System and method for determining breathing rate of subject

Abstract

A system and method for determining a respiratory rate of a subject is provided. A respiratory signal is received and a derivative of a low-pass filtered version of the respiratory signal is acquired to obtain a derivative signal. In one example, a slope summing filter is applied to transform the derivative signal, and a plurality of breathing intervals are identified from the transformed signal. Respiratory characteristics in the filtered respiratory signal are obtained using the transformed signal. Appropriate breathing intervals are selected based on breathing characteristics, and a breathing rate is calculated based on those selected breathing intervals.

Description

Technical Field

[0001] The present invention relates to monitoring the breathing rate of a subject, and in particular to monitoring the breathing rate of a subject during sleep. Background Art

[0002] Breathing during sleep is an involuntary action regulated by the autonomic nervous system. Sleep apnea is a pathological condition that causes the cessation of breathing during sleep. Sleep apnea has been found to be associated with poor sleep quality, fatigue, drowsiness, stroke, high blood pressure, heart attack, and many other complications. It is estimated that more than 1 billion people worldwide suffer from sleep apnea. However, most cases remain unidentified.

[0003] Although the diagnosis of sleep apnea is technically not complex, the procedure is far less clinically used than it should be. Limitations of diagnosis in sleep laboratories include expensive medical expenses, complex usage (depending on interpretation by experienced operators), and high invasiveness to the patient. Therefore, a non-intrusive, cost-effective, home-based screening technology solution for sleep apnea would be of great significance. A technology solution that is feasible for preliminary identification of sleep disorders in patients at home without sacrificing user comfort would be particularly desirable.

[0004] Previous studies have proposed many different methods to estimate respiratory rate. The easiest method is to estimate respiratory rate by counting the number of peaks of a filtered respiratory waveform over a specified time period. This method requires minimal computation and is able to detect other inhalation / exhalation characteristic points, such as when inhalation / exhalation begins or ends. However, the raw sensor data is often affected by motion artifacts, which impairs the performance of this method.

[0005] Respiration rate can also be estimated in the frequency domain, usually based on a periodogram. Implementations of the periodogram can be efficiently computed using Fast Fourier Transforms, with a complexity of order O(NlogN), where N represents the length of the data used for the FFT calculation. A disadvantage of the periodogram is that its resolution is limited by the length of the data and the length of the FFT. Therefore, accurate respiration rate estimation requires longer data, which makes the calculation power-consuming, which is particularly problematic for battery-based solutions. In addition, frequency-based methods cannot estimate the duration of inspiration or expiration, because this information is lost when the signal is transformed into the frequency domain.

[0006] In addition to respiration rate (ie, frequency domain information), the ability to estimate the inspiration to expiration ratio (I / E ratio) (ie, time domain information) would benefit clinical needs as an indicator of user status. Summary of the invention

[0007] The invention is defined by the claims.

[0008] According to an example according to one aspect of the present invention, there is provided a system for determining a breathing rate of a subject, the system comprising a controller configured to:

[0009] receiving a breathing signal of the subject;

[0010] low-pass filtering the respiratory signal to obtain a filtered respiratory signal;

[0011] determining a transformed signal by at least obtaining a derivative signal of the filtered respiration signal;

[0012] determining a plurality of breathing intervals from the transformed signal;

[0013] determining a respiratory characteristic in the filtered respiratory signal for the determined plurality of respiratory intervals;

[0014] selecting a breathing interval based on the breathing characteristics; and

[0015] The breathing rate is calculated based on the selected breathing interval.

[0016] The system analyzes a breathing signal (e.g., from a breathing sensor signal or a piezoelectric sensor signal that detects muscle movement) to calculate a breathing rate. The breathing rate can be used for various purposes, such as for determining a sleep apnea condition based on the breathing rate. The system can efficiently and non-intrusively track real-time breathing rate, particularly during sleep. The algorithm used to determine the breathing rate is computationally simple, e.g., performed primarily in the time domain, which makes the implementation of the device highly affordable. This also enables the acquisition of time-domain breathing characteristics as well as the underlying calculated breathing rate.

[0017] The controller may be configured to determine the occurrence and / or severity of sleep apnea based at least on the calculated breathing rate. The breathing rate and its changes over time are thereby used to detect sleep apnea events and / or their severity. For example, during a sleep apnea event, the breathing rate drops sharply to near zero within a few seconds. If the device detects that the breathing rate is below normal for a certain period of time, this is a strong indicator of a sleep apnea event. In addition, movement from the chest or abdomen caused by breathing effort may be detected. In this case, the signal pattern during apnea becomes significantly different from normal breathing, which increases the prediction change of inspiratory initiation.

[0018] The controller is, for example, configured to determine the transformed signal by:

[0019] applying a derivative filter to the filtered respiration signal and setting negative derivative filter values ​​to zero to obtain a derivative signal; and

[0020] Applies a slope sum filter to the derivative signal.

