magnetocardiograph using the reference electrocardiogram "2 zv And the division of magnetocardiographic signals into cardiocycles, removal of cardiocycles that include non-removed inharmonic interference, selection of cardiocycles synchronous with certain electrocardiographic patterns, and summation of all selected cardiocycles in different groups. - and" name) (in) - ((c) - 60 b5