Technique for extracting respiratory parameters from noisy, short-duration chest impedance measurements.

Abstract

One embodiment is a method of extracting respiratory parameters from a short-term thoracic impedance ("TI") signal, the method including: pre-processing a TI measurement signal to obtain a respiratory signal therefrom; evaluating the respiratory signal for at least one of signal quality and signal integrity; and performing at least one of an autocorrelation algorithm and a time-domain zero-crossing algorithm on the respiratory signal to extract at least one respiratory parameter therefrom, the at least one respiratory parameter including at least one of a respiration rate ("RR") and a tidal volume ("TV").

Description

Technical Field

[0001] Related Applications This disclosure claims priority to U.S. Provisional Patent Application No. 63 / 115,762, filed November 19, 2020, entitled "TECHNIQUES FOR EXTRACTING RESPIRATORY PARAMETERS FROM NOISY SHORT DURATION THORACID IMPEDANCE MEASUREMENTS", the disclosure of which is hereby incorporated by reference in its entirety.

[0002] This disclosure generally relates to techniques for detecting respiratory parameters from chest impedance measurements, and more particularly to techniques for extracting such parameters from noisy short-duration chest impedance measurements.

Summary of the Invention

Means for Solving the Problems

[0003] Embodiments of this disclosure provide a method for extracting a subject's respiratory parameters from a thoracic impedance (TI) measurement signal. The method includes performing a signal quality check on the TI measurement signal, and executing at least one of an autocorrelation algorithm and a time-domain zero-crossing algorithm on at least a portion of the TI measurement signal to extract at least one respiratory parameter of the subject from at least a portion of the TI measurement signal, wherein the at least one respiratory parameter includes at least one of respiratory rate ("RR") and tidal volume ("TV").

[0004] Another embodiment of this disclosure provides a method for determining a subject's respiratory rate (RR) from a thoracic impedance (TI) measurement signal. The method includes preprocessing the TI measurement signal to generate a respiratory signal, performing a signal quality check on the respiratory signal, and The estimated time-domain RR (TD_RR) is obtained by performing a time-domain zero-crossing algorithm on at least a portion of the respiratory signal, The autocorrelation algorithm is applied to at least a portion of the respiratory signals to obtain the estimated autocorrelation RR (AC_RR) and the confidence metric for the estimated AC_RR. Based on the confidence metric, select one of the estimated TD_RR and estimated AC_RR, This includes outputting one of the estimated TD_RR and estimated AC_RR as the final RR.

[0005] Another embodiment of the present disclosure provides a method for determining a subject's stroke volume (TV) from a chest impedance (TI) measurement signal. This method is Preprocessing the TI measurement signal to generate a respiratory signal, Perform a signal quality check on the respiratory signal, To determine the estimated TV by performing a time-domain zero-crossing algorithm on at least a portion of the respiratory signal, This includes selectively reporting the estimated TV based on the results of the signal quality check. [Brief explanation of the drawing]

[0006] For a more complete understanding of this disclosure and its features and advantages, please refer to the following description in conjunction with the attached drawings, where similar reference numbers represent similar parts.

[0007] [Figure 1] This disclosure describes exemplary environments for exemplary systems for deriving respiratory parameters from noisy, short-duration chest impedance measurements, according to several embodiments of this disclosure. [Figure 2] This block diagram shows exemplary functional components of the system in Figure 1 according to some embodiments of the present disclosure. [Figure 3A]This flowchart illustrates the operation of a method for extracting respiratory parameters from noisy, short-duration chest impedance measurements, according to several embodiments of the present disclosure. [Figure 3B] This flowchart illustrates the operation of a method for extracting respiratory rate from noisy, short-duration chest impedance measurements, according to several embodiments of the present disclosure. [Figure 4A] This disclosure describes several embodiments of the extraction of respiratory parameters from respiratory-modulated chest impedance signals using an autocorrelation-based algorithm. [Figure 4B] This disclosure describes several embodiments of the extraction of respiratory parameters from respiratory-modulated chest impedance signals using a time-domain-based (zero-crossing) algorithm. [Figure 5] This flowchart illustrates a method for performing impedance-limited signal quality checks according to some embodiments of the present disclosure. [Figure 6] This graph shows the effects of noise and motion artifacts on chest impedance signals according to some embodiments of the present disclosure. [Figure 7] This flowchart illustrates a method for performing artifact detection signal quality checks according to some embodiments of the present disclosure. [Figure 8A] This disclosure describes methods for removing artifacts detected from chest impedance signals according to several embodiments of this disclosure. [Figure 8B] This disclosure describes methods for removing artifacts detected from chest impedance signals according to several embodiments of this disclosure. [Figure 8C] This disclosure describes methods for removing artifacts detected from chest impedance signals according to several embodiments of this disclosure. [Figure 8D] This disclosure describes methods for removing artifacts detected from chest impedance signals according to several embodiments of this disclosure. [Figure 8E]A method for removing artifacts detected from a chest impedance signal according to some embodiments of the present disclosure is shown. [Figure 9] It is a flowchart showing a method for evaluating the signal quality of the longest preprocessed good segment of a chest impedance signal according to some embodiments of the present disclosure. [Figure 10A] It is a graph collectively showing the operation of an autocorrelation-based algorithm for extracting RR from a chest impedance signal according to some embodiments of the present disclosure. [Figure 10B] It is a graph collectively showing the operation of an autocorrelation-based algorithm for extracting RR from a chest impedance signal according to some embodiments of the present disclosure. [Figure 10C] It is a graph collectively showing the operation of an autocorrelation-based algorithm for extracting RR from a chest impedance signal according to some embodiments of the present disclosure. [Figure 11A] A flowchart collectively showing the operation of an autocorrelation-based algorithm according to some embodiments of the present disclosure is shown. [Figure 11B] A flowchart collectively showing the operation of an autocorrelation-based algorithm according to some embodiments of the present disclosure is shown. [Figure 12] It is a graph showing the physiological significance of a chest impedance signal and its derivative as alternatives to chest volume (liters) and flow rate (liters / minute) respectively. [Figure 13A] A flowchart collectively showing the operation of a time-domain based zero-crossing algorithm according to some embodiments of the present disclosure is shown. [Figure 13B] A flowchart collectively showing the operation of a time-domain based zero-crossing algorithm according to some embodiments of the present disclosure is shown. [Figure 14A] It includes a graph collectively showing the operation of the time-domain based zero-crossing algorithm shown in FIGS. 13A to 13B. [Figure 14B]It includes a graph that collectively shows the operations of the time-domain based zero-crossing algorithm shown in FIGS. 13A to 13B. [Figure 14C] It includes a graph that collectively shows the operations of the time-domain based zero-crossing algorithm shown in FIGS. 13A to 13B. [Figure 14D] It includes a graph that collectively shows the operations of the time-domain based zero-crossing algorithm shown in FIGS. 13A to 13B. [Figure 14E] It includes a graph that collectively shows the operations of the time-domain based zero-crossing algorithm shown in FIGS. 13A to 13B. [Figure 14F] It includes a graph that collectively shows the operations of the time-domain based zero-crossing algorithm shown in FIGS. 13A to 13B.

BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Chest impedance measurement values obtained using electrodes placed on a patient's chest provide an indirect and non-invasive method of collecting the desired respiratory parameters due to the fact that the modulation of the air level in the lungs by respiration is reflected in a proportional modulation of the chest electrical impedance. However, such measurement values are highly susceptible to the effects of very high levels of noise artifacts, for example, due to movement, coughing, and / or inappropriate skin electrode contact, making it difficult to extract parameters such as respiratory rate (RR) and tidal volume (TV) from the measurement values. In addition, certain clinical states require the extraction of events such as shallow breathing, apnea, and / or periodic or oscillatory breathing, which are even more difficult to extract, for example, in the presence of the aforementioned artifacts. The embodiments described herein include two approaches for addressing the aforementioned problems, including a time-domain based approach and an autocorrelation based approach. Both approaches closely follow the physiological aspects of the respiratory cycle, maintaining a minimum number of heuristic rules, and thus making it possible to extract most of these parameters from a single 60-second chest impedance measurement value with an error limited to within two breaths per minute (BPM) including quantization error.

[0009] Abnormal respiratory activity in a person is an early indicator of respiratory, cardiac, and / or neurological disorders. Clinically, respiratory rate (RR) is reported by counting the number of chest wall deviations during inhalation and exhalation. This method is often unreliable and depends on the nurse's skill level. The clinical method of extracting TV (volume of air inhaled and exhaled) involves breathing into a tube through the mouth with a nasal clip and is therefore not suitable for home monitoring.

[0010] As mentioned above, chest impedance monitoring via electrodes placed on the human chest can provide an indirect and non-invasive method for extracting respiratory parameters such as RR and TV. However, the accuracy of this technique is impaired by one or more of the following: the presence of very low-frequency baseline fluctuations due to improper electrode contact with the skin, high-frequency physiological interference factors such as cardiac activity, and broadband circuit noise and motion artifacts caused by coughing, hiccups, body movements, etc. Furthermore, certain physiological conditions exhibit various signatures in signal morphology, which makes it even more difficult to extract respiratory parameters with high reliability.

[0011] Conventional time-domain-based approaches, such as peak detection / counting, and frequency-domain-based approaches struggle to extract desired parameters from chest impedance measurement signals (or simply chest impedance signals) due to the non-stationary nature of the signal itself, as well as the noise embedded in the signal.

[0012] The embodiments described herein offer solutions to these problems and provide techniques for reliably extracting respiratory parameters from chest impedance signals by applying two different methodologies (time-domain based and autocorrelation-based) in the presence of different physiological signal forms and artifact conditions. A novel approach is also presented for evaluating the signal quality of chest impedance signals using input signals, accelerometer data, and filtered noise.

[0013] Time-domain-based approaches are useful for reporting RR in cases of low-confidence RR estimation (not signal quality) from autocorrelation-based techniques, as well as for estimating RR in cases of apnea and calculating TV.

[0014] Figure 1 shows an exemplary environment 100 of an exemplary embodiment of a system 102 for deriving and monitoring respiratory parameters such as RR and TV of a subject using noisy, short-duration chest impedance measurements, according to several embodiments of the present disclosure. Monitoring may be performed in a continuous or periodic manner. As shown in Figure 1, according to one exemplary embodiment, the system 102 includes a chest impedance measurement module 112 and a plurality of surface electrodes / sensors 114a-114d (e.g., four surface electrodes / sensors, or any other preferred number of surface electrodes / sensors). For example, one or more of the surface electrodes may be implemented as solid gel surface electrodes or any other preferred surface electrodes. The system 102 may be configured as a generally triangular device, or any other preferred shape, capable of operating to contact one or more of the torso, upper chest, and neck, or any other preferred part or region of the body of a subject 104 via at least a plurality of surface electrodes / sensors 114a-114d.

[0015] In various implementations, system 102 may have a configuration that allows it to be implemented as multiple patch-like devices within a wearable vest-like structure, or as any other suitable structure or device. In one possible environment, such as environment 100, system 102 may be capable of operating to engage in bidirectional communication with a smartphone 106 via a wireless communication path 116, and the smartphone 106 may further be capable of operating to engage in bidirectional communication with a communication network 108 (e.g., the Internet) via a wireless communication path 118. Alternatively, a direct link to the cloud 110 may be provided without requiring a hop via a base station or mobile phone. The smartphone 106 may further be capable of operating to engage in bidirectional communication with the cloud 110 via a wireless communication path 120 via the communication network 108, which may include resources for cloud computing, data processing, data analysis, data trend estimation, data reduction, data fusion, data storage, and other functions. System 102 may further be capable of operating to engage in bidirectional communication directly with the cloud 110 via a wireless communication path 122.

[0016] Figure 2 shows an exemplary block diagram of a system 102 for deriving and monitoring respiratory parameters such as RR and TV of a subject using noisy, short-duration chest impedance measurements, according to some embodiments of the present disclosure. As shown in Figure 2, the system includes a chest impedance measurement module 112, a processor 202 and associated memory 208, data storage 206 for storing chest impedance measurement data, and a transmitter / receiver 204. The transmitter / receiver 204 may be configured to perform Bluetooth communication, Wi-Fi communication, or any other preferred short-range communication for communicating with a smartphone 106 (Figure 1) via a wireless communication path 116. The transmitter / receiver 204 may be further configured to perform cellular communication or any other preferred long-range communication for communicating with a cloud 110 (Figure 1) via a wireless communication path 122. In certain embodiments, the chest impedance measurement module 112 may further include an electrode / sensor connection switching circuit 224 for switchable connection to a plurality of surface electrodes / sensors 114a-114d shown in Figure 1.

[0017] The processor 202 may include several processing modules, such as a data analyzer 226 and a data fusion / decision engine 228. The transmitter / receiver 204 may include at least one antenna 210 capable of transmitting and receiving radio signals, such as Bluetooth or Wi-Fi signals, via a wireless communication path 116 to and from a smartphone 106, which may be a Bluetooth or Wi-Fi enabled smartphone or any other suitable smartphone. The antenna 210 may further be capable of transmitting and receiving radio signals, such as cellular signals, to and from a cloud 110 via a wireless communication path 122.

[0018] The processor 202 may further include an autocorrelation module 230 and a time-domain module 232 for implementing autocorrelation-based and time-domain-based techniques, respectively, for deriving respiratory parameters from chest impedance signals, as described herein. The processor 202 may further include a signal quality evaluation module 234 for performing signal quality checks in relation to the chest impedance signals, as described herein.