[0021] The derivative function generates a pulse for each respiratory interval. The summation of the derivative values ​​using a slope sum filter gives an improved signal to noise ratio for subsequent peak detection (for identifying sequential respiratory intervals).

[0022] The controller is configured, for example, as follows:

[0023] detecting peaks in the transformed signal;

[0024] determining peak and trough locations in the filtered respiratory signal corresponding to detected peaks in the transformed signal; and

[0025] The breathing characteristics are determined based on the determined peak and valley locations.

[0026] The corresponding peak and valley locations are found, for example, by searching back from the time of the peak in the transformed signal to the nearest maximum and minimum values ​​of the filtered respiratory signal.

[0027] For example, by searching the filtered respiration signal for the closest local maxima and minima that occur before the peaks of the slope sum filter SSF signal, peaks and valleys are found in the filtered respiration signal (the timing of these is a first example of a respiration characteristic).

[0028] The controller is, for example, configured to determine the height difference between inhalation and exhalation as the amplitude difference between a determined peak position and an adjacent determined valley position in the filtered respiratory signal. These are further examples of respiratory characteristics. Thus, the inhalation / exhalation height difference may also be derived from the closest maximum and minimum values ​​of the filtered respiratory signal.

[0029] The controller is configured, for example, to determine the inhalation duration and the exhalation duration based on the determined peak and trough positions in the filtered respiratory signal.The controller may then be configured to calculate an inhalation to exhalation ratio based on the determined inhalation duration and the determined exhalation duration.

[0030] The controller is, for example, configured to select a plurality of breathing intervals according to the breathing characteristics, and derive a candidate breathing interval as an average or a median of a plurality of most recently selected breathing intervals, and calculate the breathing rate based on the candidate breathing intervals.

[0031] Therefore, suitable breathing intervals are identified and selected based on the breathing characteristics (e.g., a suitable height difference between inhalation and exhalation). This facilitates the selection of breathing intervals with suitable quality. In addition, these suitable breathing intervals can be used to calculate candidate breathing intervals for deriving the breathing rate.

[0032] In one example, the controller is configured to derive a final breathing interval by smoothing a plurality of selected breathing intervals, and the controller is further configured to calculate a breathing rate based on the final breathing interval.

[0033] The controller is, for example, configured to select a breathing interval having the following:

[0034] Altitude difference within a predetermined altitude range; and / or

[0035] a steadily increasing suction upstroke; and / or

[0036] a respiration signal or waveform having a mean within a predetermined mean range; and / or

[0037] a respiratory signal or waveform having one or more high-order parameters within a predetermined corresponding parameter range; and / or

[0038] Cosine similarity of consecutive respiratory waveforms of the respiratory signal within a predetermined similarity range.

[0039] In this way, high quality breathing intervals are identified and they will be the ones used for determining the breathing rate and optionally also for assessing the sleep apnea condition.High order parameters are for example skewness or kurtosis.

[0040] The controller is, for example, configured to derive the final breathing interval FBI by smoothing the candidate breathing intervals.

[0041] Smoothing is performed, for example, using:

[0042] Where FBI[n] is the nth final breathing interval, FBI[n-1] is the previous n-1th final breathing interval, and BI[n] is the nth candidate breathing interval.

[0043] In one embodiment, the most recent calculation from the candidate breath interval is given a weight of 3 compared to the weight of 1 for the previously calculated final breath interval. The breathing rate can then be calculated based on 60 / FBI[n].

[0044] The controller is configured, for example, to determine the inhalation duration and the exhalation duration. The parameters of the inhalation duration and the exhalation duration can be used as an indicator of the user's respiratory system condition and can enhance sleep apnea determination. For some patients with sleep apnea events or asthma, the inhalation to exhalation ratio may be abnormal.

[0045] The controller is, for example, configured to predict the timing of the start of the next inspiration based on the most recent trough position, a breathing interval based on the length of a previously selected breathing interval, and a phase delay caused by low pass filtering.

[0046] The timing of the start of the next inspiration can be predicted, for example, using the following formula:

[0047] t_PO=t_MIN+BI-IIR_PO

[0048] t_PO is the predicted start time.

[0049] t_MIN is the timing of the most recent breathing signal minimum (i.e., the timing of the most recent valley position). Such a breathing signal minimum can be detected without requiring high signal quality. Therefore, even the timing of a valley position from a signal with low quality can be considered as t_MIN.

[0050] The BI is the breath interval used, such as the average length (or other statistical measure) of previously selected breath intervals over a predefined period.

[0051] IIR_PO is a phase delay, such as an IIR filter phase offset determined from a best fit function.

