Method, computer program product, computer readable storage medium and system
The method improves the accuracy of estimating physiological rhythms by using prior probabilities and MAP estimation to enhance the reliability of non-intrusive measurement systems, addressing the challenges of variable signal quality and user movement.
Patent Information
- Application Number
- CN202480002319.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2023-03-24
- Filing Date
- 2024-03-20
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-03-20
AI Technical Summary
Existing non-intrusive physiological rhythm measurement systems face challenges in providing accurate estimates of instantaneous frequency due to variable signal quality caused by uncontrolled environments and user movements, which traditional algorithms developed for clinical applications cannot effectively address.
A method that utilizes a prior probability-based approach to estimate periodic characteristics of physiological rhythms by generating a second set of estimates using spectral and temporal likelihoods, combined with Maximum A Posteriori Probability (MAP) estimation to improve accuracy, particularly in unconstrained and undisturbed conditions.
Enhances the accuracy of estimating periodic characteristics of physiological rhythms, such as heart rate and breathing, even in the presence of movement or changes in posture, using non-intrusive measurement systems like accelerometers, thereby providing reliable data for health monitoring.
Smart Images

Figure CN119095536A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a method for processing a signal representing a physiological rhythm of a subject. The physiological rhythm has a periodic characteristic. The present invention further relates to a computer program product and a computer readable storage medium, each of which has instructions that, when executed by a computer, cause the computer to perform the method. In addition, the present invention relates to a system for processing a signal representing a physiological rhythm of a subject. Background Art
[0002] Accurate knowledge of physiological rhythms, such as heart rate or respiratory rate, is crucial in many healthcare applications, including diagnostics, patient monitoring, and treatment. Therefore, estimating the time-varying frequency of these rhythms from conventional clinical measurement modalities, such as electrocardiogram (ECG), pulse oximetry, or respiratory induction plethysmography, is a well-studied problem. While for some applications, simply estimating the average frequency over multiple cycles of the underlying rhythm may be sufficient, other applications, such as heart rate variability (HRV) analysis, require resolution between events.
[0003] Due to demographic changes and the increase in the number of patients with chronic diseases, home monitoring of patients’ health status will play an important role in the future healthcare system. Recent studies have shown that home telehealth approaches can significantly reduce hospital admissions and nursing days.
[0004] Since traditional measurement modes are rarely applicable outside clinical environments, a non-intrusive and unconstrained measurement system is currently emerging to allow continuous long-term monitoring of patient health conditions. These systems are intended to provide two advantages over current methods, because they generally do not require user interaction or compliance to perform measurements, and they do not interfere with the user's daily work. Non-intrusive measurement systems based on various different principles are known. These include capacitive coupling electrodes for ECG measurement of seats or car seats, capacitive respiration measurement on beds, radar, millimeter wave interferometry, and ballistocardiography (BCG) or cardiotocography (SCG) sensors integrated in beds, seats and weighing scales. Other types of non-intrusive measurement systems include wearable systems. For example, these measurement systems are implemented as watches, pendants and chest straps.
[0005] Although these measurement systems offer the above-mentioned advantages over existing methods, they also have their own set of disadvantages. Most notably, since the environment in which the measurements are taken is usually uncontrolled, the signal quality can be highly variable and unreliable. For some types of sensors, the morphology of the recorded signal can change significantly depending on the orientation and position of the user relative to the sensor. This can be observed, for example, in the case of bed-mounted BCG sensors that record the heart vibrations caused by the contraction of the heart and the ejection of blood into the aorta. In the case of wearable measurement systems, the signal can change dramatically with changes in the patient's posture.
[0006] In these cases, a reliable estimate of the instantaneous frequency (or its reciprocal: local interval length) can pose a significant challenge to algorithms originally developed for clinical applications. These algorithms are typically based on detecting specific features of the signal associated with the event of interest, such as the QRS complex in the ECG. The instantaneous frequency is then calculated by differencing the time of successive occurrences of these events. In the case of the ECG, sophisticated methods have been proposed to improve the robustness of instantaneous heart rate estimates. However, these methods are typically limited to one type of signal and require extensive prior knowledge of the expected morphology of the waveform. As mentioned above, these requirements do not apply to interference-free sensor types. Summary of the invention
[0007] It is an object of the present invention to provide an improved method for processing a signal representing a physiological rhythm of a subject.Alternatively, it is an object of the present invention to provide a method for obtaining improved information representing the physiological rhythm for different postures of a subject.
[0008] According to a first specific aspect, there is provided a method of processing a signal representing a physiological rhythm of a subject, the physiological rhythm having a periodic characteristic, the method comprising:
[0009] Receive a signal;
[0010] estimating a first set of periodic feature estimates based on the signal;
[0011] generating a priori probabilities for periodic features based on at least a subset of the first set of periodic feature estimates;
[0012] Based on the first set and the prior probabilities, a second set of period characteristic estimates is estimated.
[0013] Based on the signal, a first set of period characteristic estimates is estimated. The first set includes estimates of period characteristics for a plurality of cycles of the physiological rhythm. In most cases, the first set of period characteristic estimates provides good period characteristic estimates for some cycles, while the first set provides poor period characteristic estimates for other cycles. For example, the poor estimates are due to movement of the subject or any other type of event that negatively affects the signal accurately representing the physiological rhythm.
