Method for determining a patient's risk of cardiac decompensation

Abstract

1. A method (100) for determining a patient's risk of cardiac decompensation, comprising: at least one step (104) of processing an acceleration signal (ACC) acquired from the patient; and determining (114) the risk of cardiac decompensation based on a study of a plurality of markers (115, 117, 119, PEP) comprising intervals (S1, S2, S3) of a processed acceleration signal (ACC-T) obtained from the processing step (104). In particular, the step (104) of processing the acceleration signal comprises a step (106) of dividing the acceleration signal into time cycles (107) of equal duration; a step (108) of identifying intervals (S1, S2, S3) characteristic of heart sounds within each time cycle (107); a step (110) of selecting coherences of the identified intervals; and a step (112) of calculating an average interval.

Description

[Technical Field]

[0001] The present invention relates to the field of medical devices and systems for monitoring the cardiac health of living organisms, and more particularly to a system for determining heart failure decompensation in which an implantable medical device in communication with a computer server is configured to measure cardiac parameters. [Background technology]

[0002] Heart failure is a chronic disease that affects a large proportion of the population, especially those over 60, and can lead to episodes of decompensation, the frequency of which increases with declining cardiac function. Summary of the Invention [Problem to be solved by the invention]

[0003] Heart failure decompensation episodes often involve emergency hospitalization of patients with this heart disease because they are not detected early enough and heart failure decompensation episodes are often asymptomatic at the time of their onset. Thus, the absence of symptoms detectable by the patient means that heart failure decompensation episodes progress without any preventive measures being taken. However, once patients begin to experience early symptoms of heart failure decompensation, such as fatigue, palpitations, or shortness of breath, hospitalization is difficult to avoid.

[0004] It is clear from the above that the earlier an episode of heart failure decompensation is detected, the lower the risk of hospitalization and the costs that may come with it, since early detection of heart failure decompensation in a patient means that the patient's trait can be treated with the prescription of medications that will stabilize their cardiac function.

[0005] Early detection of heart failure decompensation is possible based on detailed analysis of various subclinical cardiac parameters, especially hemodynamic parameters. [Means for solving the problem]

[0006] The present invention is part of this context and aims to provide a system for determining episodes of heart failure decompensation in a patient.

[0007] It is therefore a primary object of the present invention to provide a method for determining a patient's risk of cardiac decompensation, wherein a step of processing acceleration signals acquired from the patient is performed, and determining the risk of cardiac decompensation based on a multiple marker study comprising segments of the processed acceleration signals resulting from the processing step, the method comprising: processing the acceleration signal, dividing the acceleration signal into time cycles of equal time length; identifying, in each time cycle, intervals characteristic of heart sounds by a first search for an interval of a first type in a first defined portion of the time cycle and a second search for at least an interval of a second type and an interval of a third type in a second defined portion of the time cycle; filtering the coherence of the identified intervals for at least a first type of interval and a second type of interval; calculating a first type of average interval, a second type of average interval, and a third type of average interval, wherein the average calculation is based on all of the selected first type of intervals, each of all of the selected second type of intervals, and each of all of the selected third type of intervals; The method is also characterized in that the determining step includes at least one step of determining an increase in amplitude of a first type of interval for the data in the memory and / or a decrease in amplitude of a third type of interval for the data in the memory.

[0008] The method according to the invention therefore makes it possible in particular to determine the risk of decompensation based on heart sounds representative of a given acquisition period, said representative heart sounds being obtained by aggregating heart sounds obtained by processing at least one acceleration signal.

[0009] The acceleration signal is divided into a plurality of time cycles of equal time length, and it should be understood that for the same time length, the time length of a time cycle of an acquired acceleration signal may be different from the time length of a time cycle of another acceleration signal acquired earlier on the same patient, but the time length of a time cycle of the same acceleration signal will be the same for each time cycle at the end of the division process.

[0010] The time length may be defined before the method is performed, or it may be calculated during the method as a function of the characteristics of the time cycle inherent in the acceleration signal being analyzed. In a non-limiting example of the present invention, the acceleration signal is first divided into cycles using one identifying parameter for starting the division of the period and another identifying parameter for ending the division of the period, or the same, so that each time cycle has its own time length reflecting the actual occurrence of these identifying parameters. The time lengths of the time cycles are then averaged in the remaining part of the method according to the present invention to determine the time length to be assigned to the time cycle. The time length of each time cycle obtained by the initial division is then adjusted to be equal to this average time length. Depending on the initial length of the time cycle, the control unit may shorten the time cycle to reduce its length to be equal to the average time length, or may extend the time cycle to increase its length. In the latter case, the control unit is configured to add one or more signal values ​​of zero to the end of the time cycle, the values ​​being added at regular intervals as a function of the sampling frequency. For example, a sampling rate of 500 Hz adds a signal value of zero to the end of the time cycle every 2 milliseconds.

[0011] For each time cycle defined in a given acquisition period, different intervals of the heart sounds are identified, and the acceleration signal is advantageously processed by interval type, in particular the first type of interval on the one hand, and the second and third types of interval on the other hand, with these last two types of intervals being processed simultaneously or in parallel. This has the advantage that useful information about one type of interval can be obtained in a given time cycle, while information about other types of intervals is excluded in that time cycle, thereby increasing the number of intervals of the same type aggregated in the acquisition period. In other words, for a given time cycle, if the first type of interval cannot be identified, especially due to temporary acquisition failure, or if the identified first type of interval appears too different from those observed in other time cycles, it is possible to extract usable information about the second and / or third type of interval.