[0052] In this way, the system additionally predicts the start time of the next inhalation to enhance sleep apnea determination. A comparison between the predicted timing of the start of the next inhalation and the actual timing of the start of the next inhalation can be made. A longer than expected (e.g., more than 0.3 seconds) delay after the predicted start of the next inhalation of a breathing interval can be used as an indicator of a possible sleep apnea event. The occurrence and / or severity of sleep apnea can be based on, for example, a comparison between the predicted timing of the start of the next inhalation and the actual timing of the start of the next inhalation, and / or a calculated inspiration to exhalation ratio.

[0053] The system may also be used for ventilation or pacing breathing, and for detecting or analyzing sleep apnea events, by predicting the onset of the next inspiration.

[0054] The phase delay (particularly the IIR phase shift) is caused by the filtering operation (low pass filtering) performed on the signal. For example, the filter may be an infinite impulse response (IIR) filter. By subtracting the phase delay, the distortion produced by filtering (in a specific frequency range used to track the breathing rate during sleep, such as 0.1 Hz to 0.6 Hz, or 6 breaths per minute to 36 breaths per minute) is compensated so that an accurate estimate of the start of the next inspiration is obtained.

[0055] The next inspiration occurs when the respiratory signal has reached its minimum (valley) and is about to increase. Although the minimum (valley) can be detected by directly comparing the value with the adjacent data, subsequent data samples are required after the minimum for this purpose. Therefore, a time delay is inevitable. Moreover, the preprocessing (filtering) applied to the raw respiratory signal will introduce additional time delays as mentioned above.

[0056] The prediction is based, for example, on the assumption that the breathing interval should remain almost constant for a short duration (e.g., about 10 to 20 seconds). Thus, given the previous few breathing intervals, the occurrence of the next inhalation can be predicted. In particular, the estimate of the next breathing interval can be taken as the average of the breathing intervals over the most recent predefined period.

[0057] IIR filters (e.g., Butterworth filters) do not have a constant phase delay across frequency. A best fit curve, such as a second order polynomial curve, can be used and applied to the phase delay within the effective respiratory frequency range (e.g., 0.1 Hz to 0.6 Hz). Using the known current respiratory rate, the appropriate phase delay can then be subtracted based on the curve fit. Thus, the additional time delay described above can be eliminated.

[0058] The controller is configured, for example, to determine an inhalation plateau and / or an exhalation plateau based on the height difference, the peak position, and the valley position.

[0059] These plateaus are defined, for example, as time periods during which the respiratory signal (low-pass filtered signal) is below a lower threshold and time periods during which the respiratory signal is above an upper threshold.

[0060] The controller is configured, for example, to determine the inhalation duration and the exhalation duration, and to determine the inhalation hold duration and the exhalation hold duration. The parameters of the inhalation duration and the exhalation duration can be used as an indicator of the user's respiratory system condition. For some patients with asthma, the ratio of the inhalation to exhalation duration may be abnormal and therefore have clinical significance for the user.

[0061] Thus, the system can achieve efficient real-time estimation of breathing rate during sleep, as well as I / E ratio estimation and inspiratory onset prediction. Those different estimates and predictions can be used together or individually to determine the occurrence and / or severity of sleep apnea, thereby enhancing the accuracy of the determination. The computational complexity of the algorithm means that the system can be implemented in an embedded system. Typically, the resources of embedded systems such as ARM Cortex-M4f are limited in both computing power and memory space.

[0062] The present invention also provides a respiratory monitoring system, which comprises:

[0063] a sensor device for generating a breathing signal of a subject; and

[0064] A system for determining breathing rate as defined above.

[0065] The present invention also provides a computer-implemented method for determining a respiration rate of a subject, the method comprising:

[0066] receiving a breathing signal of the subject;

[0067] low-pass filtering the respiratory signal to obtain a filtered respiratory signal;

[0068] obtaining a transformed signal by at least obtaining a derivative of the filtered respiratory signal;

[0069] determining a plurality of breathing intervals from the transformed signal;

[0070] determining a respiratory characteristic in the filtered respiratory signal for the determined plurality of respiratory intervals;

[0071] selecting a breathing interval based on the breathing characteristics; and

[0072] The breathing rate is calculated based on the selected breathing interval.

[0073] The method involves, for example, calculating the transformed signal by:

[0074] applying a derivative filter to the filtered respiration signal and setting negative derivative filter values ​​to zero to obtain a derivative signal; and

[0075] Applies a slope sum filter to the derivative signal.

[0076] The method obtains the breathing characteristic, for example, by detecting a peak in the transformed signal and searching back from the time of the peak in the transformed signal to the nearest maximum and minimum values ​​of the filtered breathing signal.

[0077] The invention also provides a computer program comprising computer program code means adapted to implement the method defined above when said program is run on a computer.

[0078] These and other aspects of the invention will be apparent from and elucidated with reference to the embodiment(s) described hereinafter. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] For a better understanding of the present invention, and to more clearly show how the present invention may be implemented, reference will now be made, by way of example only, to the accompanying drawings.In some of the drawings, AU stands for "arbitrary units".