[0019] The transmitter / receiver 204 may include at least one antenna 210 capable of transmitting and receiving radio signals, such as Bluetooth or Wi-Fi signals, via a wireless communication path 116 to and from a smartphone 106, which may be a Bluetooth or Wi-Fi enabled smartphone, or any other suitable smartphone. The antenna 210 may further be capable of transmitting and receiving radio signals, such as cellular signals, to and from a cloud 110 via a wireless communication path 122.

[0020] The operation of System 102 for deriving and monitoring respiratory parameters such as RR and TV using noisy, short-duration chest impedance measurements, according to several embodiments, will be further understood by referring to Figures 1 and 2 as well as the following examples. In these examples, while the subject 104 is in a supine or upright position, at fixed times each day (e.g., twice a day) or for a predetermined number of consecutive days, the subject or assistant positions System 102, configured as a generally triangular device (or a device of any other preferred shape), to contact one or more of the subject's torso and upper chest and neck (or any other preferred part or region of the body) via multiple surface electrodes / sensors 114a-114d.

[0021] After the system 102 is brought into contact with the subject's torso and / or upper chest and / or neck, the chest impedance measurement module 112 can be activated to collect, gather, sense, measure, or otherwise acquire chest impedance data from the subject 104 and generate a signal indicating it. In certain embodiments, the nature of the chest impedance data acquired using the exemplary method is that it is noisy and short in duration.

[0022] The chest impedance measurement module 112 can perform chest impedance measurements using some or all of the surface electrodes 114a-114d that come into contact with the skin of the torso, upper chest, and / or neck of the subject 104. According to the features of the embodiments described herein, respiratory parameters such as respiratory rate and tidal volume can be derived from noisy, short-term chest impedance data from the chest impedance measurement module 112, as will be described in more detail below.

[0023] In some embodiments, chest impedance data from the chest impedance measurement module 112 may be provided to the data analyzer 226 for at least partial data analysis, data trend estimation, and / or data reduction. In one embodiment, chest impedance measurement data may also be combined with other metadata such as medical history, demographic information, and other laboratory modalities, analyzed, trend estimated, and / or reduced "in the cloud," and made available in cloud-based data storage 110 with pre-configured alerts for use in various levels of clinical interventions regarding respiratory parameters.

[0024] The data analyzer 226 may provide the data fusion / decision engine 228 with at least partially analyzed chest impedance data, which can then be effectively fused or combined with other sensory data according to one or more algorithms and / or decision criteria for later use in making one or more inferences about the subject 104. The processor 202 may then provide the at least partially combined chest impedance and other sensory data to the transmitter / receiver 204, which can transmit the combined chest impedance and sensory data directly to the cloud 110 via the wireless communication path 122 or to the smartphone 106 via the wireless communication path 116. The smartphone 106 can then transmit the combined chest impedance and sensory data to the cloud 110 via the communication network 108 and wireless communication paths 118, 120, where it can be further analyzed, trend estimated, reduced, and / or fused. As described above, it will be recognized that the communication data may be communicated directly to the cloud 110 without the involvement of a smartphone / mobile phone or base station.

[0025] The resulting curated and combined sensing data can then be remotely downloaded by hospital clinicians for risk scoring / stratification, monitoring, and / or tracking purposes.

[0026] Figure 3A is a flowchart illustrating the operation of method 300 for extracting respiratory parameters from noisy, short-duration chest impedance measurements or signals, according to some embodiments of the present disclosure.

[0027] In step 302, a short-duration (e.g., 60-second) chest impedance signal obtained using electrodes (e.g., electrode 114 (Figure 1)) placed on the subject (e.g., subject 104 (Figure 1)) is preprocessed to obtain a respiratory signal therefrom. In certain embodiments, step 302 may be performed using a low-pass filter having a cutoff frequency (Fc) of 0.65 Hz.

[0028] In step 304, signal quality and signal integrity assessments are performed on the respiratory signal (e.g., by module 234 (Figure 2)). In certain embodiments, signal quality assessment may include calculating specific chest impedance limiting metrics for the signal, such as electrode contact impedance and whole-body impedance, which are compared to thresholds established based on physiological limits. Signal integrity assessment may include checking the signal signature to detect and remove major artifacts or interferences in the signal. Accelerometer data may be used to detect motion artifacts in the signal, and any motion artifacts can be removed in step 304.

[0029] As will be described in more detail below, certain embodiments of methods for extracting RR and TV from noisy, short-duration chest impedance signals, such as Method 300, take advantage of the fact that the chest impedance signal is a substitute for the subject's lung volume, and the derivative of the chest impedance signal is a substitute for the airflow velocity inside and outside the subject's lungs.

[0030] Generally, autocorrelation algorithms derive intrinsic periodic events in non-stationary signals with noise. Referring again to Figure 3A, in step 306, the breathing signal is processed using an autocorrelation-based technique (implemented, for example, by module 230 (Figure 2)). Specifically, in step 306, the breathing signal is autocorrelated to obtain the second mean of the breathing signal. Most autocorrelation-based algorithms for extracting periodicity consider only the dominant peak or maxima of the autocorrelated signal, which is prone to errors due to large high / low frequency noise. According to the features of the embodiments described herein, the autocorrelation-based technique implemented in step 306 utilizes the entire autocorrelated breathing signal to gain better insight into hidden periodicity and its fluctuations within the signal. Thus, in the illustrated embodiment, step 306 derives the estimated RR of the autocorrelation algorithm by calculating the expected value based on the time difference between peaks of the autocorrelated signal. This technique is well suited to breathing signals with anomalous morphology, periodic, oscillatory breathing patterns, and circuit noise.

[0031] In step 308, the respiratory signal is processed using time-domain-based techniques (implemented, for example, by module 232 (Figure 2)). In particular, as will be described in more detail below, in step 308, the respiratory signal is divided into inhalation and expiratory cycles by calculating zero crossings on the first derivative of the respiratory signal. Effective respiration is identified and invalid respiration is eliminated by applying heuristic rules based on physiological limitations such as invalid RR (e.g., more than 40 breaths per minute, less than 6 breaths per minute) and / or invalid inhalation relative to the expiratory ratio (e.g., 1:4 or 4:1). According to embodiments of the features described herein, the time-domain-based algorithm calculates RR by interval counting and calculates TV from the median of peak chest impedance values. The time-domain-based algorithm described herein is well suited to respiratory signals with frequency and amplitude modulated respiration and apnea.

[0032] In step 310, estimates from the autocorrelation-based and time-domain-based algorithms may be selected based on specific signal signatures representing a particular clinical condition. For example, in the case of apnea, where there is no breathing for several seconds, the number of inhalations and exhalations is best described by the time-domain-based algorithm, while the autocorrelation-based algorithm accurately identifies the number of breaths (i.e., RR) the subject is breathing before or after the apnea event. In contrast, in the case of oscillatory breathing, the RR is specified by the autocorrelation-based algorithm, and the oscillatory breathing in TV is specified by the time-domain-based "zero-crossing" algorithm.

[0033] In step 312, confidence assessments may be performed on the estimates from the autocorrelation-based algorithm and the time-domain algorithm, as described below.