[0014] Based on the first set of period feature estimates, prior probabilities are generated. Based on information from the first set of period feature estimates, the prior probabilities distinguish between estimates of the period feature that are likely to occur and estimates that are unlikely to occur. The prior probabilities are based on the entire first set or on a subset of the first set. For example, the subset includes fewer poor estimates of the period feature than the first set. For example, the subset group contains only good estimates of the period feature. Physiological rhythms have some predictable properties. For example, upper and lower limits for a subject's heart rate are known, and the heart rate can only change gradually over time. In another example, upper and lower limits for a subject's spontaneous breathing rhythm are known. For example, estimates from the first group that exceed these limits are excluded from the subset.
[0015] By estimating a second set of periodic characteristic estimates based on the first set and the prior probabilities, the periodic characteristics of the physiological rhythm are estimated more accurately. In particular, the accuracy is improved for portions of the signal for which the first set initially provided a poor estimate. Thus, even in the case of movement or a change in posture of the subject, a more accurate estimate of the periodic characteristics of the physiological rhythm is provided. In this way, an accurate estimate of the physiological rhythm can be obtained even with an undisturbed and unconstrained measurement system. Optionally, the periodic characteristics of the physiological rhythm are estimated more accurately for portions of the signal for which the first set initially provided a good estimate. In this way, an improved accuracy of the estimate of the periodic characteristics is obtained.
[0016] Physiological rhythms, such as heartbeat and spontaneous breathing rhythm, are repetitive. The repetition of a physiological rhythm is characterized by a period. A period is defined as the time between two repetitions. For example, a period is the time between two events in a physiological rhythm. For example, a period refers to the time between two heartbeats, or the time between the start of two consecutive inhalations. A period has at least one period characteristic, such as the period length, the frequency of the rhythm, or the timing of a phase in the period, such as the start or end of the period. For example, a period characteristic is a physiological event that normally occurs every period. Examples of such physiological events are the opening of the atrioventricular valve, or the change of airflow from inhalation to exhalation. Such physiological events may not occur in every period due to the subject's disease or condition.
[0017] The signal includes information about the physiological rhythm. For example, the signal is generated by a measuring device, which is configured to generate the signal in response to measuring a physiological characteristic of the subject. The measuring device is configured to measure a heartbeat, a pulse, or a respiratory movement or an airflow caused by breathing. For example, the signal is an electrical signal. For example, the signal includes signals from multiple sensors. For example, multiple sensors simultaneously measure the physiological rhythm of the subject.
[0018] Based on the signal, the method estimates a first set of periodic characteristic estimates, for example by determining a local periodicity spectrum.
[0019] Due to the predictable nature of physiological rhythms, prior probabilities are generated based on the first set of period feature estimates. Alternatively, the prior probabilities are called prior probability distributions, or simply priors. The prior probabilities use statistics to give probabilities of the values of the period feature estimates in the first set.
[0020] For example, the prior probability uses additional information. For example, the additional information includes information about the limits of the subject's physiological rhythm, such as the maximum heart rate. For example, the additional information includes information about the activity that the subject was performing when the signal was generated. For example, such activity is sleeping or sitting or lying or walking, etc.
[0021] For example, the prior probability uses only a subset of the first set of period feature estimates. Part of the signal may not reflect the physiological rhythm well for various reasons. For example, the measurement device was placed improperly, or the subject moved excessively during the acquisition of that portion of the signal. The first set of period feature estimates may give very unrealistic estimates based on that portion of the signal. By excluding these unrealistic estimates, the prior probability is improved.
[0022] Based on the first group and the prior probabilities, a second group of period characteristic estimates is estimated. By using the prior probabilities, the probability of the values of the period characteristic estimates of the first group within some periods is evaluated. If the prior probabilities indicate that the probability of the value is high, the period characteristic estimate is an accurate estimate. If the prior probabilities indicate that the probability of the value is low, the period characteristic estimate is an inaccurate estimate. In this case, the period characteristic estimates are re-estimated using the values from the first group and the prior probabilities. The second group of period characteristic estimates includes, for example, some values of the period characteristic estimates from the first group and some re-estimated values of the period characteristic estimates based on the values of the first group and the prior probabilities.
[0023] In an example, the second set of period characteristic estimates is not based on the complete first set, but is based on only a portion of the first set. For example, the second set is based on a portion of the first set that provides at least a minimum accuracy. For example, the second set is not based on a portion of the first set that has a very poor estimate of the period characteristic. The poor estimate may be due to a significant error in the measurements that generated the signal. When a signal experiences such a significant error, the signal does not contain true information representing a physiological rhythm, and therefore the signal is not further processed.
[0024] The method includes determining a spectral likelihood of at least a portion of the signal. The spectral likelihood is a likelihood that a first set of periodic feature estimates corresponding to at least a portion of the signal represents a periodic feature of a physiological rhythm. The method further includes estimating a second set of periodic feature estimates based on the spectral likelihood of the signal.
[0025] Spectral likelihood converts a signal from the time domain to the frequency domain. Spectral likelihood represents a signal as a series of periodicities. Spectral likelihood compares the similarity of one periodicity of a signal to another periodicity of the signal. For example, spectral likelihood compares two adjacent periods of a signal. Spectral likelihood represents the likelihood of the value of the period length of a period in the signal. For example, spectral likelihood represents the most likely value of the period length of a period in the signal. By considering the likelihood of the value of the period length of a period in the signal, the estimation of the periodic characteristic can be more accurately estimated.