[0080] Figure 1 A respiratory monitoring system is shown;

[0081] Figure 2 shows that after receiving the breathing signal, Figure 1 the processing steps performed by the system;

[0082] Figure 3 The raw respiratory signal (top panel), the low-pass filtered signal (bottom panel), and the derivative and SSF of the low-pass filtered signal (middle panel) are shown;

[0083] Figure 4 The raw respiratory signal (top panel), low-pass filtered signal (middle panel), and derivative and SSF (bottom panel) are shown;

[0084] Figure 5 The raw respiration signal (top panel), the low-pass filtered signal (middle panel), and the derivative and SSF (bottom panel) are shown; and

[0085] Figure 6 Shown are the raw respiration signal (top panel), the low-pass filtered signal (middle panel), and the derivative and SSF (bottom panel). DETAILED DESCRIPTION

[0086] The present invention will be described with reference to the accompanying drawings.

[0087] It should be understood that the detailed description and specific examples, although indicating exemplary embodiments of the apparatus, system and method, are intended only for illustrative purposes and are not intended to limit the scope of the invention. These and other features, aspects and advantages of the apparatus, system and method of the present invention will become better understood from the following description, the appended claims and the accompanying drawings. It should be understood that the drawings are merely schematic and are not drawn to scale. It should also be understood that throughout the drawings, the same reference numerals are used to indicate the same or similar parts.

[0088] The present invention provides a system and method for determining the respiratory rate of a subject. A respiratory signal is received, and a derivative of a low-pass filtered version of the respiratory signal is obtained to obtain a derivative signal. In one example, a slope sum filter is applied to transform the derivative signal, and a plurality of respiratory intervals are identified from the transformed signal. Respiratory characteristics in the filtered respiratory signal are obtained using the transformed signal. Appropriate respiratory intervals are selected based on the respiratory characteristics, and a respiratory rate is calculated based on those selected respiratory intervals.

[0089] The present invention can be implemented by a smart device that can efficiently and non-intrusively track real-time breathing rate during sleep. A breathing signal is received from a sensor, for example for measuring muscle movement caused by breathing. The method can be implemented by a computationally efficient algorithm, which makes the implementation of the device highly affordable. The algorithm performs analysis primarily in the time domain, so that in addition to the breathing rate, the device can also derive, for example, a time-based inspiration to expiration ratio from the breathing signal. This contains information that has important clinical implications for the user.

[0090] Furthermore, in some embodiments, the system may use a mathematical model to estimate a subsequent upcoming inspiratory event, which may extend the applicability of the invention to applications that detect when breathing begins, such as ventilation or paced breathing.

[0091] Respiratory rate is meaningful for identifying sleep apnea, and the present invention enables an affordable and non-intrusive way to identify potential patients with sleep apnea. Early screening for sleep apnea can help reduce morbidity, mortality, and the economic burden on the healthcare system. The present invention may also underpin the need for home ventilation technology solutions. Prediction of inspiratory launches and inspiratory to expiratory ratios may also be used to identify sleep apnea.

[0092] Figure 1 A breathing monitoring system 10 is shown comprising a sensor arrangement 12 for generating a breathing signal BS of a subject and a system 14 for determining a breathing rate BR of the subject. The system 14 comprises a controller which processes the breathing signal BS and derives the breathing rate and optionally also provides for the detection of sleep apnea events and determines the severity of detected sleep apnea events.

[0093] Figure 2 The processing steps after receiving the breathing signal are shown. The breathing signal can be obtained by any suitable sensor that forms part of a sleep monitoring application for detecting breathing vibrations or flow, such as a pillow or mattress, or a bedside system using radar or ultrasound, or a piezoelectric sensor, or an airflow sensor. The raw sensor data is sent to the controller for further processing.

[0094] In step 20, the respiration signal is low-pass filtered. This is a pre-processing stage. Previous studies usually apply a low-pass or band-pass filter to the raw sensor signal to attenuate unwanted frequency components. The lower limit of the low-pass filter or band-pass filter is set to 0.05 Hz or 0.1 Hz, for example, and the upper limit of the filter is usually set to 0.5 Hz or 1 Hz. For example, a low-pass second-order Butterworth filter can be used to remove high-frequency noise, where the cutoff frequency is 0.5 Hz.

[0095] Note that if a bandpass filter is applied, this includes the lowpass filter function. Therefore, the definition of "lowpass filter" is intended to include both lowpass and bandpass filtering.

[0096] In steps 21 and 22, the low pass filtered signal is transformed by obtaining a derivative signal using a derivative filter in step 21 and obtaining a transformed SSF (Slope Sum Filter) signal by applying an SSF filter to the derivative signal.

[0097] The raw sensor signal may have a drifting baseline and subtle perturbations, which makes it challenging for the algorithm to detect true peaks caused by breathing. The transform exploits waveform properties.

