FR3038508A1 - METHOD FOR CHARACTERIZING ELECTROPHYSIOLOGICAL ACTIVITY - Google Patents

Abstract

The invention is a method for characterizing a periodic electrophysiological activity using magnetometers, at least one of which is capable of measuring a magnetic field induced by said electrophysiological activity. The signal produced by each sensor is the subject of a treatment, the main stages of which are: the decomposition of the signal acquired by each magnetometer to obtain a component, called the component of interest; the identification, on this component, of a plurality of elementary signals, each elementary signal corresponding to one embodiment of said activity; - the selection of the elementary signals considered as the most representative of said activity.

Description

Method for characterizing electrophycological activity

Description

TECHNICAL AREA

The technical field of the invention is the characterization of an electrophysiological activity by detection of magnetic fields induced by this activity. One application is the characterization of cardiac activity by magnetocardiography.

PRIOR ART

Until now, when it comes to evaluating myocardial activity for diagnostic purposes in clinical cardiology, electrocardiography (ECG) is the most common measure. This technique consists in detecting the electrical potentials generated by the cardiac electrical activity. But this electrical activity also induces magnetic fields. Although being of very small amplitude - of the order of 10 '10 T at 10'14 T, to be compared with the amplitude of the terrestrial magnetic field, of approximately ΙΟ "4 T - the fluctuations of these magnetic fields This is the subject of magnetocardiography, often referred to by the acronym MCG, which, like magnetoencephalography (MEG) for the brain, is used to analyze the electrical activity of the myocardium. is a non-invasive, non-contact analysis technique that emits no radiation or magnetic field.

Although relatively old, this technique has recently benefited from advances in the detection of magnetic fields, in particular thanks to developments in the field of superconductivity or magnetic insulation. As a result, although it remains largely confined to research, its clinical applications are developing.

Recent studies have shown that magnetocardiography may be of major clinical interest in the early diagnosis of certain pathologies, such as acute coronary syndromes, ischemic myocardia, or ventricular arrhythmia.

Moreover, the magnetocardiography has certain advantages over electrocardiography, in particular a higher sensitivity, for a similar specificity, in the diagnosis of arrhythmias and coronary heart diseases. It also offers the possibility of constructing spatio-temporal maps of cardiac activity, making it possible to quickly locate an abnormal functional myocardial zone. Another advantage is that the GCM is a native record of cardiac activity, not requiring a baseline measurement. In addition, GCM can be applied to the diagnosis of fetal heart disease, where the ECG is not well adapted due to the confinement of fetal electrical signals by uterine and abdominal tissues.

However, the signals produced by MCG are often noisy and research to remedy this disadvantage is made. In this context, the inventors have invented a simple and efficient method that can be applied to signals derived from magnetometers in order to allow a reliable characterization of the electrophysiological activity studied.

SUMMARY OF THE INVENTION

An object of the invention is a method for characterizing a periodic electrophysiological activity, comprising the following steps: i) acquisition, by a first magnetometer, of a first signal, comprising a measurement of a magnetic field induced by said electrophysiological activity periodic; ii) from said first signal, extracting a component of interest representative of said induced magnetic field; iii) identifying, in said component of interest, a plurality of successive elementary signals, each elementary signal corresponding to a period of said electrophysiological activity; (iv) taking into account a basic reference signal; v) for each elementary signal identified in step iii), calculating an indicator representing a similarity between said elementary signal and said elementary reference signal; vi) according to the indicator calculated for each elementary signal, selection of at least a part of said elementary signals.

In this way, said selected elementary signals are considered to characterize said electrophysiological activity.

According to one embodiment, step i) also comprises the acquisition, by a second magnetometer, of a second signal, preferably measuring a magnetic field representative of an environment in which said first and second magnetometers are placed. . Step ii) can then comprise a decomposition of said first signal into a plurality of components, including at least one component of interest, this decomposition being performed using said second signal, by independent component analysis or by differential measurement.

According to one embodiment, step i) comprises acquiring, by a plurality of magnetometers, first signals, each first signal comprising a measurement of a magnetic field induced by said periodic electrophysiological activity; step ii) is performed using said first signals.