[0026] The method includes defining a fix point in a physiological rhythm. The fix point characterizes a phase in a cycle of the physiological rhythm. The method includes determining a temporal likelihood of a signal corresponding to the cycle. The temporal likelihood represents the likelihood that the signal represents the fix point. The method includes estimating the fix point based on the spectral likelihood, the temporal likelihood, and a priori probability.
[0027] In order to define the timing of the cycle of the physiological rhythm, a positioning point is defined. The positioning point characterizes the stage in the cycle of the physiological rhythm. For example, the positioning point represents the beginning of the cycle or the end of the cycle. For example, the positioning point indicates a periodic event in the physiological rhythm, which causes a significant feature in the signal. For example, a significant feature is a large positive value of the signal or a large negative value of the signal. For example, a significant feature is the maximum gradient in the signal or the zero gradient in the signal. In the case where the physiological rhythm is a cardiac rhythm, the stage in the cycle of the physiological rhythm is, for example, a heartbeat, or the closing or opening of a heart valve, such as an aortic value or a pulmonary artery value. For example, the stage of the cardiac rhythm causes a large acceleration or a large electrical pulse of the subject, thereby causing a significant feature in the signal. In the case where the physiological rhythm is a spontaneous breathing rhythm, the stage in the cycle of the physiological rhythm is, for example, the beginning or end of inhalation, the beginning of exhalation, the pause between inhalation and exhalation, the pause between exhalation and inhalation, the change of airflow direction or the pressure caused by the airflow from the subject. For example, phases of the spontaneous breathing rhythm cause large accelerations or maximum or minimum air pressures in the subject, causing noticeable features in the signal.
[0028] The temporal likelihood indicates the likelihood that a point in the signal represents a location point in the time domain. By estimating the location point based on the spectral likelihood, the temporal likelihood and the prior probability, the location point is accurately estimated. This allows the timing of the phases of the physiological rhythm to be accurately estimated.
[0029] In an embodiment, a fix point of a physiological rhythm indicates one of: i) the beginning of a cycle, ii) the end of a cycle, iii) the middle of a cycle.
[0030] In an embodiment, estimating the anchor point comprises using Maximum A Posteriori Probability (MAP) estimation. Maximum A Posteriori Probability (MAP) estimation is based on spectral likelihood, temporal likelihood and a priori probability.
[0031] By estimating the anchor points using the MAP estimate, the anchor points are estimated in a robust manner and provide the most likely value for the anchor points.
[0032] In an embodiment, the second set of period characteristic estimates more accurately estimates the period characteristics of the physiological rhythm than the first set of period characteristic estimates.
[0033] By estimating the periodic characteristics based on the first group and the prior probability, the periodic characteristics are estimated more accurately.
[0034] In an embodiment, the period characteristic comprises at least one of a period length and a timestamp.
[0035] The cycle length indicates the time between two consecutive cycles. For example, the cycle length indicates the time between two heartbeats or the time between the start of two breathing cycles. The timestamp indicates the position of the cycle in time. The timestamp indicates when the cycle occurs in time, such as when a certain characteristic point of the cycle occurs in time. For example, the timestamp provides information about the heart beat interval or heart rate variability.
[0036] In an embodiment, the physiological rhythm is the subject's spontaneous breathing rhythm. For example, the periodic feature represents the subject's breathing rate.
[0037] In an embodiment, the physiological rhythm is the subject's heart rhythm. For example, the periodic feature represents the subject's heartbeat frequency.
[0038] In an embodiment, the method comprises estimating at least one of the beat-to-beat interval and the heart rate variability of the heart rhythm based on the second set of period characteristic estimates.
[0039] Beat interval and heart rate variability are physiological parameters that provide valuable information about the subject's physical condition. By accurately estimating one or both of these parameters, accurate information can be obtained to help medical professionals assess the subject's physical condition.
[0040] In an embodiment, the signal comprises a signal from an accelerometer disposed on the chest of the subject.
[0041] Many people believe that an accelerometer placed on a subject's chest is a comfortable, non-intrusive, unconstrained way to obtain information about physiological rhythms. Accelerometers are configured to generate signals based on physiological rhythms (such as respiratory rhythms or cardiac rhythms). However, the signal of the accelerometer can also be affected by the movement of the subject or the posture of the subject. In addition, if the subject moves, the accelerometer may be displaced relative to the subject, resulting in changes in the signal. Due to these effects, the signal from the accelerometer is not as accurate as the signal of the ECG. However, by applying the method of the present invention to the signal of the accelerometer arranged on the subject's chest, the accuracy of the signal is significantly improved. This provides accurate estimation of periodic characteristics while using an easy-to-use measurement device, namely an accelerometer on the chest.
[0042] In an embodiment, the signal represents a circadian rhythm over a period of time during which the subject was asleep, such as sleeping in the range of 6-10 hours.
[0043] For example, the time period is greater than 1 hour, or greater than 2 hours, or greater than 4 hours. During such a long time period, the subject is likely to move. This movement is likely to reduce the accuracy of the signal. According to the method, the method can be applied to obtain signals in a long time period, and an improved accuracy of the periodic feature estimation can be provided in the long time period. By obtaining the periodic feature information in a long time period, more information about the subject's physical condition can be obtained.
[0044] In an embodiment, the method includes determining an occurrence of a sleep disordered breathing (SDB) event based on the second set of periodic feature estimates.