[0098] First the derivative filter is applied to the low-pass filtered signal. Negative values ​​of the derivative are set to zero:

[0099] y[t]=MAX(0,(x[t+window]-x[t-window]) / (2*window))

[0100] t is the current time of the signal x[t]. y[t] is the derivative signal at the current time, and x[t+window] and x[t-window] are the values ​​of the window duration after and before y[t], respectively. The window width is chosen to achieve optimal performance, for example 0.4 seconds. The function MAX() returns the larger of the two parameters. Since the first parameter is zero, y[t] will always be non-negative.

[0101] A slope sum filter (SSF) is applied to the derivatives. The SSF function sums all values ​​within a fixed width moving window (e.g., 0.8 seconds). Since all derivatives are non-negative, the value calculated from the SSF represents the increasing momentum of the filtered signal. This is because the SSF function adds up the most recent non-negative values ​​derived from the derivative filter within the most recent 0.8 seconds, for example. Therefore, the more the derivative signal increases, the larger the value calculated by the SSF.

[0102] The SSF value is calculated, for example, as:

[0103] SSF[t]=y[t-window]+y[t-window+1]+…+y[t+window-1]+y[t+window]

[0104] The transformations (derivative and SSF) have the advantage of an improved signal-to-noise ratio for peak detection, so that small disturbances are attenuated.

[0105] Figure 3The original respiration signal (top panel), the low pass filtered signal (middle panel), the scaled derivative of the low pass filtered signal (bottom panel, graph 30), and the SSF signal (bottom panel, graph 32) are shown. The value of the SSF signal 32 is much larger than the original derivative signal 30. The difference is much larger than represented because the derivative signal is multiplied by a constant (20 in this example) so that the scaled derivative signal can be seen together with the SSF signal. Similarly, in Figures 4 to 6 In the bottom panel, the derivative signal has also been multiplied by a constant to be shown together with the SSF signal.

[0106] Even though the raw respiration signal is smoothed by low pass filtering, as depicted in the middle panel, it still has small perturbations that can confuse the peak detection algorithm. After the filtered signal is transformed by the derivative and SSF functions, the smaller perturbations are significantly attenuated, making peak detection easier. The detected peak is shown as point 34 in the middle panel.

[0107] In step 23, the breathing interval BI is identified from the SSF signal.

[0108] The breathing interval is determined by fiducial detection. In particular, the breathing interval is measured from the SSF waveform (graph 32), while other inhalation / exhalation characteristics (e.g., when inhalation / exhalation starts / ends or inhalation / exhalation hold duration) are estimated from the filtered respiration signal (middle panel).

[0109] One SSF peak will correspond to one breathing interval. The algorithm finds the local maximum of the SSF peak in the above search window (i.e. from SSF[t-window] to SSF[t+window]). SSF[t] is the current value of the SSF signal and the window has a predefined duration such as 0.4 seconds, for example. Therefore, smaller peaks around a larger peak will not be identified as separate SSF peaks.

[0110] The SSF peak 34 is found from the SSF waveform as mentioned above. The SSF peak is defined as the local maximum value that exceeds the threshold in the moving window. For different users, the threshold can be automatically adapted to match the personalized signal amplitude. Otherwise, the signal caused by the subtle movement artifact of a fat person may have an amplitude comparable to the signal generated by the normal breathing of another thin person.

[0111] Once there is a newly detected SSF peak 34, the SSF peak is retrieved from the filtered signal ( Figure 3 The peaks and valleys are found in the bottom panel of the figure.

[0112] exist Figure 4The processing is shown in . The top panel shows the original respiration signal and the middle panel shows the low pass filtered signal. The bottom panel shows the derivative signal as graph 30 and the SSF signal as graph 32.

[0113] Figure 4 The bottom panel of shows that the peak position of the SSF signal is found, for example, at around 33 s. This is determined as the local maximum position in the search window using the search window as discussed above.

[0114] The search is then extended back in time for peaks and valleys in the filtered signal. As shown in the middle panel, point 40 is determined to be a peak and point 42 is determined to be a valley.

[0115] In step 24, the breathing characteristics f(BI) of those breathing intervals are obtained in the time domain from the original low-pass filtered signal.

[0116] For example, Figure 4 As shown in , the height difference between inspiration and expiration is shown as 44. It is the amplitude difference between a peak and an adjacent valley. A peak and an adjacent valley preferably belong to one breathing interval.

[0117] Other breathing properties may also be obtained.

[0118] For example, the inspiratory plateau threshold is determined as a portion of the height difference between inhalation and exhalation (e.g., 7.5%), and the expiratory plateau threshold is determined as a portion of the height difference between inhalation and exhalation (e.g., 7.5%). The inspiratory plateau begins at a position where the signal value is equal to the peak amplitude minus the threshold (i.e., 7.5% of the height difference between inhalation and exhalation), and this position is before the peak. The expiratory plateau begins (ends at inhalation) at a position where the signal value is equal to the peak amplitude minus the threshold (i.e., 7.5% of the height difference between inhalation and exhalation), and this position is after the peak. Similarly, the expiratory plateau begins at a position where the signal value is equal to the trough amplitude plus the threshold, and this position is before the trough. The inspiratory plateau begins (ends at exhalation) at a position where the signal value is equal to the trough amplitude plus the threshold, and this position is after the trough.