[0034] In step 314, RR and TV estimates are selected and, if desired, reported and / or recorded.

[0035] For chest impedance signals (e.g., those with a significant amount of noise), frequency-domain methods such as autocorrelation are more useful in deriving the RR from the chest impedance signal. However, for other chest impedance signals (e.g., chest impedance signals that are not particularly periodic), time-domain-based methods are found to be more useful in deriving the RR from the chest impedance signal. Embodiments described herein leverage the relative advantages of both approaches by using both methods to derive the RR and then selecting the one that is likely to be more accurate in that environment.

[0036] Figure 3B is a flowchart illustrating the operation of method 320 for detecting RR from noisy, short-duration chest impedance measurements or signals, according to some embodiments of the present disclosure.

[0037] In step 322, a short-duration (e.g., 60-second) chest impedance signal obtained using electrodes (e.g., electrode 114 (Figure 1)) placed on the subject (e.g., subject 104 (Figure 1)) is preprocessed to obtain a respiratory signal therefrom. In certain embodiments, step 322 may be performed using a low-pass filter having a cutoff frequency (Fc) of 0.65 Hz. In some embodiments, the chest impedance signal may be filtered to a desired bandwidth of 0.1 Hz to 0.75 Hz.

[0038] In step 324, step 308 processes the respiratory signal using a time-domain-based technique to generate an estimated time-domain RR ("TD_RR"). In addition, if apnea is detected in the respiratory signal during the implementation of the time-domain-based technique, the FLAG_APNEA_DETECTED flag is set.

[0039] In step 326, the derivative of the respiratory signal is calculated, and in step 328, the respiratory signal and / or its derivative are processed using an autocorrelation-based technique to generate an estimated autocorrelation RR ("AC_RR") and a confidence metric for the estimated AC_RR. In a particular embodiment, the confidence metric (CM) is equal to the ratio of signal power (corresponding to BPM within ±5 bpm of the estimated AC_RR) to noise power (corresponding to breaths per minute (BPM) outside the ±5 bpm range of the estimated AC_RR).

[0040] Step 328 determines whether the quality of the chest impedance signal is good (as determined by one or more signal quality checks described below). If the quality of the chest impedance signal is not good, the execution proceeds to step 332, where it is determined that there is no RR to report because the signal is unreliable / unusable.

[0041] In step 328, if the quality of the chest impedance signal is determined to be good, the execution proceeds to step 334, where it is determined whether the confidence metric is less than a predetermined threshold (e.g., 1). If the confidence metric is determined to be less than the predetermined threshold in step 334, the execution proceeds to step 336, where the estimated TD_RR is output as RR. If the confidence metric is determined to be greater than or equal to the predetermined threshold in step 334, the execution proceeds to step 338, where the estimated AC_RR is output as RR.

[0042] In certain embodiments, the RR estimate (e.g., AC_RR or TD_RR) may be used to adjust the filtering used for TV extraction. For example, if the RR is found to be 10 bpm, the center frequency Fc and bandwidth of the low-pass filter can be selected to be 10 bpm + / - 3 bpm to improve TV extraction. Note that TV information is one of the determinants of the RR confidence metric (CM_RR) reporting. For example, very low TV (possibly due to poor contact) or very high TV (possibly due to contact impedance modulation) will both reduce the confidence in the RR reporting. Furthermore, combining RR and TV information may yield important clinical insights. For example, minute ventilation is defined as the amount of air breathed in one minute and is the product of RR and TV (e.g., typically 5-8 liters / min). Furthermore, although TV is not directly detected in liters, comparing the estimated TV to baseline readings may flag potential hypoventilation / hyperventilation if there is a significant decrease / increase in minute ventilation.

[0043] Figures 4A and 4B are graphs illustrating the extraction of respiratory parameters from an oscillatory respiratory signal using an autocorrelation-based algorithm (Figure 4A) and a graph illustrating the extraction of respiratory parameters from an oscillatory respiratory signal using a time-domain-based (zero-crossing) algorithm (Figure 4B), according to some embodiments of the present disclosure.

[0044] Figure 5 shows a flowchart of Method 500 for performing an impedance-limited signal quality check according to embodiments described herein (for example, as performed in step 304 (Figure 3A)). As shown in Figure 5, step 502 checks the chest impedance ("TI") signal before processing to determine whether it is within a valid impedance range (e.g., greater than 30 Ω and less than 250 Ω). If it is determined that the chest impedance signal is not within a valid impedance range, the execution proceeds to step 504, where an error code is generated indicating that the chest impedance signal is out of range and the signal quality is rated as -1 ("not reliable"). Furthermore, in step 504, the value of the parameter valid_RR (a flag set to indicate whether the reported RR is valid) is set to 0 (i.e., the reported RR is not valid). If it is determined in step 502 that the chest impedance signal is within a valid impedance range, the execution proceeds to step 506, where the value of valid_RR is set to 1 (i.e., the reported RR is valid).

[0045] Step 508 determines whether the settling deviation of the chest impedance signal is less than a specified percentage (e.g., 10%). As used herein, “settling deviation” refers to the change in chest impedance over the measurement duration. For example, if the chest impedance changes by more than 10%, the electrode contact is likely to be unstable. If it is determined that the settling deviation of the chest impedance signal is greater than or equal to the specified percentage, the execution proceeds to step 510, where an error code is generated indicating that the chest impedance settling deviation is too large. Furthermore, in step 510, the value of the parameter gSQM_valid_TV is set to 0 and the value of the parameter gSQM_valid_RR is set to 0. If it is determined in step 508 that the settling deviation of the chest impedance signal is less than the specified percentage, the execution proceeds to step 512, where the value of gSQM_valid_TV is set to 1. gSQM_valid_TV and gSQM_valid_RR are signal quality metrics for TV and RR, respectively, where a value of "1" indicates good signal quality and a value of "0" indicates poor signal quality.

[0046] In step 514, it is determined whether the contact impedance mismatch is less than a specific value (e.g., 2000 ohms). If it is determined that the contact impedance mismatch is greater than or equal to the specific value, execution proceeds to step 516, where an error code is generated indicating that the contact impedance mismatch is too high. Furthermore, in step 516, the value of gSQM_valid_RR is set to 0. If it is determined in step 514 that the contact impedance mismatch is less than the specific value, execution proceeds to step 518.

[0047] Step 518 determines whether the contact impedance is less than a specific value (e.g., 3000 ohms). If it is determined that the contact impedance is greater than or equal to the specific value, execution proceeds to step 520, where an error code is generated indicating that the contact impedance is too high. Furthermore, in step 520, the value of gSQM_valid_RR is set to 0. If it is determined in step 518 that the contact impedance is less than the specific value, execution proceeds to step 522.

[0048] In step 522, the impedance-limited signal quality check is deemed to have passed, and the value of gSQM_valid_RR is set to 1.