[0045] A sleep disordered breathing (SDB) event is an event in which a subject's breathing is disturbed during sleep. SDB includes breathing pauses, such as obstructive sleep apnea (OSA) or central sleep apnea (CSA). During an SDB event, for example, breathing is paused, blocked, or significantly reduced. For example, an SDB event is shallow breathing. During a shallow breathing event, the subject's airflow caused by breathing decreases significantly, such as by at least 30%.
[0046] Due to the SDB event, the subject's oxygen intake is reduced and / or the subject's carbon dioxide emissions are reduced. Due to the SDB event, one or more physiological rhythms of the subject are affected. Using the method described in the present invention to determine whether an SDB event has occurred can more accurately determine whether an SDB event has occurred. For example, this allows the use of a less accurate measuring device suitable for use at home, while still being able to determine the occurrence of an SDB event with sufficient accuracy.
[0047] In an embodiment, the method includes estimating a sleep stage of the subject based on the second set of periodic characteristics.
[0048] There are four stages of sleep: three non-rapid eye movement (NREM) sleep stages and one rapid eye movement (REM) sleep stage. The non-rapid eye movement (NREM) sleep stages are N1, N2 and N3. Sleep stages N1 and N2 are light sleep, while sleep stage N3 is deep sleep. Different sleep stages have certain differences in the physiological rhythm of the subject. For example, the heart rate is slower during N3 than during N1 or N2. For example, the heart rhythm during N3 is more regular than during REM, and the breathing during N2 and N3 is slower than during N1. For example, during REM, the breathing is more irregular than during N1-N3. Based on the cycle characteristics estimated according to the method of the present invention, the sleep stage of the subject is determined. For example, the duration and frequency of each sleep stage are determined.
[0049] In an embodiment, the signal comprises a sum of three-dimensional acceleration data, wherein the sum of three-dimensional acceleration data is band-pass filtered, for example, the cut-off points of the band-pass filter are 3 Hz and 25 Hz.
[0050] By using three-dimensional acceleration data, the signal is less affected by the subject's posture or the direction of the measurement device relative to the subject. In order to make the signal more focused on the acceleration caused by the physiological rhythm, the three-dimensional acceleration data is band-pass filtered. Accelerations below a certain cutoff point and accelerations above a certain cutoff point are unlikely to be caused by the physiological rhythm and are therefore filtered out of the three-dimensional acceleration data. For example, a band-pass filter with cutoff points of 3 and 25 Hz provides a good filtering effect. By using this band-pass filter, the three-dimensional acceleration data mostly or only contains data with frequencies between 3 Hz and 25 Hz.
[0051] In an embodiment, the method includes receiving information about the subject. The generation of the prior probability is based on the information about the subject.
[0052] The prior probability is further improved by including information about the subject. This information is provided in addition to the first set of period characteristic estimates. This information includes the subject's gender, age, weight or body mass index, etc. This information includes, for example, information about the subject's physical condition or about previous physiological rhythm measurements or about previous medical examinations.
[0053] In a second aspect of the present invention, a computer program product is provided, which comprises instructions, and when the instructions are executed by a computer, the computer is caused to perform the method of the first aspect of the present invention.
[0054] In a third aspect of the present invention, a computer-readable storage medium is provided, the medium comprising instructions which, when executed by a computer, cause the computer to perform the method of the first aspect of the present invention.
[0055] In a fourth aspect of the present invention, a system for processing a signal representing a physiological rhythm of a subject, the physiological rhythm having a periodic characteristic, is provided, the system comprising a sensor and a computer, wherein the computer is coupled to the sensor, wherein the computer is configured to perform the method of the first aspect of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0056] Exemplary embodiments will now be described, by way of example only, with reference to the following drawings, in which:
[0057] Figure 1 A first embodiment according to the invention is described;
[0058] Figure 2 describes acceleration data in a signal according to a first embodiment;
[0059] Figure 3 depicts a graph showing a preprocessed signal according to a first embodiment;
[0060] Figure 4 describes a periodic spectrum according to a first embodiment;
[0061] Figure 5 The spectral likelihood and the temporal likelihood according to the first embodiment are described;
[0062] Figure 6 Describes the computation of a representation of the posterior probability according to a first embodiment;
[0063] Figure 7 describes representations of likelihood and probability calculations;
[0064] Figure 8 Various diagrams representing steps according to a first embodiment are shown;
[0065] Fig. 9 A method according to a second embodiment of the invention is described;
[0066] Fig.10 A system for processing a signal representing a physiological rhythm of a subject according to a third embodiment of the present invention is described. DETAILED DESCRIPTION
[0067] Figure 1 An embodiment of a method according to the invention is described. In the method, a signal 100 representing a physiological rhythm of a subject 1001 is processed. The subject 1001 Fig.10 As shown. The physiological rhythm has a periodic characteristic. The method includes receiving a signal 100; estimating a first set of periodic characteristic estimates based on the signal 100; generating a priori probabilities for the periodic characteristics based on at least a subset of the first set of periodic characteristic estimates; and estimating a second set of periodic characteristic estimates based on the first set and the priori probabilities.
[0068] In this embodiment, the physiological rhythm is a heart rhythm. The signal 100 is acquired by an accelerometer 1002 attached to the chest of the subject 1001, see Fig.10 In another embodiment, the physiological rhythm is the spontaneous breathing rhythm of subject 1001 .
[0069] The cycle characteristics include cycle length and timestamp.