[0119] Figure 5 The exhalation start time is shown as point 50 and the inhalation start time is shown as point 52. The exhalation plateau is shown as the signal between points 54 and 52, while the inhalation plateau is shown as the signal between points 56 and 50.

[0120] Inspiratory duration is defined as the difference between the start time of an inspiration and the start time of the subsequent exhalation. Inspiratory hold duration is defined as the difference between the start time of an inspiration plateau and the start time of the subsequent exhalation. Similarly, expiratory duration is defined as the difference between the start time of an exhalation and the start time of the subsequent inspiration. Expiratory hold duration is defined as the difference between the start time of an exhalation plateau and the start time of the subsequent inspiration.

[0121] In summary, the inhalation plateau and the exhalation plateau are determined based on the peak and valley positions and the height difference between inhalation and exhalation derived from the peak and valley positions and the threshold value. In addition, the exhalation duration, the inhalation duration, the exhalation hold duration and the inhalation hold duration are derived based on the exhalation plateau and the inhalation plateau. Preferably, if the signal quality is not sufficient for a specific breathing interval, the inhalation duration, the exhalation start time, the exhalation hold duration and the inhalation start time will not be calculated at all.

[0122] Figure 6 The exhalation duration is shown as time period 60 and the inhalation duration is shown as time period 62. The exhalation hold time is shown as 64 and the inhalation hold time is shown as 66.

[0123] Thus, there are various (time domain) respiratory characteristics that can be determined, and they can be used in various ways to assess when the breathing interval is a suitable choice for determining the breathing rate. Additionally, the respiratory characteristics can be used to assess the predicted timing of the start of the next inspiration and / or the inspiration to expiration ratio.

[0124] Therefore, based on the breathing characteristics, a suitable breathing interval is selected in step 25, which can be used to subsequently determine the breathing rate. Suitable breathing intervals are those breathing intervals with suitable quality. In this way, in order to ensure calculation accuracy, the signal of each breathing interval is screened to check the signal quality.

[0125] By way of example, two rules can be used to check for waveform anomalies:

[0126] (i) Check whether the height difference between inhalation and exhalation is too small or too large;

[0127] (ii) Check that the inspiratory upstroke is not steadily increasing (in other words, the absence of a monotonically increasing inspiratory upstroke indicates an irregular breathing pattern).

[0128] Other respiratory characteristics that may be considered are:

[0129] (iii) a respiratory waveform having a mean within a predetermined mean range;

[0130] (iv) a respiratory waveform having one or more higher-order parameters (e.g., skewness or kurtosis) within a predetermined corresponding parameter range;

[0131] (v) Cosine similarity of consecutive respiratory waveforms within a predetermined similarity range.

[0132] Those rules can be considered together or individually. If the set rules for compliance are passed (e.g., the first two rules above), the breath interval will be marked as having suitable quality. Otherwise, the quality of the breath interval is marked as poor.

[0133] According to those appropriate breathing intervals, Figure 2 The candidate breathing interval is calculated in step 26 of . This is the artificial breathing interval obtained by combining the selected suitable breathing intervals (eg the average or median of a set of the most recent suitable breathing intervals).

[0134] The number of breath intervals involved must exceed a minimum required threshold within a predefined time. For example, there should be at least 5 breaths in the last 90 seconds. If a sufficient number of breaths are not available, the calculation will generate a message indicating that no breath could be found. The breath interval value has units of seconds by dividing the signal sample index by the sampling frequency. In other words:

[0135] Respiration interval (seconds) = signal sample index (data point) / sampling frequency (data point / second).

[0136] The final breathing interval, and thus the breathing rate, is obtained based on a smoothing function in step 27. In one example, the final breathing interval is based on smoothing the current candidate breathing interval with the previous final breathing interval:

[0137] FBI[n]=(FBI[n-1]+3BI[n]) / 4

[0138] n is the index of the breathing interval. BI[n] is the latest candidate breathing interval. FBI[n] is the final breathing interval, which is converted to 60 (seconds) divided by the final breathing rate of the final breathing interval. The units are breaths per minute:

[0139] Respiration rate[n] = 60 / FBI[n]

[0140] In step 28, a sleep apnea event and optionally its severity may be assessed based on the calculated respiration rate. For example, during a sleep apnea event, the respiration rate drops sharply to near zero within a few seconds. Thus, by detecting when there is a respiration rate below normal for a certain period of time, a sleep apnea event may be detected.