According to one embodiment, step iv) is carried out using a set of elementary signals identified in step iii), the elementary reference signal being obtained, for example, from an average, or another statistical magnitude such as a median, elementary signals forming said set. Said elementary reference signal may, for example, be an average of said elementary signals of said set.

According to one embodiment, the reference elementary signal is previously stored in a memory.

According to one embodiment, the indicator calculated during step v) comprises, for each elementary signal, a correlation coefficient between said elementary signal and the elementary reference signal.

According to one embodiment, the indicator calculated during step v) results from a classification of each elementary signal, said classification being performed between at least one non-representative class of said reference signal and a representative class of the reference signal. .

The method may comprise a step vii) of determining a resultant signal, representative of the elementary signals selected during step vi), this signal characterizing said electrophysiological activity. This resulting signal can for example be obtained by an average of the selected elementary signals, or by an average of these signals weighted by the similarity indicator associated with each selected elementary signal, determined during step v). The electrophysiological activity characterized can be the activity of the myocardium.

Another object of the invention is an information recording medium, comprising instructions for the execution of a method as described above, these instructions being able to be executed by a processor, in particular a processor. microprocessor. This microprocessor can then be connected to a memory in which said instructions are stored.

Another subject of the invention is a device for characterizing a periodic electrophysiological activity, comprising: a first magnetometer, able to detect a magnetic field induced by said electrophysiological activity; a processor configured to implement a method as previously described.

The device may also include a second magnetometer, capable of detecting a magnetic field in an environment in which said device is placed.

FIGURES

FIG. 1 represents a device suitable for implementing the invention.

FIG. 2 represents the main steps of a method according to one embodiment of the invention.

FIG. 3 represents the main steps of a method according to a variant of the embodiment represented in FIG. 2.

FIG. 4A represents an experimental result, showing a signal representative of a cardiac activity obtained following an acquisition whose duration is 10 minutes.

FIG. 4B represents an experimental result, showing a signal representative of a cardiac activity, similar to that analyzed with reference to FIG. 4A, this signal being obtained following an acquisition whose duration is 30 seconds.

DESCRIPTION OF PARTICULAR EMBODIMENTS

FIG. 1 represents a device 1 for characterizing a periodic electrophysiological activity. The term electrophysiological activity refers to an electrical activity of an organ, for example the heart or the brain. Periodic means that said activity is reproduced in time, according to a determined period. This period may in particular be between 100 ms to 5 minutes.

In the example shown, we seek to characterize the electrical activity of the myocardium, this term designating a muscle tissue constituting the heart of a subject under examination. To do this, the device comprises a first magnetometer 10 disposed near the skin P of said subject, so as to minimize the distance between the member O (in this example the heart) and the magnetometer 10. The first magnetometer 10 n is not necessarily in contact with the skin. It may be disposed a few millimeters from the latter, for example 5 mm, and preferably at a distance of less than 1 cm.

The first magnetometer 10 is able to acquire a first signal li, representative of the intensity of a first magnetic field B1 along one or more axes. In the example under consideration, the intensity of the magnetic field is determined along an axis y oriented in a direction normal to the skin P of the subject under examination. Such a magnetometer is for example a SQUID magnetometer (acronym for Superconducting Quantum Interference Device) or an optically pumped magnetometer. For example, it is a magnetometer with optical pump Helium 4, this type of magnetometer being described in the application WO2013092383.

The first magnetometer 10 is connected to an electronic formatting circuit 14, ensuring the amplification of the first signal and / or its digitization, then it is transmitted to a processor 20, either by wire connection or by wireless link, or else via a recording medium of the CDROM type, USB key ... The processor 20 is connected to a memory 22 comprising instructions, the latter being able to be executed by the processor 20, to implement the method shown in Figures 2 or 3, and described below. These instructions can be saved on a recording medium, readable by the processor, hard disk type, CDROM or other type of memory. The processor may be connected to a display unit 24, for example a screen.