[0049] Referring to Figure 6, if the chest impedance signal 602 has a fault (e.g., an artifact) 600, it is recognized that the sample distribution 604 of the signal is likely to tail in one direction due to large / very small numbers. Simply put, the presence of an artifact increases the signal deviation from the mean. To evaluate this, we may consider the coefficient of variation (CoV), which increases as the noise of the chest impedance signal increases. The standard deviation (std) of the chest impedance signal would also reflect this effect, but it should be noted that defining an optimal threshold to achieve this is difficult. In contrast, CoV defines the ratio of noise to signal (std(signal) / mean(signal)). A CoV greater than 1 indicates that the sample distribution is hyperexponential, while a CoV less than 0.4 indicates that the sample distribution tails in the opposite direction.

[0050] Figure 7 shows a flowchart of method 700 for performing artifact detection signal quality check according to embodiments described herein (for example, as performed in step 304 (Figure 3A)). Referring to Figure 7, in step 702, the preprocessed chest impedance signal and corresponding accelerometer data are normalized to the range [0 to 1] to remove the effect of DC in the mean and the CoV of the chest impedance signal is calculated. Substantially simultaneously, in step 704, the preprocessed chest impedance signal and accelerometer data are normalized to zero mean and unit variance and the kurtosis is calculated.

[0051] In step 706, a determination is made for either the chest impedance signal or the accelerometer data whether (1) the CoV is greater than 1, or (2) the CoV is less than 0.4 and the kurtosis is greater than 7. If either of these conditions is true for either signal, an artifact is detected in step 708. If neither of the conditions is true for either signal in step 706, the process proceeds to step 710.

[0052] Step 710 determines whether gSQM_valid_RR=1, gSQM_valid_TV=1 (as determined in Method 500 (Figure 5)), and whether the signal length exceeds 30 seconds. If all of these conditions are true, the execution proceeds to step 712, where the signal is considered to have a high-quality data confidence (data_quality=1). If one or more of the conditions in step 710 are not true, the execution proceeds to step 714, where the signal is considered to have a low-quality data confidence (data_quality=0). As used herein, data_quality represents the final combined signal quality metric. For an artifact-free signal with sufficient duration (>30 seconds), it is the logical AND of gSQM_valid_TV and gSQM_valid_RR, and therefore it can be 1 or 0 depending on the SQM of TV and RR. If the signal has artifacts or is of insufficient length, it is set to -1 (indicating a poor / unusable signal).

[0053] Figures 8A–8E illustrate a method, according to embodiments described herein, for removing artifacts detected from a chest impedance signal to generate a signal from which RR and TV can be derived, according to embodiments described herein. Figure 8A shows the unprocessed chest impedance signal 800, including artifact 802. The unprocessed chest impedance signal 800 is normalized with respect to zero mean and unit variance. In addition, a Shannon energy envelope is calculated for the chest impedance signal and a threshold is applied. It will be recognized that Shannon energy provides better identification than mere signal energy and emphasizes mid-range artifacts compared to terminal artifacts. Figure 8B shows a waveform 810 representing the Shannon energy of the unprocessed chest impedance signal 800 and artifact 802.

[0054] As shown in Figure 8C, the mask 820 is unfolded from the waveform 810 (Figure 8B) and identifies a segment of the chest impedance signal containing artifact 802. Referring here to Figure 8D, the chest impedance signal is segmented into a bad segment 830 (containing artifact) and a good segment 832. The longest good segment, in the embodiment shown in Figure 8D, includes all of the good segment 832, is identified, preprocessed to create the longest preprocessed good segment, designated by reference no. 840 in Figure 8E. The longest preprocessed good segment 840 is then used to derive the RR and TV, as described herein. As shown in Figure 9, the signal quality of the longest preprocessed good segment 840 is evaluated.

[0055] Figure 9 shows a method 900 for evaluating the signal quality of the longest preprocessed good segment, such as segment 840 (Figure 8E). In step 902, the CoV and kurtosis of the segment are calculated. In step 904, it is determined whether the CoV is less than 1, or whether the CoV is greater than 0.4 and the kurtosis is less than 7. If a negative determination is made in step 904, the execution proceeds to step 906, where a non-quality confidence value is assigned to the segment and the segment's data_quality parameter is set to -1.

[0056] If a positive result is obtained in step 904, the execution proceeds to step 908, where it is determined whether gSQM_valid_RR is equal to 1, whether gSQM_valid_TV is equal to 1, and whether the signal length is less than 30 seconds. If all the conditions specified in step 908 are met, the execution proceeds to step 910, where a high-quality confidence value is assigned to the segment and the data_quality parameter is set to 1.

[0057] If one of the conditions specified in step 908 is not met, the execution proceeds to step 912, where it is determined whether gSQM_valid_RR is equal to 1, gSQM_valid_TV is equal to 1, and the signal length is less than 15 seconds. If all of the conditions specified in step 912 are met, the execution proceeds to step 914, where a low-quality confidence value is assigned to the segment and the data_quality parameter is set to 0.

[0058] If one of the conditions specified in step 912 is not met, execution proceeds to step 916, no quality confidence value is assigned to the segment, and the data_quality parameter is set to -1.

[0059] Depending on the details of a particular embodiment, the autocorrelation-based algorithm described herein derives the intrinsic periodicity of a respiratory signal (which does not need to be strictly periodic and / or static) without being affected by external noise. As described, extracting RR using the autocorrelation-based algorithm involves detreating the preprocessed signal to derive the static trend signal (zero mean), the autocorrelation of the signal, and the heuristic-based RR calculation from the autocorrelated signal. In addition, the signal-to-noise ratio (SNR) and TV can be calculated from the autocorrelated signal. Figures 10A–10C show the operation of the autocorrelation-based algorithm for extracting RR from a chest impedance signal 1000 (Figure 10A). As described in more detail below, the expected value of RR is calculated (Figure 10B), and a relative threshold is used to consider the peak as valid against noise (Figure 10C).

[0060] Figures 11A and 11B are flowcharts 1100 illustrating the operation of an autocorrelation-based module according to embodiments described herein. In step 1102, a 60-second chest impedance signal (e.g., a chest impedance signal segment) is input to the autocorrelation-based module. In step 1104, the input chest impedance signal segment is filtered by a low-pass filter to remove high-frequency noise. In particular, the low-pass filter may be a finite impulse response filter (FIR) having an Fc of 0.65 Hz and a tap count of length / 3.

[0061] In step 1106, the first derivative is performed, and (if the signal is not stationary) baseline fluctuations are removed to generate the difference signal (Δ amplitude / Δ time).

[0062] In step 1108, the correlation is calculated for the difference signal, which has its own time-delayed version (the delay of one sample), to generate the autocorrelated signal ((Δamplitude / Δtime)²).

[0063] In step 1110, all local maxima or peaks are identified within the autocorrelated signal.

[0064] In step 1112, a peak is discarded if its intensity is negatively correlated and its amplitude is less than 40% of the amplitude of an adjacent peak.