[0141] Other characteristics of the breathing signal may also be used as an indicator of a sleep apnea event, or to detect the severity of a sleep apnea event. For example, a breathing effort signal may also be used.

[0142] A known measure of the severity of sleep apnea events is the apnea-hypopnea index. It can also be combined with a measure of oxygen desaturation (using a pulse oximeter) to derive a sleep apnea severity score. The counting of apnea events and hypopnea events can be performed in a known manner.

[0143] Step 28 may also include other post-processing functions. The respiration rate is processed, for example, for visualization on a display screen, for storage in a memory, or for transmission to a remote device such as a smartphone via Bluetooth or Wi-Fi. The post-processing may implement customized policies, such as alerting the user or other staff if the respiration rate has been outside the normal range for a certain period of time.

[0144] The above processing is used to determine the respiratory rate and optionally other respiratory characteristics, in particular time domain characteristics of the respiratory signal. Examples of time domain characteristics include the inspiration / expiration ratio, where the inspiration duration is the elapsed time from one inspiration start point to its subsequent exhalation start point, and the exhalation duration is the remaining elapsed time when the inspiration duration is excluded from a respiratory cycle. The ratio can be estimated as the average inspiration duration divided by the average exhalation duration in the same time window as the respiratory rate.

[0145] Another option is to predict the timing of the next breathing interval, i.e. the next onset of inspiration. This exploits the estimate of the breathing interval that has been made based on the most recent breathing cycle. In addition, the prediction takes into account the phase delay of the filter used to process the raw data. The phase delay of an IIR filter is, for example, a parabolic function over the frequency range of interest, such as 0.1 Hz to 0.6 Hz in this case. By way of example, the IIR phase delay can be estimated using a regressed second-order polynomial function with a group phase delay that estimates the IIR phase delay based on the breathing rate determined from the most recent breathing interval (BI).

[0146] The upcoming inspiration start time is predicted to be:

[0147] t_PO=t_MIN+BI-IIR_PO

[0148] t_PO is the predicted start time.

[0149] t_MIN is the timing of the most recent respiratory signal minimum.

[0150] BI is the average length of previously selected breathing intervals (ie breathing intervals of suitable quality) within a predefined period.

[0151] IIR_PO is the IIR filter phase offset determined from the best fit polynomial function.

[0152] The IIR filter phase shift is used to compensate the low pass filter because the inhalation / exhalation characteristics are obtained from the filtered signal (not from the derivative or SSF signal).

[0153] Will refer to Figure 4 Explain an example.

[0154] The goal of the process is to predict where the next inspiration will be, specifically the point in time when the respiratory signal has reached its minimum (trough) and is about to increase. Figure 4 Point 42 in.

[0155] Minima can be detected by direct comparison with adjacent data, but this requires subsequent data samples after the minimum to be confirmed, giving a time delay, thus making prediction impossible. The pre-processing applied to the raw respiratory signal also introduces an additional time delay, namely the IIR phase delay discussed above.

[0156] The prediction is based on the assumption that the breathing interval duration remains nearly constant over a short duration (eg, approximately 10 to 20 seconds).

[0157] For example, in Figure 4 In FIG. 1 , an inspiratory trough has been detected at approximately 24.2 seconds. Note that this is the timing detected within the filtered signal, not the true timing, due to the time delay caused by filtering.

[0158] However, the actual timing may be desired, for example for controlling other devices.

[0159] For the prediction of this 31.8 second timing point, t_MIN is 24.2 seconds. Assuming the most recent breath interval calculation (based on the most recently selected set of breath intervals as discussed above) is 7.5 seconds, this would predict the next breath valley to be at approximately 24.2+7.5=31.7 seconds.

[0160] The filter delay is Figure 4 , as a time shift between the top and middle panels. Fitting of a curve for phase delays within the effective breathing frequency (0.1 Hz to 0.6 Hz, or 6 to 36 breaths per minute) enables the filter delay to be estimated. The phase delay is for example 0.3 seconds, and this is then correspondingly subtracted from the signal valley time (31.7 seconds). The next valley of the original signal is therefore predicted to occur at 31.7-0.3=31.4 seconds, which is consistent with Figure 4 The timing of the top panel is consistent.

[0161] In principle, under normal breathing conditions, the actual inspiratory start should be very close to the predicted value. When sleep apnea occurs, the breathing pattern has changed. Therefore, the predicted inspiratory start may be far from the actual occurrence, for example, more than 0.3 seconds.

[0162] The invention provides a method for determining a subject's breathing rate by analyzing a breathing signal of the subject. The method is preferably implemented in software controlling a processor to perform the signal analysis. The software comprises a computer program, i.e. computer program code means, which is executed by a processor forming part of an analysis system. The computer program is for example embodied as a computer program medium (i.e. a memory device).

[0163] 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 when practicing the claimed invention. In the claims, the word "comprising" does not exclude other elements or steps, and the word "a" or "an" does not exclude a plurality.