[0012] According to an optional feature of the invention, the identification step includes a sub-step of excluding intervals previously identified as first, second or third type intervals, and excluding intervals classified as artifacts in this excluding sub-step.

[0013] In particular, such artifacts may occur when signals are acquired simultaneously with patient movement during sleep and / or snoring or coughing, but these examples are not intended to limit the present invention. It will be appreciated that intervals identified as artifacts, i.e., intervals that are far removed from the shape of physiological intervals in either duration or amplitude, are excluded from the coherence selection process, thereby preventing the control unit from wasting calculation time on intervals that would not be retained after the selection process anyway and that could alter the analysis results. This improves the speed of information processing and prevents the coherence selection process from being hindered by intervals that are not characteristic of the patient's condition.

[0014] According to an optional feature of the invention, the exclusion substep comprises calculating a spectral analysis score, and intervals presenting a score greater than a threshold are identified as artifacts and excluded, in particular the spectral analysis score is a z-score and the threshold is equal to 3.

[0015] According to an optional feature of the invention, following the detected second type of interval, a third type of interval is detected in a second portion of the time cycle.

[0016] According to an optional feature of the invention, the third type of sections have a mean amplitude that is less than the mean amplitude of the second type of sections.

[0017] According to an optional feature of the invention, the method of determining includes, prior to the processing step, obtaining at least one acceleration signal from the patient.

[0018] According to an optional feature of the invention, the acquiring step is performed by a three-axis accelerometer and the processing step is performed on each of the signals detected in one of the three axes or on an overall signal obtained by combining the signals acquired in each of the axes.

[0019] According to an optional feature of the invention, the step of obtaining at least one acceleration signal involves obtaining an electrocardiogram signal.

[0020] According to an optional feature of the invention, the signal splitting step includes synchronizing the acceleration signal based on the electrocardiogram signal.

[0021] According to an optional feature of the invention, the identifying step includes calculating an envelope of the signal, and an interval is identified when the amplitude of the envelope is greater than a determined threshold in a defined portion of the time cycle.

[0022] According to an optional feature of the invention, the signal envelope is calculated over the entire cycle.

[0023] According to an optional feature of the invention, the determined threshold is on the order of 10% of the maximum amplitude value.

[0024] According to an optional feature of the invention, the envelope calculation step includes both a squared envelope calculation and an absolute value calculation.

[0025] According to an optional feature of the invention, the first defined portion of the time cycle in the identification step is equal to the first third of the time length of the time cycle, and the second defined portion of the time cycle is equal to the last two-thirds of the time cycle.

[0026] According to an optional feature of the invention, the sorting step includes at least two series of correlation calculations performed in parallel, including a first series of correlation calculations performed on all of the intervals of a first type and a second series of correlation calculations performed on all of the intervals of a second type.

[0027] According to an optional feature of the invention, each series of correlation calculations performed for a given type of interval includes a first calculation of cross-correlations between all intervals identified as this type of interval to define a reference interval for this type of interval, followed by a step of calculating a correlation for each of the intervals identified as this type of interval against the reference interval.

[0028] According to an optional feature of the present invention, the intervals selected in each series of correlation calculations are those having a correlation index with the reference interval that is equal to or greater than a predetermined correlation threshold. As an example, the correlation threshold is at least 50%, such as 60%. In other words, any intervals having a correlation index below the correlation threshold (60% in this example) are ignored and are not considered in the average interval calculation process.

[0029] According to an optional feature of the invention, the sorting step includes only two series of correlation calculations performed in parallel, including a first series of correlation calculations performed on all of the intervals of a first type and a second series of correlation calculations performed on all of the intervals of a second type.

[0030] In this variant, the third type of intervals have already been identified, and no cross-correlation calculations are performed on the third type of intervals. In this case, the coherence-based sorting for the reference interval is performed using only the second type of intervals. The third type of intervals are sorted in the same way as the second type of intervals that exist in the same time cycle, and if the second type of intervals are retained after the coherence sorting process, the third type of intervals in the same time cycle are retained, and if the second type of intervals are excluded after the coherence sorting process, the third type of intervals in the same time cycle are also excluded.

[0031] In other words, signal processing is performed to take into account the third type of interval because it is one of the most specific in determining heart failure, but this signal processing is specific in that the third type of average interval is calculated based on the excluded or retained third type intervals as a function of the exclusion or retention of the second type intervals of the same time cycle, making the identification of this second type interval more reliable.

[0032] The acceleration signal is segmented into three sections, and the first section on the one hand and the second and third sections on the other hand are processed in two separate procedures, with the second and third sections being processed based on cross-correlation and selection based only on the second section.

[0033] According to an optional feature of the invention, the sorting step includes three series of correlation calculations performed in parallel, including a first series of correlation calculations performed for all of the intervals S1 of the first type, a second series of correlation calculations performed for all of the intervals S2 of the second type, and a third series of correlation calculations performed for all of the intervals of the third type. In other words, a cross-correlation calculation is performed for each of the pre-identified intervals.

[0034] According to an optional feature of the invention, between the steps of selecting and calculating the average interval, a step is performed for each type of interval in which the selected interval is reset to the time of the corresponding reference interval.

[0035] According to an optional feature of the invention, the determining step further comprises determining an increase in amplitude of the second type of interval for the data in the memory, such an increase being indicative of pulmonary arterial hypertension.