The device 1 preferably comprises a second magnetometer 12, held at a distance from the first magnetometer 10. As previously described, the first magnetometer has the function of detecting an intensity of a magnetic field generated by the physiological activity to be characterized. But the first signal measured by the first magnetometer also has a parasitic component related to the ambient magnetic field of the environment in which the device 1 is placed. The function of the second magnetometer is to detect only, or at least a very large majority, this field. ambient magnetic. It is then held at a sufficient distance from the organ generating the electrophysiological activity, so that its exposure to the magnetic field induced by this activity is negligible.

This second magnetometer 12 is for example placed at a distance greater than 10 cm from the first magnetometer 10. It is connected to the electronic shaping circuit 12 as well as to the processor 20. The second magnetometer is able to acquire a second signal h, representative of the intensity of a second magnetic field B2, along the same axis, or the same axes, as those considered during the measurement of the first magnetic field.

Thus, the first magnetic field B1 whose intensity is measured by the first magnetometer 10 is such that

:

Ba represents the magnetic field induced by the electrical activity of the organ O. This field is of low amplitude, typically between 10 '10 T at 10'14 T when the organ is a human heart.

Benv represents the ambient magnetic field, in the environment in which are placed the first magnetometer and the second magnetometer.

In this example, the second magnetic field B2, whose intensity is measured by the second magnetometer 12, is such that B2 "Benv,

Figure 2 shows the main steps of a method according to the invention.

Step 100: Data Acquisition During this step, each magnetometer acquires a signal whose intensity is representative of the magnetic field at said magnetometer, and in particular one of its components along one or more axes. In this example, where two magnetometers are used, two signals li (t) and h (t) are obtained representative of the respective intensity, at time t, of the first magnetic field B1 and the second magnetic field B2, according to y axis previously mentioned.

Step 110: Preprocessing the data. The signal delivered by each magnetometer is preferably pretreated prior to step 120 of extracting a component of interest. It is first of all a digitization of the signal, the time sampling frequency being for example 2 kHz. This pretreatment may also include a substep aimed at centering each signal with respect to its average value. In other words, we subtract from each signal, sampled in time, an estimate of the average value of this signal.

The pretreatment may also include the application of frequency filters, so as to select certain spectral bands. For example, a band-pass filter, whose bandwidth extends between 0.5 Hz and 45 Hz, as well as a Notch-type filter, can be applied to reject a narrow spectral range of a few Hz or less, located around the frequency 50 Hz.

Steps 110 to 170 can be implemented in post-processing, that is to say following the acquisition of all the data, or during data acquisition. In the latter case, there is a learning period, in which the acquired data are used to establish the quantities necessary for the data processing, in this case the average, the variance, or the matrix W described in link with step 120.

The pretreatment may also include a data whitening substep, i.e., process each h and l2 signal, so as to minimize the covariance between the signals from different magnetometers, and to make the variance of each signal equal to 1. / '(t) As a result of this pretreatment step, a measuring vector I (t) = ^ is available.

'2 (where each component is a reduced centered temporal variable.

Note that this sub-step of bleaching is optional. It is carried out if step 120 involves the implementation of an Independent Component Analysis algorithm, as described below.

It will be understood that these pre-processing operations performed during step 110 are adapted to the method of extracting a component of interest applied during step 120.

Step 120: Extraction of a component of interest. This step aims to break down each term of the measurement vector into a linear combination of components statistically independent of each other. A first component Ci is representative of the magnetic field Ba induced by the electrophysiological activity, and a second component C2 is representative of the ambient magnetic field Benv previously described. This step can be performed by an Independent Component Analysis algorithm, frequently referred to as ICA (Independent Component Analysis). This method, known to those skilled in the art, allows a decomposition of the measurement vector I (t) such that: l (t) = M × C (t), M being a weighting matrix and C (t) being a vector of components. So,

Μι, ι, Μι, 2, Μ2, ι and Μ2,2 being the terms of the weighting matrix M, Ci (t) and C2 (t) denoting said components. The independent component analysis algorithm aims at determining an extraction matrix W, such that C = ΜΛχ I. This type of algorithm is easily implemented using mathematical calculation software, for example the software well known Matlab®.

Note that step 120 does not necessarily implement an independent component analysis. For example, it is possible to make differential measurements from the signals of each magnetometer, so as to determine the first component Ci. The latter can then be obtained on the basis of a difference between the signals delivered by each magnetometer, in this case a subtraction of the second signal b to the first signal li.

