Detecting sniffs and differentiating artifacts from electromyogram and accelerometer signals.

Abstract

A non-invasive system and method for quantifying respiratory muscle effort (RME) (or respiratory effort, work of breathing) is provided. The system and method utilize simultaneously measured EMG and accelerometer signals. The measured EMG signals are pre-processed to generate both a signal that emphasizes normal respiratory activity and a signal that emphasizes sniff activity (deep, sharp inspiration). Candidate sniff time intervals are determined from the pre-processed EMG signals. The measured accelerometer signals are pre-processed to generate a number of signals that emphasize activity in upper or lower frequency bands. Features of the pre-processed EMG and accelerometer signals corresponding to the candidate sniff time intervals are analyzed to determine whether the candidate sniff constitutes an actual sniff or an artifact. The maximum sniff effort is then identified, and the RME is then quantified by the ratio of the average of the maximum normal respiratory activity to the value of the maximum sniff effort.

Description

[Technical field]

[0001]

[0001] The disclosed concepts relate to methods and systems for quantifying a patient's respiratory effort (i.e., respiratory muscle effort, or "work of breathing"), and in particular to methods and systems for distinguishing electromyogram (EMG) signals indicative of sniff (i.e., deep, sharp inspiration) effort from EMG signal artifacts. [Background technology]

[0002] Electromyography (EMG) can be used to assess a patient's respiratory status by estimating the activity of muscles involved in breathing, such as the intercostal spaces on either side of the sternum (parasternal) or abdominal areas close to the diaphragm. Although respiratory rate can be easily measured non-invasively, respiratory rate alone does not indicate the effort a patient makes while breathing. In comparison, EMG measurements of the inspiratory muscles are an index of the balance between respiratory muscle load and respiratory muscle capacity. EMG signals provide a non-invasive method to obtain an objective measure of respiratory effort. In particular, respiratory EMG activity measured during inspiration represents the neural respiratory drive, which is the signal output by the brain to the respiratory muscles and is an index of the balance between respiratory muscle load and respiratory muscle capacity.

[0003]

[0003] Objective measurements of respiratory muscle activity derived from EMG signals are considered important for monitoring the respiratory status of patients, such as hospitalized patients with chronic obstructive pulmonary disease (COPD). In addition to EMG measurements of normal respiratory activity that occurs naturally when a patient is relaxed, EMG measurements of sniffs are also very useful in assessing a patient's respiratory status. A sniff is a deep, sharp inhalation (maximal effort movement). However, it is not easy to isolate sniff activity in the EMG signal, as signal artifacts due to non-respiratory activity give rise to strong activity in the EMG signal that resembles sniff activity. Such signal artifacts may be due to, for example, disturbance of the electrodes due to patient movement or pressure on the connected wiring.

[0004]

[0004] The ability to reliably detect sniffs and distinguish them from artifacts has a direct impact on the estimation of respiratory muscle activity from EMG signals and the estimation of the patient's respiratory state. Independent measurement of respiratory activity by devices such as nasal cannulas or esophageal electrodes provides additional respiratory data that can be compared with the EMG signal to distinguish sniffs from signal artifacts on the EMG waveform. However, nasal cannulas, esophageal electrodes and other similar devices are highly invasive and can be cumbersome or even burdensome for the patient. Summary of the Invention [Problem to be solved by the invention]

[0005]

[0005] Therefore, there is room for improvement in the methods and systems used to distinguish sniffs from other activity in EMG signals. [Means for solving the problem]

[0006]

[0006] It is therefore an object of the present invention, in one embodiment, to provide a method for quantifying a patient's respiratory muscle effort (i.e., respiratory activity, work of breathing) during a respiratory interval, the method comprising the steps of: measuring the patient's respiratory muscle activity using a plurality of EMG electrodes; measuring acceleration in multiple planes of the patient's chest simultaneously with the respiratory activity using an accelerometer; receiving raw EMG signals measured by the EMG electrodes and raw accelerometer signals measured by the accelerometer using a controller; generating a plurality of preprocessed EMG signals by preprocessing the raw EMG signals in the controller; generating a plurality of preprocessed accelerometer signals by preprocessing the raw accelerometer signals in the controller; receiving a plurality of preprocessed EMG signals by preprocessing the raw accelerometer signals in the controller; identifying a portion of the G signal as a candidate sniff; using a controller, determining a plurality of EMG-derived features from a plurality of pre-processed EMG signals associated with a time interval of the candidate sniff; using a controller, determining a plurality of accelerometer signal features from a plurality of pre-processed accelerometer signals associated with a time interval of the candidate sniff; using a controller, comparing the plurality of EMG-derived features and the accelerometer signal features to a plurality of sniff detection thresholds; using a controller, classifying the candidate sniff as a confirmed sniff if a result of the comparison indicates sniff activity; and using a controller, quantifying the patient's respiratory muscle effort by comparing a plurality of attributes of the plurality of pre-processed EMG signals to a plurality of attributes of the confirmed sniff.