[0036] Other characteristics, details and advantages of the invention will become more apparent on reading the following description on the one hand and the examples of embodiments given by way of indication and non-limitingly with reference to the attached schematic drawings on the other hand. [Brief explanation of the drawings]

[0037] [Figure 1] FIG. 1 shows the main steps of the method for determining cardiac decompensation according to the present invention. [Figure 2] FIG. 2 illustrates a first embodiment of the method shown in FIG. 1. [Figure 3] 4 shows an acceleration signal divided into time cycles in one of the steps of the method; FIG. [Figure 4] 4 shows the acceleration signal according to FIG. 3 subjected to squared envelope and magnitude envelope calculations which enable the next steps of the method to be implemented. [Figure 5] FIG. 1 shows an average time cycle representative of a processed acceleration signal, with a first average type interval in the first third of the cycle and a second average type interval and a third average type interval in the remaining two-thirds. [Figure 6] FIG. 2 illustrates a second embodiment of the method shown in FIG. DETAILED DESCRIPTION OF THE INVENTION

[0038] First of all, it should be noted that although the figures illustrate the invention in detail for its implementation, these figures may of course be used to better define the invention as needed, and it should also be noted that these figures only show examples of embodiments of the invention.

[0039] The features, variations, and different embodiments of the invention can be associated with one another in various combinations, provided they are not incompatible or mutually exclusive. In particular, it is possible to envision variations of the invention that include only a selection of features described below, in isolation from other features described, if this selection of features confers a technical advantage or is sufficient to distinguish the invention from the prior art.

[0040] In the figures, elements common to several figures retain the same reference numerals.

[0041] FIG. 1 illustrates a method 100 for determining a patient's risk of cardiac decompensation according to the present invention, the main steps of which are processing acceleration signals acquired from the patient and determining the risk of cardiac decompensation based on a study of a plurality of markers comprising the processed acceleration signals.

[0042] FIG. 1 shows in more detail the general flow of the determination method 100, which comprises step 102 of acquiring at least one acceleration signal ACC, step 104 of processing the acceleration signal, which includes at least step 106 of dividing the acceleration signal into time cycles 107 of a determined time length, step 108 of identifying intervals S1, S2, S3 characteristic of heart sounds in each time cycle 107, step 110 of selecting the coherence of the identified intervals, step 112 of calculating an average interval representative of the processed acceleration signal ACC-T, and step 114 of determining the risk of cardiac decompensation.

[0043] Step 102 of acquiring at least one acceleration signal ACC may be performed by any possible acquisition means, as long as the selected acquisition means allows for the acquisition of a signal that is as representative of cardiac activity as possible with as few artifacts as possible, without this choice limiting the invention. For example, to acquire such a signal, it may be desirable to provide a subcutaneous implant at the thoracic level and set it to perform while the patient is asleep, in order to ensure that the acquisition is performed when the patient is not undergoing parasitic movement. Such physiological stability ensures the acquisition of reproducible data, making it possible to determine that any fluctuations observed by the method according to the invention after processing the acceleration signals are due to a pathology.

[0044] The accelerometer used may be a three-axis accelerometer capable of acquiring the same acceleration signal representing cardiac activity in three separate axes over a given period of time. In the following description, the processing of the acceleration signal ACC may be processing of a signal extracted in one axis direction, depending on whether it is obtained by selecting from three axes or by an accelerometer capable of acquiring a signal in one axis direction. However, it should be noted that the description also applies when the processed acceleration signal is a signal obtained by combining values ​​acquired in each of the three axes.

[0045] The acquisition means used are configured to transmit the acquired data to a control unit, which may be integrated into a remote server, the patient's mobile device or a computing device connected to the acquisition means, but this is not a limitation of the invention, and the data may be transmitted by wireless communication protocols or by wired means, if possible depending on the type of acquisition means used and the location of the control unit. The control unit is configured to perform step 104 of processing the acceleration signal ACC.

[0046] More specifically, the step 106 of dividing the acceleration signal consists of evaluating the acquired acceleration signal ACC over the entire acquisition period, which may be, for example, 30 seconds, and dividing it into time cycles 107 of equal time length, which are the same for each time cycle obtained by dividing the same acceleration signal, which are then analyzed in the remainder of the method.

[0047] Means for generating time cycles of the same time length are described in the remainder of the description, and as an example, the time length of the time cycles of the same acceleration signal may be on the order of 300 to 1500 milliseconds.

[0048] As will be explained below, in particular with reference to FIG. 2, step 106 of dividing this acceleration signal may advantageously be associated with simultaneous processing of the acceleration signal ACC and an electrocardiogram signal ECG acquired simultaneously with the acceleration signal, so that specific features of the cardiac activity used to set at least one start point and, where appropriate, an end point of a time cycle 107 of the acceleration signal ACC can be detected by analysis of the electrocardiogram signal ECG, for example corresponding to the appearance of a peak R.

[0049] Once the signal has been divided into different time cycles 107, at least some of these time cycles are analyzed to identify intervals S1, S2, S3 characteristic of heart sounds in an identification step 108. In particular, these characteristic intervals are identified when the amplitude of the acceleration signal, more particularly the amplitude of the envelope of the acceleration signal obtained by the signal rectification operation, is greater than a predetermined value, for example greater than a certain percentage of the maximum amplitude of the envelope observed over the duration of a time cycle 107.

