The invention is part of the field of processing acoustic signals, in particular tracheal or pulmonary sounds captured from an individual.
More specifically, the invention relates to a technique for recognising the respiratory phases of an individual from an acoustic signal representative of the respiratory activity of this individual.
The invention applies in particular, but not exclusively, in the field of acoustic instrumentation dedicated to the recording and monitoring of the respiratory activity of individuals, for example for the study of respiratory pathologies or respiratory disorders during sleep.
In the remainder of this document, focus is more particularly made to describe the problem existing in the context of the study of the respiratory activity in patients suffering from “sleep apnoea syndrome” (known as SAS syndrome). The invention is of course not limited to this particular context of application, but is of interest for any technique for recognising respiratory phases having to deal with a close or similar problem.
The sleep apnoea syndrome is manifested by repeated and uncontrolled interruptions in breathing during sleep. They cause non-conscious micro-arousals that can harm the quality of sleep. This can therefore result in daytime drowsiness, difficulties with concentration or memory, or even cardiovascular complications, possible sources of corollary accidents such as road or work accidents. It is therefore important to have medical equipment capable of detecting this type of pathology, in particular for screening and treatment purposes.
The diagnosis of sleep apnoea syndrome is mainly based on the exploration of the individual's ventilation in order to observe movements of breathing and the evolution of the upper airway during sleep. Apnoeas and hypopneas are investigated by analysing the amplitude of acoustic signals obtained for example during ventilatory polygraphy or polysomnography. It is generally recommended to use a certain number of measuring instruments to measure the individual's respiratory behaviour during sleep, such as nasal cannula, thoraco-abdominal straps, pulse oximeter, etc. An analysis software analyses the signals recorded by the different measuring instruments in order to identify possible pathological events, such as apnoeas and hypopneas. Other instruments, such as tracheal sound sensors, can also be used in a complementary manner to study more specifically certain parameters associated with the inhalation and exhalation phases of the individual's respiratory activity, and allow refining the diagnosis. Indeed, the latest studies carried out on respiratory cycles obtained from tracheal sound signals have allowed highlighting a certain number of pathological indicators whose consideration proves to be relevant in the detection and evaluation of respiratory disorders.
However, even if there are techniques for analysing respiratory cycles from tracheal sounds, such as the non-invasive phase detection technique proposed by ZK Moussavi et Al. in the scientific article “Computerised acoustical respiratory phase detection without airflow measurement” (March 2000), these lack accuracy. The respiratory activity of an individual (typically composed of successive respiratory cycles, each respiratory cycle comprising an inhalation phase and an exhalation phase) can be observed from the following three representations: a temporal representation, a frequency representation and a frequency-intensity representation over time (“three-dimensional” representation). However, it proves that it is often difficult to accurately distinguish the inhalation phase from the exhalation phase of a respiratory cycle in the frequency or time domain of a tracheal sound signal, in particular when the captured signal is located below a certain sound threshold (case of an individual at rest for example). Moreover, such known techniques only provide a limited amount of relatively imprecise information on the patient's respiratory activity.
Moreover, certain respiratory profiles prove to be particularly complex to process, for example in the case of snoring or in the case of apnoea during which the reduction in the acoustic signal is accompanied by noises linked to the individual's efforts to enter or exit air via the respiratory tract. Other noises, such as swallowing, coughing, throat clearing, or even changes in the individual's position during his sleep complicate the recognition of the respiratory phases.
There is therefore a real need to provide a technique which allows an effective recognition of the respiratory phases of an individual in a tracheal sound signal, and more generally in an acoustic signal representative of a respiratory activity. It would also be particularly interesting to provide such a technique which is minimally invasive and easy to implement.