[0007]

[0007] The method may further include using the controller to: identify a local normal breathing EMG maximum associated with normal breathing in the plurality of pre-processed EMG signals; determine an average of the local normal breathing EMG maximum; and identify a maximum sniff value in the plurality of pre-processed EMG signals associated with a confirmed sniff. Quantifying the respiratory muscle effort may include comparing the average of the local normal breathing EMG maximum to a maximum sniff value. The method may further include using the controller to: generate a normal breathing EMG signal by pre-processing the raw EMG signal to highlight normal breathing activity in the raw EMG signal and minimize artifacts; generate a sniff EMG signal by pre-processing the raw EMG signal to highlight sniff activity in the raw EMG signal; identify a local normal breathing EMG maximum in the normal breathing EMG signal; identify a local normal breathing EMG minimum in the normal breathing EMG signal; and identify a maximum sniff value in the sniff EMG signal. Quantifying the patient's respiratory muscle effort may comprise finding the ratio of the mean of the local EMG maxima to the maximum sniff value.

[0008]

[0008] The method may further include using the controller to: identify, for each candidate sniff, a bump in the candidate sniff such that all values ​​of the sniff EMG signal therein are equal to or greater than a predetermined threshold sniff value; identify a midpoint in each bump, the midpoint being the median such that the area under the curve of the left half of the bump is equal to the area under the curve of the right half of the bump; calculate an offset value by linearly interpolating between the local sniff EMG minimum immediately preceding the bump and the local sniff EMG minimum immediately following the bump; determine the amplitude of the bump by finding the difference between the maximum sniff EMG value within the bump and the offset value; and classifying the candidate sniff as an artifact if a plurality of predetermined amplitude conditions indicate artifact activity. The method may further include using the controller to determine a first asymmetric characteristic of the bump by finding a ratio of an average of a local sniff EMG minimum immediately preceding the bump and a local sniff EMG minimum immediately following the bump relative to the amplitude; determining a second asymmetric characteristic of the bump by finding a first difference between the local sniff EMG minimum immediately preceding the bump and the local sniff EMG minimum immediately following the bump, finding a second difference between a maximum sniff value within the bump and the local sniff EMG minimum immediately preceding the bump, finding a third difference between a maximum sniff value within the bump and the local sniff EMG minimum immediately following the bump, and finding a ratio of the first difference to the smaller of the second and third differences; determining a third asymmetric characteristic of the bump by determining a skewness of the bump; and classifying the candidate sniff as artifactual if a plurality of predetermined asymmetry conditions indicate artifactual activity.

[0009]

[0009] The method may further include using the controller: low pass filtering, rectifying and smoothing the raw accelerometer signal to generate a lower (low side) frequency band power signal; high pass filtering, rectifying and smoothing the raw accelerometer signal to generate an upper (high side) frequency band power signal; summing the lower frequency band power signal and the upper frequency band power signal to generate a summed frequency band power signal; determining a first ratio of the upper frequency band power signal to the summed frequency band power signal, on a first axis of the accelerometer, during a time interval associated with each of the candidate sniffs; determining a second ratio of the upper frequency band power signal to the summed frequency band power signal, on a second axis of the accelerometer, during a time interval associated with each of the candidate sniffs; comparing the first and second ratios to a predetermined frequency band ratio; and identifying as an artifact any candidate sniff whose first and second ratios exceed the predetermined frequency band ratio.

[0010]

[0010] The method may further include using the controller to: high pass filter, rectify and smooth the raw accelerometer signal to produce an upper high frequency band power signal; determine the number of times the upper high frequency band power signal crosses a predetermined threshold during a time interval associated with each of the candidate sniffs; and identify as an artifact any candidate sniff in which the upper frequency band power signal crosses the predetermined threshold more than a predetermined number of times during a time interval associated with the candidate sniff. The method may further include using the controller to: low pass filter the raw accelerometer signal to produce a lower frequency band signal; high pass filter the raw accelerometer signal to produce an upper frequency band signal; determine a standard deviation of the lower frequency band signal and the upper frequency band signal; and identify as an artifact any candidate sniff in which the standard deviation exceeds a predetermined value.