[0050] In particular, in the identification step 108 it is possible to identify at least a first type of interval S1 in a first defined portion of one of the analyzed time cycles 107 and at least a second type of interval S2 and a third type of interval S3 in a second defined portion of this same time cycle 107. Thus, for a given acquisition period, the control unit is configured to, on the one hand, store in a suitable database all of the first type of intervals S1 identified over all time cycles 107, and to distinguishably store in a suitable database the identified intervals S2, S3 in the second portion of each analyzed time cycle 107.

[0051] According to the embodiment described below, the second type section S2 and the third type section S3 may then be processed simultaneously by processing only the second type section S2, or may be processed in parallel by processing both the second type section S2 and the third type section S3.

[0052] The next step is therefore performed for at least the intervals of the first type S1 and the intervals of the second type S2 and / or the intervals of the third type S3, with reference to what has been mentioned above, and consists of a step 110 of sorting the coherence of the identified intervals. This step will be explained in more detail below, in particular with reference to Figures 2 and 3, and in particular consists of estimating a reference interval for this type of interval and then calculating the cross-correlations for all the intervals of the same type previously identified, in order to make it possible to store only the intervals most similar to this reference interval, i.e. the coherent intervals of each type of interval.

[0053] Regardless of the number of interval types processed simultaneously in the previous step, the processing of the signal ends with step 112, in which an average interval is calculated for each of the interval types. In other words, regardless of whether the third type of interval is processed separately from the second type of interval or added to the latter, the calculation step makes it possible to obtain a first type of average interval S1_moy, a second type of average interval S2_moy, and a third type of average interval S3_moy.

[0054] The processed acceleration signal resulting from the signal processing step and reflecting the patient's cardiac activity over the signal acquisition period (in this case approximately 30 seconds) is formed by concatenating each of the previously calculated averaging intervals S1_moy, S2_moy, S3_moy.

[0055] As mentioned above, the processed acceleration signal is a marker used in determining the risk of cardiac decompensation, step 114. This determination step may take into account other markers including, inter alia, electrophysiological markers 115 determined by the electrocardiogram signal mentioned above, respiratory frequency 117, the tilt of the patient's body during the night 119, or the pre-ejection period, denoted by the acronym PEP.

[0056] In particular, a marker corresponding to the processed acceleration signal may be used to compare the amplitude of the average interval of the first type with data stored in memory regarding the occurrence of this first type interval over a given period of time, and to compare the amplitude of the average interval of the third type of interval S3 with data stored in memory regarding the occurrence of this first type of interval over a given period of time, which may be different from the period over which the first type of interval is considered.

[0057] In particular, if it is observed that the first type of interval S1 of the processed acceleration signal ACC_T is smaller than that previously recorded and the third type of interval S3 of the processed acceleration signal ACC_T is larger than that previously recorded, the decision step 114 may indicate information regarding the risk of cardiac decompensation.

[0058] Analysis of heart sounds provides information about hemodynamic function. In heart failure, both the first heart sound (i.e., the first type S1 segment) and the third heart sound (i.e., the third type S3 segment) are important for predicting decompensation.

[0059] The amplitude of the first type S1 interval is strongly correlated with ventricular contractility and overall cardiac performance. A decrease in amplitude indicates a decrease in cardiac contractility. In contrast, an increase in the amplitude of the third type S3 interval is a specific sign of increased filling pressures that occur secondary to a decrease in cardiac contractility. Several studies have shown that the specificity of an increase in the amplitude of the third type S3 interval in cardiac decompensation is very high, at approximately 80%.

[0060] With reference to FIG. 2, a first embodiment of the present invention will now be described.

[0061] As mentioned above, in this particular embodiment, the electrocardiogram signal ECG is taken into account in the step 106 of dividing the acceleration signal ACC into time cycles 107 of a given time length. The result is a step of acquiring this electrocardiogram signal ECG, which is carried out before this division step 106 and simultaneously with the step of acquiring the acceleration signal in a common acquisition step 102.

[0062] These two separate signals are acquired synchronously, ie by simultaneous triggering of the acquisition operations and acquisition on the same time scale.

[0063] The control unit is configured to execute in parallel a first signal analysis step 121 in which the electrocardiogram signal ECG is analyzed and a second signal analysis step 122 in which the acceleration signal ACC is analyzed.

[0064] The first signal analysis step 121 comprises at least one step of filtering the electrocardiogram signal 124 to remove artifacts and detecting signal peaks 126 representative of defined cardiac features. The control unit is configured to define at least one datum on the time position of these peaks relative to the start of the acquisition period at the output of this first analysis step 121.

[0065] In particular, the first signal analysis step 121 may include at least one filtering of the electrocardiogram signal ECG over a frequency band of, for example, about 5-60 Hz, without limiting the scope of the present invention, followed by a step of detecting the locations of various peaks R occurring successively during the signal acquisition period, which are representative of defined cardiac features.

[0066] The positions of these peaks (in this case peak R) are considered in combination with the filtered acceleration signal SCG obtained in the second signal analysis step 122 .

[0067] The control unit is configured to execute, in parallel with the first signal analysis step 121, a second signal analysis step 122 in which the acquired acceleration signal ACC is analyzed.

[0068] The second signal analysis step 122 includes at least one step of filtering the acceleration signal ACC. More specifically, the acceleration signal may be filtered over a band of 5 to 100 Hz. In particular, the frequency may be around 20 Hz to avoid low-frequency noise and frequencies associated with cardiac motion.