In a particular embodiment of the invention, a method is proposed for determining respiratory phases in an acoustic signal representative of a respiratory activity of an individual, the respiratory activity being defined by respiratory cycles each comprising an inhalation phase and an exhalation phase, the method is remarkable in that it comprises the following steps:
Thus, the principle of the invention is based on a search in the acoustic signal for pauses of the signal to deduce the inhalation and exhalation phases therefrom thanks to their duration and their position in the signal. This new approach is clever because, rather than directly seeking to process the inhalation and exhalation phases as in the state of the art, the present invention focuses first on recognising pauses in the signal, these being easier to recognise in an acoustic signal. Thus, the invention offers a technique for recognising respiratory phases which is more effective than state-of-the-art techniques. Moreover, being able to determine the inhalation and exhalation phases using only an acoustic signal offers prospects for simplified examination of medical equipment.
According to a particular aspect, the method implements a mechanism for distinguishing among the pause intervals, between a first type of pause interval, called short pause, of duration less than a second type of pause interval, called long pause.
According to a particular feature, the acoustic signal belongs to the group comprising: a tracheal acoustic signal or a pulmonary acoustic signal captured from an individual.
According to a particular implementation of the method, if at the end of the detection step at least one succession of two pause intervals which are separated in the signal by a time interval, called aberrant interval, of duration less than a predetermined duration threshold is identified, the method comprises a step of replacing, for each detected succession, the set consisting of the two pause intervals and said aberrant interval by a new pause interval. The mechanism for recognising the inhalation and exhalation phases according to the invention is thus optimised. Indeed, the predetermined threshold is dimensioned such that when a point defect appears in the signal between two pause intervals, this is considered as an integral part, with the two pause intervals, of the same and single pause. This allows eliminating the interference noise contained in the signal which would disrupt the recognition of the respiratory phases.
According to a particular feature, if at the end of the determination step at least one time interval, called indeterminate interval, meeting a predetermined inconsistency criterion, is identified: —processing the signal portion corresponding to said indeterminate interval, by means of a machine learning method; —associating said at least one indeterminate interval with an inhalation or exhalation phase depending on the results of the processing step. The mechanism for recognising the inhalation and exhalation phases is thus further improved according to the invention.
According to a particular aspect of the invention, the step of detecting pause intervals is based on a processing belonging to the group comprising a frequency processing of the acoustic signal, an energy processing of the acoustic signal. Indeed, it is by observing the energy and time-frequency representation spaces of sound signals that has been considered, surprisingly, that the respiratory pauses are more easily recognisable in these representation spaces than the inhalation and exhalation phases themselves. The pause recognition mechanism according to the invention can be implemented either by frequency processing of the signal, or by energy processing of the signal, or by simultaneous frequency and energy processing of the signal, thus making the method even more reliable.
According to a particular implementation, the frequency processing comprises the following steps:
According to a particular implementation, the energy processing comprises the steps:
According to a particularly advantageous aspect of the invention, the method comprises a step of shaping, by processing said first and second transient logic signals by an OR-Inclusive logic function, a final logic signal whose high logic levels correspond to said detected pause intervals. Since the frequency and energy treatments can be complementary, taking these two processings into account can allow detecting pauses which would not have been detected by the implementation of only one of the two processings.
According to a particular feature, the method comprises a step consisting in, if at the end of the shaping step, at least one succession of two pause intervals which are separated by an aberrant interval detected in said final logic signal of duration less than the predetermined duration threshold is identified: replacing, for each succession detected in said final logic signal, the set consisting of the two pause intervals and said aberrant interval by a new pause interval. This facilitates distinguishing between the long pauses and short pauses, and in fact allows a more reliable recognition of respiratory phases.
In another embodiment of the invention, a computer program product is proposed, which comprises program code instructions for implementing the aforementioned method (in any one of the different embodiments thereof), when said program is executed on a computer.
In another embodiment of the invention, a computer-readable and non-transitory storage medium is proposed, storing a computer program product comprising a set of instructions executable by a computer to implement the aforementioned method (in any one of the different embodiments thereof).