[0070] The method makes use of a latent variable model 108. The latent variable model 108 is a statistical model that relates observed variables to (unobserved) latent variables. The latent variable θ is a periodic feature. The observed variable is the acceleration data D in the signal 100. In order to estimate the latent variable θ from the observed data D, maximum a posteriori estimation or MAP estimation is used. The purpose of MAP estimation is to find the most likely value of the heartbeat θ given the acceleration data D: θ MAP = argmax θ {P(θ|D)} (1) P(D|θ) is the posterior, which is the probability of θ given D, calculated using Bayes’ theorem: where P(D|θ) is the likelihood, P(θ) is the prior, and P(D) is the evidence. More information on these formulas can be found in Lambert, B. (2018), A Student's Guide to Bayesian Statistics, Sage. The advantage of using MAP estimation is that it allows the use of prior probabilities during estimation, making it possible to prevent estimates that are unlikely to occur, making the method more robust.
[0071] like Figure 1 The method shown includes the following: signal 100, preprocessing step 102, spectrum conversion step 104, time conversion step 106, latent variable θ, latent variable model 108, prior probability P(θ), calculation step 110, MAP search step 112, and latent variable θ update step 114. Figure 1 Further description of the various parts of the method outlined follows, and reference is made to the other figures.
[0072] Figure 2The acceleration data D in the signal 100 according to the first embodiment is described as a function of time t. The acceleration data D shows a portion of the complete signal 100. The portion is about 6 seconds in time, from 6 hours 15 minutes 7 seconds to 6 hours 15 minutes 13 seconds. The complete signal 100 represents a physiological rhythm during a period of time, during which the sleep time of the subject 1001 is between 6 and 10 hours.
[0073] After receiving the signal 100, the signal 100 is preprocessed in a preprocessing step 102. The signal 100 includes original three-dimensional acceleration data x, y and z. In the preprocessing step 102, the original three-dimensional acceleration data is summed to obtain a one-dimensional signal a: a=x+y+z. (3) Subsequently, the one-dimensional signal a is band-pass filtered to obtain a preprocessed signal s, the cutoff points of the band-pass filter are 3 Hz and 25 Hz, and the sampling frequency of the one-dimensional signal a is 250 Hz.
[0074] Figure 3 Two periods 301, 302 of the preprocessed signal s are shown. m The periodic characteristics of the two cycles 301 and 302 are estimated for the center, and the function of the periodic spectrum with T is determined. T is the period length. A peak is found in the periodic spectrum at T = 1130ms, corresponding to the maximum similarity between s[tT] and s[t], such as the signal 303 of the preprocessed signal s in the cycle 302 is superimposed on the cycle 301, which is the signal 304 shown. In this way, the cycle length T for the cycle 301 is estimated. So in the cycle 301, the estimated cycle length in the first group is 1130ms. Repeat this process for other cycles in the preprocessed signal s. This produces a first set of periodic feature estimates based on the signal 100.
[0075] In the spectrum conversion step 104, the local periodic spectrum is calculated using the preprocessed signal s: S m (T) = p(s[tT], s[t]) (3) Where T is Figure 3 The estimated period length in , s is the preprocessed signal, and p is the periodic function. The periodic function p represents the similarity between s[tT] and s[t], for t m <t<t m +T, where t mis the center time of the spectrum. The periodicity function is based on a modified version of the autocorrelation, the mean amplitude difference function (AMDF) and the maximum amplitude pair. This calculation can be obtained from Brüser, C., Winter, S. and Leonhardt, S. (2013) "Robust interbeat intervalestimation in cardiac vibration signals", Physiological Measurement, 34 (2), pp. 123-138.
[0076] The periodic spectrum is calculated every 200 ms, thereby obtaining a periodic spectrum 400, such as Figure 4 As shown. Spectrum 400 depicts a function of estimated cycle length over time. Spectrum 400 is relatively stable within and across cardiac intervals of equal length. When the cardiac interval changes, a change occurs, i.e., a heartbeat occurs. Around these changes, the periodicity gradually changes, which is visible in spectrogram 400 as artifacts 401 and 403 at t=06:15:08 and t=06:15:13, respectively. In contrast, at 402 at approximately t=06:15:12, spectrogram 400 is relatively stable over 2.5 seconds, while the cardiac interval is 1250 ms. There is a heartbeat in the middle of this interval, and the cardiac intervals before and after this heartbeat are both 1250 ms long.
[0077] The spectral conversion step 104 improves the accurate determination of the heartbeat interval, which is the period length T between consecutive heartbeats. In order to further improve the accuracy of the heartbeat timestamp (i.e., heartbeat location), a time conversion step 106 is performed. The purpose of the time conversion step 106 is to locate the heartbeat at a specific phase in the cardiac cycle. By using a certain phase for each cycle, the time difference between the timestamps corresponds to the time difference between the R peaks in the ECG signal. In this embodiment, the maximum amplitude point of the acceleration data D in the preprocessed signal s is used.
[0078] The method includes defining a setpoint in a physiological rhythm, wherein the setpoint characterizes a phase in a cycle of the physiological rhythm. The method includes determining a temporal likelihood of the signal 100 corresponding to the cycle. The temporal likelihood represents the likelihood that the signal 100 represents the setpoint. The method includes estimating the setpoint based on the spectral likelihood, the temporal likelihood, and a priori probability.
[0079] For example, an anchor point of a physiological rhythm represents the beginning of a cycle of the physiological rhythm, the end of a cycle, or the middle of a cycle.