[0065] In step 1114, the relative amplitude and relative time difference between peaks are calculated to generate an array of relative time differences and relative amplitudes equal to the harmonic period, or a signal power.

[0066] In step 1116, the array of breaths per minute (BPM) is calculated using an array of relative time differences (e.g., 60 / Δtime / sampling rate).

[0067] In step 1118, a BPM value may be excluded from the array of BPMs calculated in step 1116 if (1) it is greater than 44 or less than 6, or (2) the difference between that value and an adjacent BPM value is 10 or greater. The result is an array of valid relative BPM values.

[0068] In step 1120, the average of the valid relative BPM values ​​is calculated and considered the estimated average RR.

[0069] In step 1122, the highest peak in the autocorrelated signal corresponding to the estimated mean RR is identified. This is the estimated dominant RR. The change in breath impedance is equal to the square root of the highest signal peak.

[0070] In step 1124, the RR of the time difference corresponding to the highest peak from the origin is calculated and considered the estimated dominant RR.

[0071] In step 1126, the acceptable deviation of the instantaneous BPM is calculated (e.g., estimated mean RR ± 5).

[0072] In step 1128, the SNR is calculated by summing up all the relative signal powers that enter both the inside (signal) and outside (noise) of the signal band.

[0073] The expected value of all relative time differences represents the RR, which is affected by harmonics of highly periodic sequences in the chest impedance signal, frequency increases / decreases between cycles, low-frequency artifacts, and non-uniform signal amplitudes (e.g., due to shallow breathing, apnea). Whenever the number of overlapping signals is limited, only correlated data is represented as peaks, and all uncorrelated data is canceled out. Because this algorithm does not depend entirely on signal amplitude, large artifacts have little effect. Relative thresholds, rather than global thresholds, are applied to identify valid peaks in the autocorrelated plot.

[0074] As described herein, a time-domain-based approach is also provided, which is useful for reporting RR in the case of low-confidence RR estimation (not signal quality) from autocorrelation-based techniques, estimating RR in the case of apnea, and calculating TV.

[0075] Figure 12 shows the physiological significance of the chest impedance signal and its derivative, as shown in Graph 1200, as a substitute for chest volume (liters), and as shown in Graph 1202, as a substitute for flow rate (liters / minute), respectively. Figures 13A-13B illustrate the operation of Method 1300 for implementing time-domain based zero-crossing according to the features of the embodiments described herein. The time-domain based algorithm is necessary, as described, to report RR based on time-domain counts in the case of unreliable RR estimates from autocorrelation-based algorithms, to estimate RR in the case of apnea, and to calculate TV.

[0076] Referring to Figure 13A, in step 1302, a short-duration chest impedance signal is input to the time-domain module (as shown in Figure 14A). In step 1304, the input signal is pre-processed by a low-pass filter (e.g., at 0.65 Hz) to produce the filtered signal shown in Figure 14B. In step 1306, the derivative of the pre-processed input signal is expanded (Figure 14C), and in step 1308, zero-crossings in the derivative signal are identified (Figure 14D). In step 1310, peaks and troughs are identified within the derivative signal. Heuristic rules applied to identify valid peaks may include excluding peaks that are below a minimum threshold for valid impedance peaks (e.g., 5% of the highest peak after artifact removal), excluding inhalation peaks that are less than 1.5 seconds (40 bpm) apart from adjacent inhalation peaks, and / or excluding peaks where the peak inhalation and peak expiratory values ​​change by 90% (e.g., peak inhalation is 40 milliohms (mohm) and peak expiratory is 400 mohm).

[0077] In step 1312, a shallow breathing threshold (described in more detail in Figure 13B) is applied. In step 1314, the median chest impedance value is calculated (Figure 14E) to generate a TV estimate. In step 1316, each effective peak is counted along with the trough (Figure 14F) to generate an estimated RR using the time-domain method ("RRt_estimate").

[0078] Figure 13B is a flowchart illustrating the application of the shallow breathing threshold method 1350 according to the embodiments described herein. As shown in Figure 13B, applying the shallow breathing threshold involves integrating one inhalation and exhalation cycle (step 1352), and then determining whether the peak value exceeds 20 mohm or 5% of the maximum peak value (step 1354). If the peak value is 20 mohm or less or 5% or less of the maximum peak value, the inhalation / exhalation cycle is excluded from the count (step 1356). If the peak value exceeds 20 mohm or 5% of the maximum peak value, the inhalation / exhalation cycle is included in the count (step 1358). These aforementioned steps are repeated for all inhalation / exhalation cycles (step 1360).

[0079] Example 1 provides a method for extracting respiratory parameters for a subject from a chest impedance (TI) measurement signal, the method comprising: performing a signal quality check on the TI measurement signal; and performing at least one of an autocorrelation algorithm and a time-domain zero-crossing algorithm on at least a portion of the TI measurement signal to extract at least one respiratory parameter of the subject from at least a portion of the TI measurement signal, the at least one respiratory parameter comprising at least one of respiratory rate ("RR") and tidal volume ("TV").

[0080] Example 2 provides the method of Example 1, further comprising low-pass filtering the TI measurement signal before implementation and execution.

[0081] Example 3 provides the method of Example 2, wherein the cutoff frequency of the filter used to perform low-pass filtering is 0.65 Hz.

[0082] Example 4 provides a method for signal quality checking that includes impedance-limited signal quality checking, as described in any of Examples 1 to 3.

[0083] Example 5 provides the method of Example 4, wherein the impedance-limited signal quality check includes checking at least one of electrode contact impedance and whole-body impedance by referring to a threshold based on physiological limits.

[0084] Example 6 provides a method according to any of Examples 1 to 5, wherein the signal quality check includes identifying at least one signal artifact in the TI measurement signal.

[0085] Example 7 provides the method of Example 6, further comprising removing at least one artifact from the TI measurement signal to generate at least a portion of the TI measurement signal.

[0086] Example 8 provides the method of Example 6, wherein at least one artifact is noise.

[0087] Example 9 provides the method of Example 6, wherein at least one artifact is a result of the subject's movement.

[0088] Example 10 provides a method according to any one of Examples 1 to 9, further comprising performing at least one of an autocorrelation algorithm and a time-domain zero-crossing algorithm on the TI measurement signal to autocorrelate the TI measurement signal to obtain the second mean of the TI measurement signal, and calculating an expected value based on the time difference between peaks of the autocorrelated TI measurement signal to derive an estimated respiratory rate ("RR").

[0089] Example 11 provides the method of Example 10, further comprising deriving the signal-to-noise ratio (SNR) of the TI measurement signal from the autocorrelated TI measurement signal.

[0090] Example 12 provides a method by any of Examples 10-11, further comprising calculating a confidence metric for the estimated RR.

[0091] Example 13 provides a method according to any of Examples 1 to 12, wherein performing at least one of an autocorrelation algorithm and a time-domain zero-crossing algorithm on the TI measurement signal includes counting zero-crossings on the first derivative of the TI measurement signal to divide the TI signal into inhalation and respiratory cycles to calculate the respiratory rate (RR), and calculating the stroke volume ("TV") from the median of the peak TI values.