In another embodiment of the invention, a device is proposed for determining respiratory phases in an acoustic signal representative of a respiratory activity of an individual, the respiratory activity being defined by respiratory cycles each comprising an inhalation phase and an exhalation phase, the device being characterised in that it comprises: —means for detecting pause intervals in the acoustic signal; —means for determining activity intervals corresponding to inhalation phases and activity intervals corresponding to exhalation phases by comparison of the duration of the pause intervals
Advantageously, the device comprises means for implementing the steps which it performs in the obtaining method as previously described, in any one of the different embodiments thereof.
Other features and advantages of the invention will appear on reading the following description, given by way of indicative and non-limiting example, and the appended drawings, in which:
In all figures of the present document, identical elements and steps are designated by the same reference numerals.
In the remainder of the description, we consider an example of implementation of the invention in the context of a tracheal acoustic signal, that is to say an acoustic signal obtained on the trachea of an individual to carry out an examination of respiratory activity of said individual during sleep (for example to diagnose a sleep apnoea syndrome). The invention is of course not limited to this particular context, and can be applied to any type of acoustic signals representative of a respiratory activity of a living being.
In the remainder of the document, the term “acoustic signal” means the electrical signals representative of the respiratory activity of the individual X which have been collected and recorded by the processing unit 2 during the examination.
The individual's respiratory activity is typically composed of respiratory cycles, each cycle comprising an inhalation phase and an exhalation phase which produce sound pressure waves within the individual. The inhalation phase corresponds to a filling of the lungs during which the diaphragm and the intercostal muscles are contracted. The exhalation phase corresponds to an emptying of the lungs during which the thoracic muscles are relaxed.
The method is initialised by activating the recording of the respiratory activity of the individual X or after a predetermined recording duration.
In step S1 (referenced “OBT_S”), the device acquires the tracheal acoustic signal of the individual X. An example of a tracheal acoustic signal (referenced “SAT”) is represented in
In step S2 (referenced “DET_PA”), the device proceeds, by signal analysis, to a detection of the pause intervals contained in the SAT signal. This detection step is based on the implementation, by the device, of a frequency processing of the SAT signal and/or an energy processing of the SAT signal. The principle of these two processings is described in detail below in relation to the blocks A and B of the flowchart in
The term “pause interval” means a time interval during which the corresponding signal portion is representative of a respiratory pause, in other words of an absence of respiratory activity. This absence of respiratory activity characteristic of a cycle of conventional respiratory activity is easily identified in the signal both in the frequency domain and in the energy domain.
In step S3 (referenced “DIS_P”), the device performs a distinction, among the pause intervals detected in step S2, between a first type of pause interval, called “short pause” and a second type of pause interval, called “long pause”, depending on the estimated duration for each of the pause intervals. It is considered here that a “short pause” type interval is of a duration less than a “long pause” type interval.
According to a particular implementation, after having estimated the duration of the detected pause intervals, the device carries out a comparison of the duration of the pause intervals which follow each other in pairs over time to determine which one is the short pause and which one is the long pause.
Complementarily or alternatively, the device compares the duration of each of the pause intervals depending on a predetermined threshold. Such a threshold consists of a duration chosen to differentiate the long pauses from the short pauses among the detected pause intervals. For example, a pause interval is considered to correspond to a long pause if the estimated duration for this interval is greater than said threshold. Otherwise, it corresponds to a short pause if the estimated duration for this interval is less than said threshold.
In step S4 (referenced “DET_PH”), the device determines the time intervals of the SAT signal corresponding to the inhalation and exhalation phases taking into account the following scheduling rule: a time interval corresponds to an inhalation phase if it is comprised between a long pause and a short pause, and an exhalation phase if it is comprised between a short pause and a long pause. Indeed, in a respiratory cycle in a normal resting situation, the inhalation phase is followed by an inhalation pause (or short pause), which is followed by an exhalation phase, itself followed by an exhalation pause (or long pause), the short pause being of a duration less than the long pause. In other words, the principle is based on the comparison, for each time interval comprised between two successive pauses, of the pause which precedes and the pause which follows said interval. An inhalation is followed by a pause shorter than the one preceding it, an exhalation is followed by a pause longer than the one preceding it. It is based on this observation that the aforementioned scheduling rule was established for the present invention.