[0080] The latent variable model 108 calculates how well the data fits the latent variable θ in the form of a likelihood P(D|θ). The latent variable model 108 takes as input the values of the latent variables (heartbeat timestamps and heartbeat intervals) and calculates how well the spectral and time-transformed data fits these timestamps and intervals using the following relationship: Among them, f is the likelihood function, D T Transform data for time, D S is the spectrum conversion data, f T is the likelihood function, f S is the spectral likelihood function. The first equation states, without loss of generality, that the likelihood is a function of D and θ, while the second equation factors this function into the modeling choice we made.
[0081] Then calculate the overall spectrum likelihood f Stot as follows: The latter equation decomposes the likelihood into a product of functions (f S ), which is not unreasonable, because the periodic spectrum (D S ) depends only locally on the occurrence of a heartbeat. The heartbeat timestamp t i and the associated heart beat interval T i The likelihood of is calculated as follows: f S (D S , t i , T i )=(Π m S m (T i )) 1 / p (6) For all t i ≤t m ≤t i +T i where m is the number of spectra used and p is the number of spectra used. The calculation includes the heart beat interval (t i to i +T i ) The center time is t m The multiplication of all spectra of makes (6) similar to the joint probability. The exponent 1 / p represents the geometric mean, which prevents short intervals (smaller p) from being favored over longer intervals (larger p).
[0082] The method includes determining a spectral likelihood of at least a portion of the signal 100. The spectral likelihood indicates a likelihood that a first set of periodic feature estimates corresponding to at least a portion of the signal 100 represents a periodic feature of a physiological rhythm. The method includes estimating a second set of periodic feature estimates 100 based on the spectral likelihood of the signal. The second set of periodic feature estimates more accurately estimates the periodic feature of the physiological rhythm than the first set of periodic feature estimates.
[0083] Figure 5 In the upper part of the figure, two heartbeat intervals 501 and 502 are shown as horizontal lines. The duration of the first interval 501 is 880 ms. The first interval 501 covers 4 vertical lines, i.e. 4 spectral lines (p=4), all of which have a peak value at 880 ms. Therefore, the spectral likelihood (6) reaches a maximum value at T=880 ms. The duration of the second interval 502 is 1225 ms. The second interval 502 covers 6 vertical lines, i.e. 6 spectral lines, all of which have a peak value at 1225 ms. Therefore, the spectral likelihood is also maximum at T=1225 ms.
[0084] For the total time likelihood f Ttot , which is a similar decomposition as used for the spectral likelihood: The latter equation decomposes the likelihood into the function f T The product of is equal to t i The value of the temporal likelihood at : f T (D T , t) = llh(t) (8)
[0085] exist Figure 5 The following figure shows the preprocessed signal s, which is divided into two parts 503 and 504. Part 503 shows Figure 3 The positive value of the preprocessed signal s in FIG. 504 shows Figure 3 The negative values of the preprocessed signal s are represented as positive values. Figure 5 Two consecutive heartbeat timestamps are shown, vertical dashed lines 505 and 506. The probability llh(t i ) and llh(t i +T i ) are located at the maximum values at lines 505 and 506 in portion 503 of the preprocessed signal s, respectively. Figure 5 The spectral likelihood and temporal likelihood of the two heartbeats in are at their maximum, so the overall likelihood (6) is also at its maximum.
[0086] The prior probability gives information about the probability of values of the latent variable θ and helps prevent foolish estimates, such as a 100 ms heart beat interval (corresponding to 600 bpm). The prior probability uses knowledge of the distribution of the latent variable θ over the population of measurements. In addition to knowing that subject 1001 is human, the prior probability is explicit, for example, by using the knowledge that subject 1001 is lying down, which discourages shorter heart beat intervals (due to higher heart rate). Optionally, the prior probability takes into account the probability of the sequence, such as by using a Markov model or information from a neural network model. For example, the prior probability is a personalized prior probability.
[0087] For example, the prior probability is a Gaussian prior Where μ = μ(t). The value of σ is empirically set to twice the standard deviation of the difference between the input and output of the first set of median filter cycle lengths.
[0088] Based on the prior likelihood, spectral likelihood and temporal likelihood, the posterior probability is calculated and the posterior probability over the latent variable θ is maximized. In order to reduce the computational complexity of this problem, the method is to decompose the problem into smaller sub-problems and solve each sub-problem cyclically. Assume that there is a partial solution θ n : θ n =(t n , T n-1 )=(t1,...,t n , T1, ..., T n-1 ) (9) θ1=t1,see Figure 6 . n Expanded to a heart beat interval, and knowing θ n =P(θ n |D), then the extended solution θ n+1 The posterior probability of can be calculated as follows: P(θ n+1 |D)=P(θ n , T n |D)=P(T n |D,θ n )P(θ n |D) (10) The first equation uses (9) and t n+1 =t n +T n , and the second equation uses P(A, B) = P(A|B)P(B). Equation (10) is the basis of the MAP search engine because it is an expression for optimizing the posterior probability using MAP estimation. The right side of (10) contains the posterior P(T n |D,θ n), in calculation step 110, Bayes' theorem is used to calculate: Here we discard P(D) since it is constant when maximizing the posterior, leaving only the numerator. The numerator consists of the prior P(T n |θ n ) and the likelihood P(D|T) calculated using the spectral likelihood function and temporal likelihood function of equations (6) and (7) n ,θ n ) to obtain: P(D|T n ,θ n )=f S (D S , t n , T n )f T (D T , t n+1 ) (12)
[0089] The numerator of (11) is calculated as Figure 7 As shown, Figure 7 The likelihood and probability are shown in the logarithmic domain, making the multiplications in (11) and (12) additions, since if A = BC, then log(A) = log(B) + log(C)). Figure 7 The top graph 700 is the time log-likelihood f T , the middle graph 702 shows the spectral log-likelihood f S With the logarithmic prior P(T n |θ n ), while the bottom graph 703 is the posterior P(T n |D,θ n ) is the logarithm of .