However, an independent component analysis is preferable because it allows better discrimination of the component of interest with respect to noise.

Thus, step 120 makes it possible to extract, from the first and second signals li and I2, a component of interest, in this case the first component Ci, representative of the electrophysiological activity to be characterized.

Step 130: Time segmentation. Since the physiological electrical activity is periodic, the component of interest Ci extracted during step 120 is also. Step 130 is a step intended to identify, on said component of interest, time periods, each time period corresponding to one embodiment of the physiological electrical activity. In this example, applied to the myocardial activity, each period includes a local maximum, corresponding to contraction of the ventricular muscle. This contraction generates three waves, commonly designated by the respective terms Q, R, S, or QRS complex, the R wave being the one having a peak amplitude of the highest, and this in each period. Also, by applying an amplitude thresholding on the component of interest Ci, it is possible to identify instants ίζ, ... t £, ... t $, each instant t £ corresponding to the peak of a wave R an nth period, n being an integer between 1 and N, N denoting the number of periods included in the analyzed interest component.

The value of the threshold can be determined a priori, or manually, for example according to a graphical representation of the time course of the first component on the screen 24, or automatically, from an estimate of the standard deviation σ of the first component Ci, the threshold then corresponding to a weighting of this standard deviation, for example 2σ or 3σ. From each instant tf1, we can define a start of period period - δ ^ and an end of period period t% + δί2, δ ^ and ôt2 corresponding to predetermined durations. For example, 5 ^ = δί2 = 300 ms. These durations are determined a priori, based for example on the heart rate of the subject under examination.

An elementary signal Sn, representing an nth period of Ci, is then extracted, such that

with: T = Δt + δί2, T denoting the duration of an elementary signal, determined according to the period of the cardiac activity, t i being the initial moment of each elementary signal.

Thus, each elementary signal Sn is defined according to a time range between t = t1 = 0 and t = T. T is in this example equal to 600 ms.

This step 130 makes it possible to segment the component of interest Ci according to a sequence of N elementary signals Sn, each elementary signal being representative of a period of rank n of the electrophysiological activity to be characterized.

Step 140: Obtaining an elementary reference signal. This step consists in obtaining an elementary reference signal Sref, considered as representative of the electrophysiological activity studied. This elementary signal may be stored in the memory 22. It may be a signal established on the basis of a previously memorized model or signal corresponding to the subject under examination.

Preferably, in connection with the variant represented in FIG. 3, the reference elementary signal Sref is obtained during a step 135, considering a set of P elementary signals Sn identified in the component of interest Ci. This set may for example consist of the P first identified elementary signals, or by the set of identified elementary signals, P being an integer less than or equal to N. The elementary reference signal is then obtained by performing an average of these P elementary signals , in such a way that :

where mean means the average operator. P is preferably greater than 10, or even greater than 30. The component of interest may comprise a sequence N elementary signals, whose first P elementary signals, in chronological order, are used to establish the reference signal Srep the signals The following elemental elements are used to characterize electrophysiological activity. This allows a characterization of the electrophysiological activity simultaneously with the acquisition of the measurements. P may be equal to N, in which case the reference signal Sref is progressively updated as the number of elementary signals increases. It is updated constantly.

Alternatively, the characterization is performed post-processing, in which case all the elementary signals can be used to establish the reference signal. Other statistical quantities than the average can be used to calculate the reference signal, for example by using the middle operator instead of the average operator.

Step 150: calculation of an indicator expressing a similarity between each elementary signal S "and the reference signal Sref. The similarity can be a correlation. The indicator can then be a correlation coefficient rn, expressed for example according to the expression:

where ti denotes each instant resulting from the temporal sampling of the signals Sn and Srep, these signals being sampled between the instants tt = ti and tt = T.