[0069] The control unit is configured to perform a step of cross-analyzing the electrocardiogram signal, in particular the position of the peak R identified in the above-mentioned filtered electrocardiogram signal ECG_1, with the filtered acceleration signal SGC obtained from the second signal analysis step 122.

[0070] The mutual analysis step consists of defining, in a synchronized operation, the start of a time cycle based on the appearance of peaks R of the filtered electrocardiogram signal ECG_1, thereby enabling a step 106 of dividing the acceleration signal, here the filtered SGC, into a number of time cycles 107 of the same time length, each starting at each identified peak R.

[0071] It may be assumed that a predefined fixed time length is applied to each time cycle from the appearance of peak R. In the example embodiment of the invention described herein, each time cycle beginning with the appearance of peak R ends with the appearance of the next peak R, and therefore the appearance of the next peak R serves to define the end of time cycle n and the start of time cycle n+1. To ensure that each time cycle has the same time length, the control unit is configured to adjust the time length of the time cycle by adding or subtracting time from the average time length of all time cycles identified by mutual analysis of the positions of peak R in the acceleration signal and the electrocardiogram signal.

[0072] Such division of the acceleration signal can be seen in FIG. 3, which shows the division of the acceleration signal into six time cycles 107 of equal duration.

[0073] Once the filtered acceleration signal SCG has been segmented into time cycles 107, the identification step 108 may include an envelope calculation step for each time cycle 107, in which an envelope approximation 109 is made to a plot of a curve representing the cardiac signal over the entire time cycle, which may be either a squared envelope calculation or an absolute value envelope calculation.

[0074] This kind of envelope 109 calculation for each time cycle can be seen in Figure 4, which shows the squared envelope and the absolute value envelope for each time cycle. It is understood that any method that allows signal rectification can be implemented here without departing from the context of the present invention.

[0075] The identification step 108 also includes identifying a particular section S1, S2, or S3 of the processed time cycle by detecting a previously calculated amplitude value of one or the other envelope that deviates significantly from a flat waveform of the cardiac signal. Thus, the control unit identifies the presence of a first type of section S1 when the signal amplitude, more specifically, the amplitude of one and / or the other envelope, is greater than a first threshold. Similarly, the control unit identifies the presence of a second type of section S2 when the signal amplitude, more specifically, the amplitude of one and / or the other envelope, is greater than a second threshold, and identifies the presence of a third type of section S3 when the signal amplitude, more specifically, the amplitude of one and / or the other envelope, is greater than a third threshold.

[0076] Since intervals are defined over a given time range in which the amplitude of the envelope of the acceleration signal is greater than a corresponding threshold, it is understood that envelope amplitudes greater than the corresponding threshold for very short lengths of time may be considered artifacts and will not be captured as characteristic intervals.

[0077] To reliably and accurately identify the first, second, and third intervals, the control unit performs a comparison of the amplitude value of such envelope with a threshold value over two separate portions of each time cycle 107.

[0078] More specifically, the control unit searches for a signal amplitude value greater than the first threshold value over the first third T1 of the time cycle 107. In other words, in the first type of section S1, the search and identification are performed only in the first part of the cycle (in this case, the first third T1). The first third T1 of the time cycle 107 refers to the period of the time cycle that elapses from the start point over a period equal to one-third of the entire time length of the time cycle.

[0079] Figure 5, which illustrates a representative time cycle of a processed acceleration signal and is described below, shows how the time cycle is divided into three thirds, T1, T2, and T3. It is important to understand that in this identification process, each time cycle is divided into three thirds, and searches are performed in the corresponding thirds for different types of intervals.

[0080] The control unit is also configured to search for envelope amplitude values ​​greater than a threshold different from the first threshold over the remaining time cycle. In other words, in the second type of interval S2 and the third type of interval S3, the search and identification are performed only in the part of the time cycle complementary to the first part of the cycle, i.e., here, the last two-thirds T2, T3 of the time cycle 107.

[0081] An envelope amplitude value greater than the second threshold is captured as a second type section S2, and an envelope amplitude value greater than the third threshold and occurring after the amplitude value corresponding to the second type section S2 is captured as a third type section S3.

[0082] It is noteworthy that in this identification process, a third type interval S3 is searched following a detected second type interval S2 during the same part of the time cycle, i.e., the remaining part of the time cycle during which a first type interval S1 has not been searched.

[0083] The control unit is configured to perform step 130 of sorting the identified intervals, in particular removing from these intervals those that may be considered artifacts (i.e., intervals that are foreign to what is expected for the patient, here by spectral energy calculation as presented above, i.e., by considering the z-score of each interval).

[0084] In other words, this sorting step 130 consists of an exclusion substep of the identification step 108, which consists of excluding intervals previously identified as characteristic of the identified cardiac sounds if they are classified as artifacts in this sorting step.

[0085] An interval identified here is considered an artifact if its associated z-score is greater than 3.

[0086] At the end of this identification process 108, the control unit can store all of the first type intervals S1 identified over all time cycles 107 in a first database, and can store all of the second type intervals S2 and third type intervals S3 identified over all time cycles in a second database in which each second type interval is associated with a third type interval that exists in the same time cycle.

[0087] Alternatively, it should be noted that in the embodiment described below, the control unit may be configured to store the third type of interval S3 in a third database separately from the second type of interval S2.