[0090] Solve for θ in the expanded part n When T is not considered n All values of , and only consider T with peak values in the posterior probability (13) n value, thereby discretizing the search space. Figure 7 This is outlined in . Figure 7 The peak values are indicated by arrows 704 , 705 , 706 .
[0091] The constrained lookahead recursive evaluation starts at the starting time t1 and forms the first partial solution θ1. The posterior probability of θ1 is calculated as: P(θ1|D)=f T (D T , t1) (13) It is the initial value in equation (13). Starting from the first partial solution, a tree search is performed, where the branches are the peaks in the posterior (such as Figure 7 As shown), and the leaves are partial solutions θ d , where d is the look-ahead depth. For example, d = 8 in order to achieve a good balance between performance and computational load. During the recursive tree exploration, the posterior probability P(θ d |D). Then, the path with the highest posterior probability is selected, and θ1 is expanded with the first heartbeat interval of the selected path to obtain θ2. This process is repeated until the partial solution covers all the data and becomes a complete solution.
[0092] Formula (11) defines the interval T for a single heartbeat n The posterior probability P(T n |D,θ n ). The posterior probability of the numerator of (11) is approximately: Q(T n )=P(D|T n ,θ n )P(T n |θ n )=f S (D S , t n , T n )f T (D T , t n+1 )P(T n |θ n ) (14)
[0093] The method includes estimating the timestamp using a maximum a posteriori probability (MAP) estimate. The maximum a posteriori probability (MAP) estimate is based on spectral likelihood, temporal likelihood and a priori probability. The MAP estimate is calculated in a MAP search step 112.
[0094] The result of the MAP search step 112 is used to update the latent variable θ in the calculation step 114 .
[0095] Figure 8 Various diagrams representing steps in a method according to the invention are shown. These diagrams are functions of time t.
[0096] The first graph 801 starts with the raw acceleration data D, x, y and z. The lines are plotted to visualize the adjusted offset. The peak to peak acceleration is about 50 mg.
[0097] The second figure 802 shows the preprocessed one-dimensional signal. This is the preprocessed signal s.
[0098] The third diagram 803 shows the spectral transformation of the pre-processed signal s.
[0099] The fourth diagram 804 shows the time transformation of the pre-processed signal s.
[0100] The fifth figure 805 shows the result of the MAP estimation. The MAP estimation provides the most likely value of the latent variable θ, namely the interbeat interval (IBI) in [ms] as a horizontal line 807 and the heartbeat timestamp as a dashed vertical line 808. For clarity, only two lines 807 and only two vertical lines 808 are numbered. This method allows to estimate the interbeat interval and the heart rate variability of the heart rhythm based on the second set of period feature estimates.
[0101] The sixth graph 806 shows an ECG of the same heart rhythm represented by the signal. Comparison of the fifth graph 805 and the sixth graph 806 shows that the heart beat intervals and heart beats estimated according to the method of the present invention accurately match the ECG.
[0102] In an embodiment, the method includes determining the occurrence of a sleep disordered breathing (SDB) event based on the second set of cycle feature estimates. An SDB event may be identified by typical changes in the heart rhythm.
[0103] In an embodiment, the method includes estimating a sleep stage of subject 1001 based on the second set of periodic characteristics.
[0104] Fig. 9 A method according to a second embodiment of the present invention is described. The method is used to process a signal representing a physiological rhythm of a subject 1001. The physiological rhythm has a periodic characteristic. The method includes receiving the signal 100 in step 901. The method includes estimating a first set of periodic characteristic estimates based on the signal 100 in step 902. The method includes generating a priori probabilities for the periodic characteristics based on at least a subset of the first set of periodic characteristic estimates in step 903. The method includes estimating a second set of periodic characteristic estimates based on the first set of periodic characteristic estimates and the prior probabilities in step 904.
[0105] In optional step 905, the method includes determining a spectral likelihood of at least a portion of the signal 100. The spectral likelihood indicates a likelihood that a first set of periodic feature estimates corresponding to at least a portion of the signal 100 indicates periodic features of a physiological rhythm. The method includes estimating a second set of periodic feature estimates 100 based on the spectral likelihood of the signal.
[0106] In optional step 906, the method includes defining a fix point in the physiological rhythm. A fix point is a phase in a cycle of the physiological rhythm. The method includes determining a temporal likelihood of the signal 100 corresponding to the cycle. The temporal likelihood represents the likelihood that the signal 100 represents the fix point. The method includes estimating the fix point based on the spectral likelihood, the temporal likelihood, and the prior probability.