[0092] Example 14 provides the method of Example 13, further comprising applying the shallow breathing threshold to the first derivative before calculating RR and TV.

[0093] Example 15 provides a method of any of Examples 1 to 14, further comprising selecting estimates generated by at least one of the autocorrelation algorithm and the time-domain zero-crossing algorithm based on a confidence metric associated with the autocorrelation algorithm.

[0094] Example 16 provides a method of any of Examples 1 to 15, further comprising selecting estimates generated by at least one of an autocorrelation algorithm and a time-domain zero-crossing algorithm based on a signal signature indicating a clinical condition.

[0095] Example 17 provides a method according to any of Examples 1 to 16, wherein the TI measurement signal has a duration of less than 60 seconds.

[0096] Example 18 provides a method according to any of Examples 1 to 17, wherein the TI measurement signal has a duration of less than 30 seconds.

[0097] Example 19 provides a method for determining a subject's respiratory rate (RR) from a chest impedance (TI) measurement signal, comprising: preprocessing the TI measurement signal to generate a respiratory signal; performing a signal quality check on the respiratory signal; performing a time-domain zero-crossing algorithm on at least a portion of the respiratory signal to obtain an estimated time-domain RR (TD_RR); performing an autocorrelation algorithm on at least a portion of the respiratory signal to obtain an estimated autocorrelation RR (AC_RR) and a confidence metric for the estimated AC_RR; selecting one of the estimated TD_RR and estimated AC_RR based on the confidence metric; and outputting the selected one of the estimated TD_RR and estimated AC_RR as the final RR.

[0098] Example 20 provides the method of Example 19, wherein the selection of one of the estimated TD_RR and estimated AC_RR based on a confidence metric includes selecting the estimated AC_RR if the confidence metric is above a threshold, and selecting the estimated TD_RR if the confidence metric is below a threshold.

[0099] Example 21 provides a method of any of Examples 19-20, further comprising refraining from outputting one of the estimated TD_RR and estimated AC_RR as the final RR if the signal quality check results are poor.

[0100] Example 22 provides a method according to any of Examples 19-21, wherein the preprocessing includes filtering the TI measurement signal using a low-pass filter.

[0101] Example 23 provides a method of any of Examples 19-22 in which the signal quality check includes an impedance-limited signal quality check.

[0102] Example 24 provides the method of Example 23, wherein the impedance-limited signal quality check includes checking at least one of electrode contact impedance and whole-body impedance by referring to a threshold based on physiological limits.

[0103] Example 25 provides a method according to any of Examples 19-24, wherein the signal quality check includes identifying at least one signal artifact in the respiratory signal.

[0104] Example 26 provides the method of Example 25, further comprising removing at least one artifact from the respiratory signal to generate at least a portion of the respiratory signal.

[0105] Example 27 provides a method of any of Examples 25-26, wherein at least one artifact includes noise.

[0106] Example 28 provides a method of any of Examples 25–27, wherein at least one artifact is a result of the subject's movement.

[0107] Example 29 provides a method according to any of Examples 19–28, further comprising: running an autocorrelation algorithm on at least a portion of the respiratory signal to obtain an autocorrelated signal; and deriving an estimated respiratory rate ("RR") by calculating an expected value based on the time difference between the peaks of the autocorrelated signal.

[0108] Example 30 provides the method according to Example 29, wherein the confidence metric is the ratio of signal power to noise power of the autocorrelated signal.

[0109] Example 31 provides a method according to any of Examples 19-30, which involves performing a time-domain zero-crossing algorithm on at least a portion of the respiratory signals, counting the number of zero-crossings for the first derivative signals of at least a portion of the respiratory signals, wherein the number of zero-crossings corresponds to the estimated TD_RR.

[0110] Example 32 provides the method of Example 31, further comprising performing a time-domain zero-crossing algorithm on at least a portion of the respiratory signals to flag apnea in relation to at least a portion of the respiratory signals.

[0111] Example 33 provides a method of any of Examples 31-32, further comprising performing a time-domain zero-crossing algorithm on at least a portion of the respiratory signals to flag shallow respiratory states in relation to at least a portion of the respiratory signals.

[0112] Example 34 provides a method for determining a subject's breath volume (TV) from a chest impedance (TI) measurement signal, comprising: preprocessing the TI measurement signal to generate a breath signal; performing a signal quality check on the breath signal; performing a time-domain zero-crossing algorithm on at least a portion of the breath signal to determine the estimated TV; and selectively reporting the estimated TV based on the results of the signal quality check.

[0113] Example 35 provides the method of Example 34, further comprising refraining from reporting the estimated TV if the result of the signal quality check is poor.

[0114] Example 36 provides the method according to any of Examples 34-35, wherein the preprocessing includes filtering the TI measurement signal using a low-pass filter.

[0115] Example 37 provides a method of any of Examples 34-36 in which the signal quality check includes an impedance-limited signal quality check.

[0116] Example 38 provides the method of Example 37, wherein the impedance-limited signal quality check includes checking at least one of electrode contact impedance and whole-body impedance by referring to a threshold based on physiological limits.

[0117] Example 39 provides a method according to any of Examples 34-39, wherein the signal quality check includes identifying at least one signal artifact in the respiratory signal.

[0118] Example 40 provides the method of Example 39, further comprising removing at least one artifact from the respiratory signal to generate at least a portion of the respiratory signal.

[0119] Example 41 provides a method of any of Examples 34–40, further comprising performing a time-domain zero-crossing algorithm on at least a portion of the respiratory signal to estimate TV from the median of peak TI values.

[0120] It should be noted that all specifications, dimensions, and relationships (e.g., the number of elements, actions, steps, etc.) outlined herein are provided for illustrative and teaching purposes only. Such information may vary significantly without departing from the spirit of this disclosure or the scope of the appended claims. This specification applies to only one non-limiting example, and therefore they should be interpreted as such. In the foregoing description, exemplary embodiments are described with reference to the arrangement of specific components. Various modifications and changes can be made to such embodiments without departing from the scope of the appended claims. Accordingly, the description and drawings should be taken as illustrative, not restrictive.

[0121] It should be noted that in many of the examples provided herein, interactions may be described in terms of two, three, four, or more electrical components. However, this is done for clarification and illustrative purposes only. It should be recognized that the system can be integrated in any preferred manner. In line with similar design choices, any of the parts, modules, and elements illustrated in the drawings can be combined in a variety of possible configurations, all of which are clearly within the broad scope of this specification. In certain cases, it may be easier to describe one or more functions of a given set of flows by referring to only a limited number of electrical elements. It should be understood that the electrical circuits in the drawings and their teachings are readily expandable and can accommodate a large number of components, as well as more complex / sophisticated arrangements and configurations. Therefore, the examples provided should not limit or impede the broad teaching of electrical circuits when potentially applied to a vast number of other architectures.