[0088] In this first embodiment, the step 110 of sorting the coherence of the identified intervals, which is performed after the identification step 108, is divided into two sub-steps performed in parallel, each of which processes all of the previously identified intervals of the same type.

[0089] The first substep 1101 consists of calculating the cross-correlation for all the first type of intervals previously identified in order to select the first type of coherent intervals which will then be used to calculate the first type of average interval.

[0090] More specifically, the first substep 1101 includes a step 1101_1 of calculating the cross-correlation for each of the first-type intervals S1 to define a first reference-type interval S1_ref. In other words, the cross-correlation of each of the first-type intervals S1 with all other intervals of the first type is calculated, and an average correlation score is assigned to each time cycle based on the correlation score established as a function of coherence between the first-type interval S1 associated with each time cycle and each of the other intervals of the first type associated with the other time cycles. The reference interval and the corresponding reference time cycle are the ones that exhibit the best cross-correlation with all others.

[0091] Then, in the first sub-step 1101, a coherence sorting step 1101_2 is performed by comparing each of the first type of intervals S1 present in the database that have not been removed by the previous spectral energy analysis with a first type of reference interval S1_ref to define a correlation index. Without limiting the present invention, the control unit is configured not to retain any first type of intervals whose correlation index is below a specified threshold value of at least about 50%, which may be about 60% here.

[0092] Finally, in a reset step 1101_3 of this first sub-step 1101, the control unit is configured to reset each of the retained first-type intervals S1, i.e., each interval that has not been excluded from the analysis by the spectral energy calculation or coherence sorting step, to the time of the first-type reference interval S1_ref. The resetting is performed so as to obtain maximum correlation between the reset interval and the corresponding reference interval.

[0093] The method then performs a step 112 of calculating an average interval based on these reset intervals and the reference interval, in this case a step 1121 of calculating an average interval of the first type S1_moy representative of each of the identified and not excluded intervals of the first type S1. This calculation step consists in carrying out an averaging of the amplitude values ​​of each of the intervals of the first type S1 at each instant of a time cycle, for example every millisecond.

[0094] The control unit is configured to store a first type of averaging interval S1_moy associated with the analysis of the acceleration signal.

[0095] The control unit executes a second sub-step 1102 simultaneously with the execution of the first sub-step 1101. This second sub-step 1102 includes the same steps as the first sub-step 1101, namely a cross-correlation calculation step 1102_1 for defining a second reference type interval S2_ref, a coherence sorting step 1102_2 for the reference interval, and a time reset step 1102_3, and then also executes a step 1122 for calculating an average interval. Again, this second sub-step and the various steps it includes are performed only on intervals that have passed the spectral energy analysis step, which means that the calculation can be focused only on useful intervals.

[0096] In this first embodiment, the second sub-step 1102 is performed by considering only the second type of intervals S2, and the third type of intervals S3 are retained or discarded depending on what is done to the second type of intervals S2 present in the corresponding time cycle.

[0097] More specifically, the intervals, and thus the time cycles, are sorted based on their spectral energy by calculating the z-score of the second type of interval S2. If the second type of interval has a z-score below a threshold, here equal to 0.3, the entire time cycle, i.e., the second type of interval and the associated third type of interval, are excluded.

[0098] Similarly, a cross-correlation calculation step 1102_1 is performed based on the second type of intervals to select a second type of reference interval S2_ref. Then, each of the second type of intervals S2 stored in the second database that are not excluded by the spectral analysis is compared with the second type of reference interval S2_ref. As described above for the first substep 1101, the control unit excludes from the analysis any second type of interval that, when compared with the second type of reference interval, results in a correlation threshold that is deemed insufficient (for example, about 60% here, but this value is not limited thereto). However, in this case, excluding this second type of interval S2 also has the effect of excluding a third type of interval S3 identified in the time cycle 107 corresponding to the excluded second type of interval S2.

[0099] Next, an interval average value calculation step is performed for each interval type, ie, intervals of the second type S2 and intervals of the third type S3.

[0100] The control unit is configured to store the second type of averaging interval S2_moy and the third type of averaging interval S3_moy associated with the analysis of the acceleration signal. These averaging intervals representative of the processed acceleration signal ACC-T are then used in step 114 to determine the risk of cardiac decompensation, in particular by analyzing the change in amplitude of the intervals from one signal to the next.

[0101] The control unit may be configured to reconstruct a time cycle representative of the processed acceleration signal ACC-T based on the first averaging type interval S1_moy, the second averaging type interval S2_moy, and the third averaging type interval S3_moy for the purpose of displaying the signal.

[0102] FIG. 5 shows, by way of example, an average time cycle 107 comprising a first average type interval S1_moy present in the first third T1 of the time cycle, and a second average type interval S2_moy and a third type interval S3_moy present in the remaining parts of the time cycle, i.e. the second two-thirds T2 and T3.

[0103] It will be appreciated that this first embodiment is particular in that the cross-correlation calculation is performed not on these types of sections but on the second type of sections having larger amplitudes, thereby enabling more reliable cross-correlation and coherence calculations and therefore signal reconstruction based on the third type of average section S3_moy.

[0104] 6, the second embodiment will be described, in which the second sub-step 1102 differs from the second sub-step of the first embodiment in that the steps of calculating the cross-correlation to determine the reference interval and sorting by comparing the interval with the reference interval are performed for both the second type intervals and the third type intervals S3. The third type average interval S3_moy is calculated by the control unit based on the sorting of the identified third type intervals, which is performed independently from the sorting of the identified second type intervals.