[0107] Fig.10 A system 1000 for processing a signal 100 representing a physiological rhythm of a subject 1001 1001 is described. The physiological rhythm has a periodic characteristic. The system 1000 includes a sensor 1002 and a computer 1010. The computer 1010 is coupled to the sensor 1002 via a signal line 1008. The computer 1010 is configured to perform the method of any one of the above embodiments. The sensor 1002 is an accelerometer attached to the chest of the subject 1001 1001. The sensor 1002 is attached to the chest with a belt 1002. The subject 1001 1001 is lying in bed and is asleep. The computer 1010 has an input device 1012 connected to the sensor 1002 and receives the signal 100 from the sensor 1002. The input device 1012 forwards the signal 100 to a processor 1014. The processor 1014 processes the signal 100 according to the method of any one of the above embodiments. The processor 1014 is connected to an output device 1016. The output device 1016 outputs an output signal about the periodic characteristic estimated by the method. For example, the output device 1016 generates a video signal or a Bluetooth signal or any other type of signal suitable for conveying information about the estimated periodic characteristic to a person.
[0108] The computer 1010 executes the computer program product, which includes instructions, which, when executed by the computer 1010, cause the computer 1010 to perform the method according to the present invention. For example, the instructions are provided to the computer 1010 via a computer-readable storage medium containing instructions. For example, the computer 1010 has a memory storing the computer program product. The processor 1014 is configured to receive the instructions of the computer program product from the memory.
[0109] By studying the drawings, the present disclosure and the appended claims, those skilled in the art can understand and implement variations of the disclosed embodiments by practicing the principles and techniques described herein. In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude pluralities. A single processor or other unit may perform the functions of several items listed in the claims. The fact that certain measures are listed in mutually different dependent claims does not indicate that a combination of these measures cannot be used for advantageous purposes. The computer program product may be stored or distributed on a suitable medium, such as an optical storage medium or a solid-state medium provided together with or as part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunications systems. Any reference symbols in the claims should not be interpreted as limiting their scope.
Claims
1. A method for processing a signal (100) representing a physiological rhythm of a subject (1001), the physiological rhythm having a periodic characteristic (T, t i ), the method comprising: receiving the signal (100); Based on the signal (100), a first set of periodic features (T, t i )estimate; Based on the first set of periodic features (T, t i ) estimates, generating a periodic feature (T, t i ) (P(θ)); Based on the first group and the prior probability (P(θ)), a second group of periodic features (T, t i )estimate; Determining a spectral likelihood (f) of at least a portion of the signal (100) S ); The spectral likelihood (f S ) represents the first set of periodic features (T, t) corresponding to at least a portion of the signal (100) i ) estimates the periodic feature (T, t i )’s likelihood; The second set of periodic features (T, t i ) is estimated based on the spectral likelihood (f S ); defining a location point in the physiological rhythm, wherein the location point represents a phase in a cycle of the physiological rhythm; Determine the time likelihood (f) of the signal (100) corresponding to the period T ); The temporal likelihood (f T ) represents the likelihood that the signal (100) represents the positioning point; Based on the spectral likelihood (f S ), the time likelihood (f T ) and the prior probability (P(θ)), estimate the positioning point.
2. The method according to claim 1, comprising: wherein estimating the anchor point comprises using maximum a posteriori probability (MAP) estimation, The maximum a posteriori probability (MAP) estimate is based on the spectral likelihood (f S ), the time likelihood (f T ) and the prior probability P(θ)).
3. The method according to claim 1 or 2, wherein: Compared with the first set of periodic features (T, t i ) estimates that the second set of periodic features (T, t i ) estimates the periodic characteristics (T, t) of the physiological rhythm more accurately i ).
4. The method according to any one of the preceding claims, wherein the periodic characteristic (T, t i ) includes the cycle length (T) and the timestamp (t i ) 5. The method according to any of the preceding claims, wherein the physiological rhythm is the spontaneous breathing rhythm of the subject (1001).
6. The method according to any of the preceding claims, wherein the physiological rhythm is a cardiac rhythm of the subject (1001).
7. A method according to any one of the preceding claims, comprising: Based on the second set of periodic features (T, t i ) estimates at least one of the heart beat interval and heart rate variability of the heart rhythm.
8. The method according to any one of claims 5 to 7, wherein the signal (100) comprises a signal from an accelerometer arranged on the chest of the subject (1001).
9. A method according to any one of the preceding claims, comprising: Based on the second set of periodic features (T, t i ) estimates and determines the occurrence of sleep disordered breathing (SDB) events.
10. A method according to any one of the preceding claims, comprising: Based on the second set of periodic features (T, t i ) estimates and determines the sleep stage of the subject (1001).
11. A computer program product comprising instructions which, when executed by a computer (1010), cause the computer (1010) to perform the method according to any one of claims 1 to 10.
12. A computer-readable storage medium comprising instructions, which, when executed by a computer (1010), cause the computer (1010) to perform the method according to any one of claims 1 to 10.
13. A system (1000) for processing a signal (100) representing a physiological rhythm of a subject (1001), the physiological rhythm having a periodic characteristic (T, t i ), the system (1000) comprises a sensor (1002) and a computer (1010), wherein the computer (1010) is coupled to the sensor (1002), wherein the computer (1010) is configured to execute the method according to any one of claims 1 to 10.
Citation Information
Patent Citations
Apparatus, system, and method for monitoring physiological signs
CN101489478A
A method of processing a signal representing a physiological rhythm
CN104144636A
Multi-target respiratory rate estimation method, device and equipment and storage medium
CN114767087A
Measuring respiration or other periodic physiological processes
US20160171170A1
System and method for smart pooling
US20230041884A1