The denominator of this expression is an optional term of normalization. Other methods than calculating a correlation coefficient are possible. For example, each elementary signal may be classified so as to retain only those most correlated to the elementary reference signal. The classification makes it possible to classify each elementary signal Sn according to at least one class representative of the reference signal Sref and a class that is not representative of this reference signal. Such a classification can for example result from the application of an algorithm for reducing the dimensionality of each elementary signal, for example by principal component analysis. It is then possible to establish a distance between each elementary signal Sn and the reference signal Srep in a space defined by basic vectors determined by the principal component analysis. Such a distance constitutes a classification parameter, according to which an elementary signal is or not considered to belong to a class representative of the reference signal. From this classification, the similarity indicator rn associated with the elementary signal Sn is established from the distance, in which case the smaller it is, the more the elementary signal is considered to be close to the reference signal.

Alternatively, the similarity indicator may take the value 1 when the elementary signal is classified in the representative class of the reference signal; it takes the value 0 otherwise.

Also, in general, the similarity indicator represents a similarity between the elementary signal considered and the reference signal. This indicator can then be: a correlation coefficient, expressing a correlation between the elementary signal Sn considered and the reference signal Sref. a membership in a class representative of the reference signal Srep following a classification of the elementary signal Sn considered.

Step 160: selection of the elementary signals Sn as a function of the value of the indicator calculated during step 150, in order to retain only the elementary signals best correlated with the reference signal, and considered as characterizing the electrophysiological activity studied. Indeed, although an amplitude thresholding is previously applied during step 130, the inventors considered that it was preferable to make a selection, based on the indicator of similarity with the reference signal. This makes it possible to improve the characterization of the electrophysiological activity, by rejecting signals considered as not representative of this activity, for example signals presenting artifacts.

The selection can be made on the basis of a thresholding with respect to the value of the correlation coefficient rn defined above, the elementary signals Sn, for which the correlation coefficient rn exceeds a certain threshold rthreshold, being selected.

The value of the threshold rthreshoid can be predetermined. It can also be adjusted automatically, so as to retain a certain percentage of elementary signals.

Step 170: Characterization of the electrophysiological activity on the basis of the selected elementary signals. During these steps, the selected elementary signals, denoted by may be averaged, so as to provide a resultant signal Soutput that is considered to characterize the electrophysiological activity. Thus, Soutput (t) = mean (S ^ (t)), mean means the mean operator. The calculation of the resulting signal Soutput can also be performed by averaging the selected elementary signals S ^, weighted by the similarity indicator associated with them rn, so as to modulate the weight of each selected elementary signal in the calculation of the signal. resulting Soutput, according to this indicator.

Tests carried out:

Comparative tests were carried out in order to highlight the advantages generated by the invention. Acquired magnetic signals generated by the myocardium were made by implementing, on a healthy subject, a device as shown in Figure 1. The first magnetometer is placed 5 mm from the skin of the subject and the second magnetometer is placed at a distance of 10 cm from the first magnetometer. The tests are carried out inside a magnetically isolated enclosure.

In a first test, the acquisition time is 10 minutes, the sampling frequency being 2.056 kHz.

FIG. 4A represents a resultant signal Soutput, characterizing the electrophysiological activity, the latter being established from a obtained average: curve 1: from 88 elementary signals Sn, identified during the amplitude thresholding applied during the step 130, without implementation of steps 140,150 and 160. In other words, the elementary signals identified in step 130 are all considered representative of the electrophysiological activity. curve 2: from 119 elementary selected Sen signals implementing the method shown in Figure 3, the reference signal being determined from the set of selected elementary signals. The x-axis designates the time t, the unit being the second.

It is specified that the value of the threshold applied during step 130 has been adjusted between the acquisitions respectively corresponding to curve 1 and curve 2, so as to obtain a resulting signal calculated on the basis of a comparable number of elementary signals.

Curve 2 is better resolved. The amplitudes of the QRS complex are more marked, which allows a more precise estimation of its duration. Moreover, the P and T waves are sharper than on curve 1. The electrophysiological activity is therefore characterized with more precision.

FIG. 4B represents the results obtained considering only 30 seconds of recording. The curves are similar to those shown in relation to FIG. 4A, the only difference being the number of elementary signals considered: in curve 1, 11 elementary signals S "identified for 30 seconds have been averaged, without implementation of the steps 140 , 150, 160. On the curve 2, we have averaged 14 selected signals S ^, after implementation of the method shown in FIG.