[0105] The steps prior to the identification step 108 are not shown in FIG. 6, as they are substantially identical here.

[0106] As mentioned above, the step 110 of sorting the coherence of the identified intervals, which is carried out after the identification step 108, is divided into several substeps carried out in parallel, in which all the previously identified intervals of the same type are processed in each step. As mentioned above, the first substep 1101 consists in sorting coherent intervals of a first type and then calculating the cross-correlation for all the previously identified intervals of the first type in order to calculate an average interval of the first type S1_moy.

[0107] As described above, in this second embodiment, the control unit is configured to store the third type of section S3 in a database separate from the database in which the second type of section S2 is stored.

[0108] In this second embodiment, the interval coherence sorting step 110 is divided into three substeps. The control unit executes the second substep 1102 and the third substep 1103 simultaneously with the execution of the first substep 1101. The second substep 1102 and the third substep 1103 include the same steps as the first substep 1101, namely, cross-correlation calculation steps 1102_1 and 1103_1 that define reference intervals S2_ref and S3_ref, coherence sorting steps 1102_2 and 1103_2 for the reference intervals, and time reset steps 1102_3 and 1103_3, respectively. Thereafter, each of these two substeps executes steps 1122 and 1123 that calculate average intervals S2_moy and S3_moy, respectively.

[0109] It will therefore be appreciated that in this second embodiment, unlike the first embodiment, the third type reference interval S3_ref is determined by a cross-correlation calculation specific to the third type interval S3. The second sub-step 1102 is performed taking into account only the second type interval S2, and the third sub-step 1103 is performed taking into account only the third type interval S3, and these intervals are retained or excluded in each sub-step based solely on what is done in the corresponding sub-step.

[0110] As an example, in this second embodiment, a third type interval S3 present in the defined time cycle 107 can be retained because it is coherent with the third type reference interval S3_ref, while at the same time, a second type interval S2 present in this same defined time cycle 107 is excluded because it is not coherent with the second type reference interval S2_ref. Thus, the averaging step 1122 at the end of the second sub-step 1102 may be based on a number of second type intervals S2 that is different from the number of third type intervals S3 on which the averaging step 1123 at the end of the third sub-step 1103 is based.

[0111] Again, sorting of intervals by spectral analysis 130 is performed, in this case, on all of the first type of intervals in the first sub-step 1101, the second type of intervals S2 in the second sub-step 1102, and the third type of intervals S3 in the third sub-step 1103.

[0112] This sorting serves the same purpose as the first embodiment, namely to avoid including in the cross-correlation calculation intervals that are clearly atypical and should be considered artifacts, thereby reducing the calculation time associated with each sub-step.

[0113] Having described the first embodiment, the controller can then store the average intervals for each of these types of intervals and restore them for step 114 of determining the risk of cardiac decompensation, in particular, where the change in amplitude of the interval from one signal to another is analyzed.

[0114] Here too, based on the first type of average interval S1_moy, the second type of average interval S2_moy and the third type of average interval S3_moy, the processed acceleration signal ACC-T can be reconstructed for display purposes.

[0115] It will be appreciated that this second embodiment is particular in that it allows the reconstruction of a signal having a third type of mean interval S3_moy based on a direct analysis of this third type of interval.

[0116] The invention as described above has achieved its stated objective, namely, to efficiently process acceleration signals acquired from a patient and define the processed signals to enable a reliable determination of increases or decreases in the amplitude of specific data within said signals, thereby providing the practitioner with a tool for diagnosing any cardiac decompensation. Variations not described herein may be implemented without departing from the context of the invention, in particular as long as they form part of the determination method that allows the invention to define at least three types of averaging intervals representative of the acceleration signals acquired in the patient.

[0117] As a non-exhaustive example, a possible variation provides an additional step in which the number of intervals of each type retained after a step of the method is analyzed and compared with a threshold value. If the number of intervals of one of these types, for example, the number of intervals of the first type, is less than the threshold value, the acquired signal is considered to be of low quality and is not retained in the calculation of the average interval, and the acquired signal is not processed for the purpose of determining the risk of cardiac decompensation. The threshold value may, for example, be a number corresponding to 50% of the number of time cycles obtained by dividing the acceleration signal. The steps of the method before this additional step may, in particular, be an interval identification step and an exclusion substep based on the calculation of spectral analysis scores. The number of intervals of the first type S1, the second type S2, and the third type S3 retained after this exclusion substep are compared with a threshold value, and depending on the result of this comparison, a decision is made to continue the method, in particular to start a coherence sorting step, which requires performing a large number of cross-correlation calculations. As mentioned above, if one of the number of intervals of a given type is below a threshold, the acquisition of the acceleration signal as a whole is considered unreliable and no further calculations based on this acquired acceleration signal are performed.

[0118] Another possible variation provides another exclusion substep performed in the time cycle division step, which specifically consists of excluding from the method time cycles whose time length, defined by the interval between two consecutive peaks R, is much longer than the average or median of the time lengths of the other time cycles. For example, time cycles having an original time length exceeding 1.7 times the average or median are excluded from the remainder of the method, and are therefore not reset to the average time length of the time cycle or analyzed to detect characteristic intervals. Similarly, time cycles having very short time lengths, for example, original time lengths less than 0.3 times the average or median, are also excluded.