[0164] The functions implemented by a processor may be implemented by a single processor or by multiple separate processing units, which may be collectively considered to constitute a “processor.” Such processing units may be remote from each other in some cases and communicate with each other in a wired or wireless manner.

[0165] The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

[0166] The computer program may be stored / distributed on a suitable medium, such as an optical storage medium or solid-state medium supplied with or as part of other hardware, but the computer program may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunications systems.

[0167] If the term "suitable for" is used in the claims or the specification, it should be noted that the term "suitable for" is intended to be equivalent to the term "configured to". If the term "arrangement" is used in the claims or the specification, it should be noted that the term "arrangement" is intended to be equivalent to the term "system" and vice versa.

[0168] Any reference signs in the claims should not be construed as limiting the scope.

Claims

1. A system for determining a breathing rate of a subject, comprising a controller (14), the controller being configured to: receiving a breathing signal (BS) of the subject; (20) performing low-pass filtering on the respiratory signal to obtain a filtered respiratory signal; (21, 22) determining a transformed signal by at least obtaining a derivative signal of the filtered respiratory signal; (23) determining a plurality of breathing intervals from the transformed signal; (24) determining a respiratory characteristic in the filtered respiratory signal for each of the determined plurality of respiratory intervals; (25) selecting a breathing interval based on the breathing characteristics; as well as (27) Calculate the respiratory rate based on the selected breathing interval.

2. The system of claim 1, wherein the controller is further configured to: (28) Determining the occurrence and / or severity of sleep apnea based on the calculated breathing rate.

3. The system according to claim 1 or 2, wherein the controller is configured to determine the transformed signal by: applying a derivative filter to the filtered respiration signal and setting negative derivative filter values ​​to zero to obtain the derivative signal; and A slope sum filter is applied to the derivative signal.

4. The system according to any one of claims 1 to 3, wherein the controller is configured to: detecting peaks in the transformed signal; determining peak locations and trough locations in the filtered respiratory signal, the peak locations and the trough locations corresponding to the detected peaks in the transformed signal; and The breathing characteristic is determined based on the determined peak position and the valley position.

5. The system of claim 4, wherein the controller is configured to determine the height difference between inhalation and exhalation as a difference in amplitude between the determined peak location and an adjacent determined valley location in the filtered respiratory signal.

6. The system of claim 4 or 5, wherein the controller is configured to determine an inhalation duration and an exhalation duration based on the determined peak positions and trough positions in the filtered respiratory signal. 7 . The system of claim 6 , wherein the controller is configured to calculate an inspiration to expiration ratio based on the determined inspiration duration and the determined expiration duration.

8. The system according to any one of claims 1 to 7, wherein the controller is configured to derive a candidate breathing interval as an average or median of a plurality of most recently selected breathing intervals, and the controller is further configured to calculate the breathing rate based on the candidate breathing intervals.

9. The system of claim 8, wherein the controller is configured to derive a final breathing interval by smoothing a plurality of selected breathing intervals, and the controller is further configured to calculate the breathing rate based on the final breathing interval.

10. The system of any one of claims 1 to 9, wherein the controller is configured to select the plurality of breathing intervals having one or more of the following breathing characteristics: Altitude difference within a predetermined altitude range; Steadily increasing suction upstroke; a respiration signal having a mean within a predetermined mean range; a respiration signal having one or more high-order parameters within a predetermined corresponding parameter range; and / or Cosine similarity of consecutive breathing waveforms of the breathing signal within a predetermined similarity range.

11. The system of claim 10, wherein the controller is configured to predict the timing of the start of the next inspiration based on: a most recent trough position, a breath interval based on the length of a previously selected breath interval, and a phase delay caused by the low pass filtering.

12. The system of claim 11, wherein the controller is configured to determine the occurrence and / or severity of sleep apnea based on: a comparison between the predicted timing of the start of the next inhalation and the actual timing of the start of the next inhalation, and / or a calculated inspiration to exhalation ratio.

13. A respiratory monitoring system comprising: a sensor device for generating a breathing signal of a subject; as well as A system according to any one of claims 1 to 12, for determining respiratory rate.

14. A computer-implemented method for determining a respiration rate of a subject, comprising: receiving a breathing signal of the subject; (20) performing low-pass filtering on the respiratory signal to obtain a filtered respiratory signal; (21, 22) determining a transformed signal by at least obtaining a derivative signal of the filtered respiratory signal; (23) determining a plurality of breathing intervals from the transformed signal; (24) determining a respiratory characteristic in the filtered respiratory signal for each of the determined plurality of respiratory intervals; (25) selecting a breathing interval based on the breathing characteristics; as well as (27) Calculate the respiratory rate based on the selected breathing interval.

15. A computer program comprising computer program code means adapted to implement the method according to claim 14 when said program is run on a computer.

Images