Thus, at the end of step S4 of the method, the device has a segmentation of the SAT signal into time intervals each associated with one of the following three categories: an inhalation phase, an exhalation phase, a pause. This information can be provided in the form of a transient logic signal representative of the respiratory activity of the individual X, the logic states of which correspond to the respiratory phases and pauses of said individual. The device communicates the logic signal representative of the respiratory activity of the individual X to the man-machine interface, for display and diagnosis purposes by the medical staff.
Thus, the general principle of the method is based on searching in a tracheal acoustic signal for pauses in the respiratory activity of an individual to deduce the inhalation and exhalation phases contained in this signal. This respiratory phase recognition technique is particularly effective because it focuses on priority processing of signal portions corresponding to respiratory pauses, easily recognisable in the frequency and energy domains. It is indeed by observing the temporal evolution of tracheal acoustic signals in the frequency and energy domains that the inventors of the present noticed that the pauses are more easily identified than the inhalation and exhalation phases as such. By offering the possibility of determining the inhalation and exhalation phases solely using tracheal sounds captured from an individual, the method of the invention thus opens up prospects for simplified and less invasive examination, further requiring lighter diagnosis medical equipment.
In relation to
After recovering the tracheal acoustic signal SAT in step S01 (referenced “ACQ_S”), the device first applies an anti-noise high-pass filter to the signal in step S02 (referenced “FPH”), by example a Butterworth type filter of order 4 having a cutoff frequency equal to 100 Hz. This high-pass filter has the function of removing, from the SAT signal, a predefined frequency range corresponding to the cardiovascular noises emitted by the individual X through his trachea. The signal obtained at the output of the high-pass filter is a signal filtered in the time domain, subsequently called filtered signal ‘xf’.
The observation of the temporal frequency-energy representation revealed that it is on this type of representation that the pauses in respiratory activity are easiest to recognise. Two physical magnitudes distinguish the pauses of the respiratory phases (inhalation/exhalation): the frequency content and the energy. Indeed; it appears that the energy of the tracheal acoustic signal is minimal during the pauses since it corresponds to an absence of breathing and that the spectrum of the signal corresponding to a pause is close to that of white noise in the frequency band comprised between 100 and 2000 Hz.
Two particular techniques are proposed for processing the filtered signal ‘xf’: a frequency processing (represented by the dotted block A) and an energy processing (represented by the dotted block B). The second technique is subsequently described.
In step S11, the device carries out a segmentation of the filtered signal xf to obtain a plurality of samples of this signal. For example, the filtered signal xf is divided into segments of 200 samples each corresponding to 50 ms. Each segment, denoted ‘l’, is further divided into three sub-segments which overlap at 50%. Then, the device estimates the power spectral density (also known as “PSD”) of each segment l so as to transform the filtered signal xf into a frequency signal, denoted ‘Fcent’ in the time domain.
The estimated PSD of the segment l, denoted ‘Pxx[k,l]’, is equal to the mean of the PSD of the three sub-segments composing the segment l. The index k represents the index of the frequency ‘Fk’ such that Fk=k. Fs/Nfft, with the frequency Fs equal to 4 kHz, Nfft the number of points of the Fourier transform (assuming that Nfft is an even number) and k an integer comprised in the interval [0−Nfft/2]. For each segment I of the signal xf, the device estimates in step S12 (“EST_F”) the centroid frequency Fcent[l] by means of the following relationship:
Step S13 (“DET_SF”) consists in determining a variable frequency threshold, denoted Thf, which will be used in step S14 to identify the pause intervals contained in the acoustic signal. To do this, the device defines a sliding time window of predefined size substantially corresponding to the average duration of a respiratory cycle, then applies this sliding window to the frequency signal Fcent to determine the frequency threshold whose value is a function of the frequency maximum of the frequency signal Fcent in the considered sliding window. By way of example, as illustrated in
Thus, rather than defining a frequency threshold of predetermined value, the device according to the invention proposes a calculation of a variable threshold coming as close as possible to the shape of the frequency signal, which allows providing a method with increased reliability.