Claims

1. A method (100) for determining a risk of cardiac decompensation in a patient, the method (100) comprising: a control unit executing at least one step (104) of processing an acceleration signal (ACC) acquired from the patient, and performing a step (114) of determining the risk of cardiac decompensation based on a study of a plurality of markers (115, 117, 119, PEP) comprising intervals (S1, S2, S3) of the processed acceleration signal (ACC-T) obtained from the processing step (104), The step (104) of processing the acceleration signal performed by the control unit comprises: Dividing (106) the acceleration signal into time cycles (107) of equal time length; identifying (108) intervals (S1, S2, S3) characteristic of heart sounds in each time cycle (107) by a first search for an interval (S1) of a first type in a first defined portion of said time cycle (107) and a second search for at least an interval (S2) of a second type and an interval (S3) of a third type in a second defined portion of said time cycle (107); a step (110) of sorting the coherences of the identified intervals, at least for the intervals of the first type (S1) and the second type (S2), consisting in calculating the cross-correlations of all the previously identified intervals of the same type, in order to make it possible to deduce a reference interval for this type of interval and to preserve in the second step only the intervals most similar to this reference interval, i.e. the coherent intervals of each of the types of intervals; a step (112) of calculating a first type of average interval (S1_moy), a second type of average interval (S2_moy), and a third type of average interval (S3_moy), wherein the average calculation is based on all of the selected first type intervals (S1), each of all of the selected second type intervals (S2), and each of all of the selected third type intervals (S3); The invention is characterized in that it comprises the determining step (114) comprises at least one step of determining an increase in the amplitude of the intervals of the first type (S1) relative to the data in the memory and / or a decrease in the amplitude of the intervals of the third type (S3) relative to the data in the memory, the control unit being configured on the one hand to store the average intervals of the first type (S1_moy) related to the analysis of the acceleration signal, and on the other hand to store the average intervals of the second type (S2_moy) and the average intervals of the third type (S3_moy) related to the analysis of the acceleration signal, these average intervals being representative of the processed acceleration signal (ACC-T) and being used in this determining step, in particular to analyze the changes in the amplitude of the intervals from one signal to the next; The determination method (100) is also characterized by:

2. 2. The method of claim 1, wherein the identifying step (108) includes a substep (130) of excluding intervals previously identified as first, second, or third type intervals, the substep excluding the intervals classified as artifacts in the excluding substep.

3. 3. The method of claim 2, wherein the exclusion substep (130) comprises calculating a spectral analysis score, and intervals exhibiting a score greater than a threshold are identified as artifacts and excluded.

4. 2. The method (100) of claim 1, wherein the processing step (104) is performed for each signal detected in one of three acquisition axes of the acceleration signal (ACC) or for an overall signal obtained by combining the signals acquired in each of the axes.

5. The method (100) of claim 1, comprising processing an electrocardiogram signal (ECG) obtained from the patient.

6. 6. The method (100) of claim 5, wherein the step (106) of dividing the signal comprises synchronizing the acceleration signal (ACC) based on the electrocardiogram signal (ECG).

7. 7. The method (100) according to claim 1, wherein the step (108) of identifying comprises a step of calculating an envelope (109) of the acceleration signal, and wherein intervals (S1, S2, S3) are identified when the amplitude of the envelope (109) is greater than a determined threshold value in a defined portion of the time cycle (107).

8. 8. The method (100) of claim 1, wherein the first defined portion of the time cycle (107) in the identifying step (108) is equal to the first third (T1) of the time length of the time cycle (107) and the second defined portion of the time cycle is equal to the last two-thirds (T2, T3) of the time cycle (107).

9. 9. The method (100) of claim 1, wherein the sorting step (110) comprises at least two series of correlation calculations performed in parallel, including a first series of correlation calculations performed for all of the first type of intervals (S1) and a second series of correlation calculations performed for all of the second type of intervals (S2).

10. 10. A method (100) for determining whether a given type of interval (S1, S2, S3) is being used, characterized in that each of the series of correlation calculations performed for the given type of interval (S1, S2, S3) comprises a first calculation of cross-correlations (1101_1, 1102_1, 1103_1) between all intervals identified as this type of interval to define a reference interval (S1_ref, S2_ref, S3_ref) for the type of interval, followed by a step of calculating the correlation of each of the intervals identified as this type of interval with respect to the reference interval.

11. The method (100) according to claim 10, characterized in that the intervals selected in each series of correlation calculations are intervals having a correlation index with the reference intervals (S1_ref, S2_ref, S3_ref) that is greater than or equal to a predetermined correlation threshold.

12. 12. A method (100) for determining whether a signal is amplified by a first correlation calculation performed for all of the first type of intervals (S1) and a second correlation calculation performed for all of the second type of intervals (S2).

13. 12. The method (100) of claim 9, wherein the sorting step (110) comprises three series of correlation calculations performed in parallel, including a first series of correlation calculations performed for all of the first type of intervals (S1), a second series of correlation calculations performed for all of the second type of intervals (S2), and a third series of correlation calculations performed for all of the third type of intervals (S3).

14. A determination method (100) that combines the determination method (100) described in any one of claims 1 to 13 with claim 10, characterized in that in the selection step (110), before the step (112) of calculating the average interval, a step (1101_3, 1102_3, 1103_3) of time resetting the selected interval to the corresponding reference interval (S1_ref, S2_ref, S3_ref) is performed for each type of interval.

Images