Note that the curve 2, implementing the invention, allows a better definition of the QRS complex as well as P and T waves, in comparison with the curve 1. This result is remarkable for the small number of elementary signals considered to obtain a satisfactory representation of the electrophysiological activity. This suggests an acquisition time of a few tens of seconds, comparable to the acquisition time required to obtain a usable electrocardiogram.

In addition to a better characterization of the electrophysiological activity, the invention allows a significant reduction in the acquisition time, the latter being limited to a few tens of seconds.

Although described in connection with the characterization of the myocardial activity, the invention can be applied to other organs, in particular electroencephalography or the characterization of muscle activity in response to stimulation, for example stimulation trans-cranial magnetic motor cortex.

Moreover, although the example described is made on the basis of two magnetometers, the use of a higher number of magnetometers is quite possible, the step 120 allowing the extraction of at least one component of the magnetometers. interest from the different measured signals. For example, a plurality of magnetometers may be provided, each magnetometer measuring a magnetic field induced by the electrophysiological activity. Step 120 may be performed from the signals recorded by these magnetometers, in order to extract at least one component of interest representative of the electrophysiological activity to be characterized.

Claims (13)

1. A method of characterizing a periodic electrophysiological activity, comprising the following steps: i) acquisition, by a first magnetometer (10), of a first signal (li), comprising a measurement of an induced magnetic field (Ba) by said periodic electrophysiological activity; ii) from said first signal, extraction of a component of interest (CJ representative of the induced magnetic field (Ba), iii) identification, in said component of interest, of a plurality of successive elementary signals (Sn), each elementary signal corresponding to a period of said electrophysiological activity; iv) taking into account an elementary reference signal (Sre ^); v) for each elementary signal (Sn) identified in step iii), calculating an indicator (rn) representing a similarity between said elementary signal and said elementary reference signal (Sref); vi) according to the indicator (rn) calculated for each elementary signal (Sn), selection of at least a part of said elementary signals, said selected elementary signals (S £) being considered as characterizing said electrophysiological activity. 2. The method of claim 1, wherein step i) also comprises the acquisition, by a second magnetometer (12), of a second signal (b), measuring a magnetic field (Bgnv) representative of an environment. wherein said first and second magnetometers are placed. 3. Method according to claim 2, wherein step ii) includes a decomposition of said first signal (h, i) into a plurality of components (Ci, C2), at least said component of interest (C {), this decomposition being carried out, using said second signal (12, 2), by independent component analysis or by differential measurement. 4. Method according to any one of claims 1 to 3, wherein step iv) is performed using a set of elementary signals (Sn) identified during step iii), the elementary signal of reference (Sref) being obtained from an average of the elementary signals forming said set. The method of claim 4, wherein said reference elementary signal (Srej) is an average of said elementary signals of said set. 6. Method according to any one of claims 1 to 3, wherein the elementary reference signal (Sref) is previously stored in a memory (22). 7. Method according to any one of the preceding claims, in which the indicator calculated during step v) comprises, for each elementary signal (Sn), a correlation coefficient (rn) between said elementary signal and the elementary signal. reference (Sref). Process according to any one of Claims 1 to 6, in which the indicator calculated during step v) results from a classification of each elementary signal (Sn), said classification being carried out between at least one non-representative class. said reference signal (Sref) and a class representative of said reference signal. 9. Method according to any one of the preceding claims, comprising a step vii) of determining a resulting signal (Soutput), calculated as a function of the elementary signals (S ^) selected in step vi), this signal resulting thus characterizing said electrophysiological activity. The method of any of the preceding claims, wherein said electrophysiological activity is myocardial activity. 11. An information recording medium, comprising instructions for the execution of a method according to any one of claims 1 to 10, these instructions being able to be executed by a processor (20). 12. Device (1) for characterizing a periodic electrophysiological activity, comprising: a first magnetometer (10) capable of detecting a magnetic field (Ba) induced by said electrophysiological activity; a processor (20) configured to implement a method according to one of claims 1 to 8. 13. Device according to claim 12, also comprising a second magnetometer (12) capable of detecting a magnetic field (Benv) of an environment in which said device (1) is placed.