It should be noted that the centroid frequency is close to 1000 Hz during the pauses, which corresponds to the Nyquist frequency divided by two. This result is consistent with the hypothesis of white noise in a frequency band comprised between 100 and 2000 Hz.
In step S14 (“DET_PA1”), the device proceeds to shape, from the frequency signal Fcent, a transient logic signal PauseS whose logic levels are depending on the frequency threshold values Thf determined in step S13. An example of a logic signal PauseS, resulting from a frequency processing, is represented in
The device first applies, in step S21, a low-pass filter to the filtered signal xf (referenced “FPB”) in order to remove the frequencies for which the tracheal sound intensity level is low. The application of a Butterworth type filter of order 6 having a cutoff frequency equal to 1500 Hz, for example, leads to a filtered signal xf′ whose frequency range is comprised between 100 and 1500 Hz, corresponding to the most energetic frequency band of the tracheal sounds.
In step S22 (“EST_I”), the device defines a first sliding time window of predefined size, then applies this first window to the filtered signal xf′ in order to estimate the acoustic intensity of this filtered signal. By way of example, as illustrated in
Step S23 (“DET_SI”) consists in determining the variable intensity threshold, denoted Tht, which will be used in step S24 to identify the pause intervals contained in the acoustic signal.
To do this, the device determines a second sliding time window, but this time of variable size W depending on the respiratory period PR. The respiratory period PR is itself determined by means of an autocorrelation function, denoted rI
For l≥2000 (i.e. ½ s), the respiratory period PR is deduced from the location of the maximum of the function rI
Then, the device applies the second sliding windows determined above to the energy signal If in order to identify the minimum value of the sound intensity of the energy signal If (denoted IMin) appearing for each second sliding window determined above. To do this, the device searches, for a given moment of the signal, the minimum sound intensity 0.2×PR before and 0.2×PR after said moment. Then, for each considered sliding window, a predefined intensity value Δl is added to the minimum sound intensity value determined for said considered window, so as to deduce the intensity threshold value to be taken into account for said considered window. The predefined intensity Al is typically comprised between 1 and 2 dB.
Thus, the device of the invention proposes a variable threshold calculation getting as close as possible to the shape of the energy signal, which allows providing a method with increased reliability.
In step S24 (“DET_PA2”), the device proceeds to shape, from the energy signal If, a transient logic signal PauseE whose logic levels depend on the intensity threshold values ThI determined in step S13. An example of logic signal PauseE is represented in
The pause intervals thus detected are then stored in the local table of the device. The frequency processing ends at the end of this step S24.
At the end of processing steps S14 and S24, the device can decide to return the chronograms of the current logic signals PauseE and PauseS to the user via the man-machine interface, in order to highlight the respiratory pauses contained in the SAT signal via the frequency processing and the energy processing.
In step S03 (“REG_&_DIST-P”), the device has a first set of pause intervals resulting from the frequency processing (chronogram of the logic signal PauseS) and a second set of pause intervals resulting from the energy processing (chronogram of the logic signal PauseE). The device will then process the logic signals PauseS and PauseE by an ‘OR-Inclusive’ logic function so as to obtain a final logic signal ‘Pause’ whose high logic levels correspond to the detected pause intervals. As illustrated in
Thus, as previously described, the embodiment described here is based on a simultaneous implementation of the two signal processing operations. Such an embodiment makes the method even more reliable and robust. It is entirely possible, as an alternative, to implement only one of the two aforementioned processing operations to identify the pause intervals in the acoustic signal, for example to gain processing speed or even because one of the two processings is more adapted to the nature of the studied acoustic signal.