[0122] Furthermore, any references in this specification to various features (e.g., elements, structures, modules, components, steps, operations, characteristics, etc.) included in "one embodiment," "exemplary embodiment," "embodiment," "another embodiment," "several embodiments," "various embodiments," "other embodiments," or "alternative embodiments" are intended to mean that any such features are included in one or more embodiments of this disclosure, but may be combined in the same embodiment or not.

[0123] Furthermore, it should be noted that the functions related to the circuit architecture shown in the figures represent only a portion of the possible circuit architecture functions that can be performed by or within the system shown. Some of these operations may be deleted or removed as necessary, or they may be substantially modified or changed without departing from the scope of this disclosure. Moreover, the timing of these operations may be substantially changed. The aforementioned operation flows are provided for illustrative purposes and consideration. Substantial flexibility is provided by the embodiments described herein in that any preferred arrangement, time series, configuration, and timing mechanism may be provided without departing from the teachings of this disclosure.

[0124] Numerous other changes, substitutions, modifications, alterations, and modifications may be apparent to those skilled in the art, and this disclosure is intended to encompass all such changes, substitutions, modifications, alterations, and modifications as those found in the appended claims.

[0125] Any feature of the above-described devices and systems may be implemented in relation to the methods or processes described herein, and it should be noted that the details of the examples may be used in any of one or more embodiments. The “means to do” in these (above-described) examples include, but are not limited to, the use of any suitable components described herein, along with any suitable software, circuitry, hubs, computer code, logic, algorithms, hardware, controllers, interfaces, links, buses, communication paths, etc.

[0126] It should be noted that, using the examples provided above, as well as numerous other examples provided herein, interactions can be described in terms of two, three, or four network elements. However, this is done for clarification and merely illustrative purposes. In certain cases, it may be easier to describe one or more functions of a given set of flows by referring to only a limited number of network elements. It should be understood that the topologies illustrated and described with reference to the accompanying drawings (and their teachings) are readily extensible and can accommodate a large number of components, as well as more complex / advanced arrangements and configurations. Therefore, the examples provided should not limit or hinder the extensive teachings of the illustrated topologies when they are potentially applicable to a vast number of other architectures.

[0127] It is also important to note that the steps in the aforementioned flowchart represent only a portion of the possible signaling scenarios and patterns that may be performed by or within the communication system shown in the drawings. Some of these steps may be deleted or removed as appropriate, or they may be significantly modified or changed without departing from the scope of this disclosure. In addition, many of these operations are described as being performed simultaneously with or in parallel with one or more additional operations. However, the timing of these operations may be significantly altered. The aforementioned operation flow is provided for illustrative purposes and consideration. The communication system shown in the drawings offers substantial flexibility in that any preferred arrangement, sequence, configuration, and timing mechanism may be provided without departing from the teachings of this disclosure.

[0128] While this disclosure is described in detail with reference to specific configurations and arrangements, these exemplary configurations and arrangements can be substantially modified without departing from the scope of this disclosure. For example, while this disclosure is described with reference to specific communication exchanges, embodiments described herein may be applicable to other architectures.

[0129] Numerous other changes, substitutions, variations, alterations, and modifications may be apparent to those skilled in the art, and this disclosure is intended to encompass all such changes, substitutions, variations, alterations, and modifications as they fall within the scope of the attached claims. In order to help the United States Patent and Trademark Office (USPTO), and indeed any reader of any patent obtained in connection with this application, interpret the attached claims herein, the applicant wishes to note that (a) none of the attached claims are intended to exercise Section 142, paragraph 6 of the United States Patent Act as of the filing date of this application, unless the words “means for” or “steps for” are specifically used in a particular claim, and (b) nothing in the specification is intended to limit this disclosure in any way not specifically reflected in the attached claims.

Claims

1. A method for determining a subject's respiratory rate (RR) from a chest impedance (TI) measurement signal, wherein the method is: The TI measurement signal is preprocessed to generate a respiratory signal, The aforementioned respiratory signal will be subjected to a signal quality check, A time-domain zero-crossing algorithm is performed on at least a portion of the respiratory signal, the execution of which includes counting the number of zero-crossings with respect to the first derivative signal of the respiratory signal, dividing the respiratory signal into inhalation and exhalation cycles, calculating the RR, applying heuristic rules to identify effective breaths and eliminate ineffective breaths, the application of which includes identifying ineffective RRs and identifying ineffective ratios of inspiration to expiration, and the execution of the time-domain zero-crossing algorithm is performed to obtain an estimated time-domain RR (TD_RR). An autocorrelation algorithm is performed on at least a portion of the respiratory signals, the execution of which includes autocorrelation the respiratory signals to obtain an autocorrelated respiratory signal, and calculating an expected value based on the time difference between peaks in the autocorrelated respiratory signal, thereby obtaining an estimated autocorrelation RR (AC_RR) and a confidence metric for the estimated AC_RR. Based on the aforementioned confidence metric, one of the estimated TD_RR and the estimated AC_RR is selected. A method comprising outputting the selected one of the estimated TD_RR and the estimated AC_RR as the final RR.

2. Selecting one of the estimated TD_RR and the estimated AC_RR based on the confidence metric is If the confidence metric is above a threshold, select the estimated AC_RR, The method according to claim 1, comprising selecting the estimated TD_RR when the confidence metric is less than the threshold.

3. The method according to claim 1 or 2, further comprising refraining from outputting the selected one of the estimated TD_RR and the estimated AC_RR as the final RR if the result of the signal quality check is unsatisfactory.

4. The method according to any one of claims 1 to 3, wherein the preprocessing includes filtering the TI measurement signal using a low-pass filter.

5. The method according to any one of claims 1 to 4, wherein the signal quality check includes a signal quality check specifically for the impedance related to the respiratory signal.

6. The method according to claim 5, wherein the impedance-specific signal quality check includes checking whether at least one of the electrode contact impedance and whole-body impedance relating to the respiratory signal is within a threshold range.

7. The method according to any one of claims 1 to 6, wherein the signal quality check includes identifying at least one signal artifact in the respiratory signal.

8. The method according to claim 7, further comprising removing the at least one signal artifact from the breathing signal to generate the at least portion of the breathing signal.

9. The method according to claim 7 or 8, wherein the at least one signal artifact includes noise.

10. The method according to any one of claims 7 to 9, wherein the at least one signal artifact is a result of the subject's movement.

11. The method according to any one of claims 1 to 10, wherein the reliability metric is the ratio of the signal power to the noise power of the autocorrelated signal.

12. The method according to any one of claims 1 to 11, wherein performing a time-domain zero-crossing algorithm on at least a portion of the respiratory signal further comprises flagging an apnea state in relation to at least a portion of the respiratory signal.

13. The method according to any one of claims 1 to 12, wherein performing a time-domain zero-crossing algorithm on at least a portion of the respiratory signal further comprises flagging a shallow breathing state in relation to at least a portion of the respiratory signal.

Images