The device first checks whether successions of pause intervals separated by a time interval— called aberrant interval— of a duration which is less than a predetermined threshold are present in the logic signal obtained at the end of step S03. The value of this predetermined threshold is chosen to be of very short duration, that is to say a duration typically less than ¼ s. Thus, a time interval comprised between two pauses of duration less than the predetermined threshold is considered by the device as an aberration of the signal and must be part of a single pause interval. In the case of positive check for two given successive pauses, the time interval bringing together the two pauses of said succession and the aberrant interval is replaced, in the final logic signal, by a new pause interval (of logic level 1). In the case of a negative check, no replacement is made.
In the following step S04 (“DET_PH”), the device determines the inhalation and exhalation phases of the acoustic signal by comparing, for each remaining time interval, the duration of the preceding pause relative to the duration which follows said interval. The device considers that a time interval of the low logic level signal corresponds to:
Indeed, it is recalled that in a typical respiratory cycle, an inhalation is preceded by a long pause and followed by a short pause and an exhalation is preceded by a short pause and followed by a long pause.
A simple comparison of the relative duration of the pauses preceding and following each time interval is thus carried out by the device. It is therefore not necessarily useful to measure the duration then to allocate a duration state depending on this measurement to the detected pauses. Alternatively, it is possible to configure the device so that it calculates the absolute value of the duration of the detected pauses, then to compare the calculated values to deduce the exhalation and inhalation phases of the signal.
This information is stored in a local table of the device.
As illustrated in
Step S05 (“CLAJND”) consists in further improving the determination algorithm according to the invention by checking whether, for certain time intervals, the respiratory phase to which they are allocated would not be inconsistent with a conventional respiratory cycle. This step is carried out depending on one or more predetermined inconsistency criteria relating to the human respiratory cycle. A succession of at least two inhalation or exhalation phases constitutes an example of an inconsistency criterion applicable to the present invention. The cycles that do not meet the consistency criteria are then classified as indeterminate. The person skilled in the art is able to adapt the list of possible inconsistency criteria and to possibly combine them depending on the weight he wishes to give to this step of the algorithm. In the case of a positive check for a given interval, the latter is considered an “indeterminate interval”. For each indeterminate interval of the signal, the device processes the signal portion corresponding to said indeterminate interval by means of a supervised machine learning method, such as for example the method known as “Random forest” based on trees of random decision. This method is adapted to each patient by carrying out the learning phase using the supposedly well-classified cycles of said patient. The features are extracted from the filtered signal xf, for example the duration, energy, power, shape, spectral content, or any other feature that the person skilled in the art could determine.
In a complementary or alternative manner, a simple processing of the indeterminate intervals based on the coherence of the successions of the respiratory phases and the known durations of the inhalation and exhalation phases, can be implemented prior to or simultaneously with the learning method for allow quickly and easily reclassifying all or part of the indeterminate intervals.
Thanks to this step, the mechanism for recognising the inhalation and exhalation phases is further refined when certain determined phases meet said inconsistency criterion.
Finally, in step S06 (“FOU_SL”), the device provides, via the man-machine interface, the information relating to the respiratory phases which are previously determined in the form of a transient logic signal capable of being interpreted by qualified medical staff. As illustrated by way of example in the chronogram in
This
In the case where the invention is implemented on a reprogrammable calculation machine, the corresponding program (that is to say the sequence of instructions) could be stored in a removable storage medium (such as for example a floppy disk, a CD-ROM or a DVD-ROM) or not, this storage medium being partially or totally readable by a computer or a processor.
The particular embodiment presented above is based on the processing of a tracheal acoustic signal. Of course, it is entirely possible to apply the method of the invention to an acoustic signal from another organ of the individual, such as a pulmonary acoustic signal for example (to mention only one example of application). More generally, the technique of the invention applies to any type of acoustic signals representative of the respiratory activity of an individual.
Number | Date | Country | Kind |
---|---|---|---|
FR2103511 | Apr 2021 | FR | national |
Filing Document | Filing Date | Country | Kind |
---|---|---|---|
PCT/EP2022/058849 | 4/4/2022 | WO |