[0011]

[0011] In another embodiment, a system for quantifying a patient's respiratory effort during a breath interval includes: a plurality of EMG electrodes configured to measure the patient's respiratory muscle activity; an accelerometer configured to measure acceleration in multiple axes of the patient's chest; and a controller. The controller is configured to receive raw EMG signals measured by the EMG electrodes and raw accelerometer signals measured by the accelerometer; generate a plurality of preprocessed EMG signals by preprocessing the raw EMG signals; generate a plurality of preprocessed accelerometer signals by preprocessing the raw accelerometer signals; identify a portion of the plurality of preprocessed EMG signals as a candidate sniff; determine a plurality of EMG-derived features from the plurality of preprocessed EMG signals associated with a time interval of the candidate sniff; determine a plurality of accelerometer signal features from the plurality of preprocessed accelerometer signals associated with a time interval of the candidate sniff; compare the plurality of EMG-derived features and the accelerometer signal features to a plurality of sniff detection thresholds; classify the candidate sniff as a confirmed sniff or a signal artifact based on a comparison of the plurality of EMG-derived features and the accelerometer signal features to a plurality of sniff detection thresholds; and quantify the patient's respiratory muscle effort by comparing a plurality of attributes of the plurality of preprocessed EMG signals to a plurality of attributes of the confirmed sniff.

[0012]

[0012] The controller of the system may be further configured to: identify a local normal breathing EMG maximum associated with normal breathing in the plurality of preprocessed EMG signals; determine an average of the local normal breathing EMG maximum; identify a maximum sniff value in the plurality of preprocessed EMG signals associated with a confirmed sniff; and quantify respiratory muscle effort by comparing the average of the local normal breathing EMG maximum to the maximum sniff value. The controller of the system may be further configured to generate a normal breathing EMG signal by preprocessing the raw EMG signal to emphasize normal breathing activity in the EMG signal and minimize artifacts; generate a sniff EMG signal by preprocessing the raw EMG signal to emphasize sniff activity in the EMG signal; identify a local normal breathing EMG maximum in the normal breathing EMG signal; and identify a maximum sniff value in the sniff EMG signal. Comparing the average of the local EMG maximum to the maximum sniff value may include finding a ratio of the average of the local EMG maximum to the maximum sniff value.

[0013]

[0013] The controller of the system may further be configured to: identify, for each candidate sniff, a bump in the candidate sniff within which all values ​​of the sniff EMG signal are equal to or greater than a predetermined threshold sniff value; identify a midpoint in each bump, where the midpoint is the median such that the area under the curve of the left half of the bump is equal to the area under the curve of the right half of the bump; determine an offset value at the midpoint of each bump by linearly interpolating between the local sniff EMG minimum immediately preceding the bump and the local sniff EMG minimum immediately following the bump; determine the amplitude of the bump by finding the difference between the maximum sniff EMG value within the bump and the offset value; and classify the candidate sniff as artifactual if a plurality of predetermined amplitude conditions indicate artifactual activity.

[0014]

[0014] The controller of the system may be further configured to: high-pass filter, rectify and smooth the raw accelerometer signal to generate an upper high frequency band power signal; determine the number of times that the upper high frequency band power signal crosses a predetermined threshold during a time interval associated with each of the candidate sniffs; and identify as an artifact any candidate sniff in which the upper frequency band power signal crosses the predetermined threshold more than the predetermined number of times during a time interval associated with the candidate sniff. The controller of the system may be further configured to: low-pass filter the raw accelerometer signal to generate a lower frequency band signal; high-pass filter the raw accelerometer signal to generate an upper frequency band signal; determine a standard deviation of the lower frequency band signal and the upper frequency band signal; and identify as an artifact any candidate sniff in which the standard deviation exceeds a predetermined value.

[0015]

[0015] The above and other objects, features and characteristics of the present invention, as well as the method of operation and function of the associated components, combination of parts and economies of manufacture, will become more apparent from the following description and the appended claims, taken in conjunction with the accompanying drawings, all of which form a part of this specification, and like reference numerals indicate corresponding parts in the various views. It is to be expressly understood, however, that the drawings are for the purpose of illustration and description only, and are not intended to define the limits of the present invention. [Brief description of the drawings]

[0016] [Figure 1A] FIG. 1A is a perspective view of the front of a patient's torso with multiple EMG electrodes and accelerometers attached to the torso in accordance with an exemplary embodiment of the present invention, showing the x and y axes of the accelerometers defined relative to the torso. [Figure 1B]

[0017] FIG. 1B is a perspective view of the right side of the patient shown in FIG. 1A in accordance with an exemplary embodiment of the present invention, illustrating an xz plane defined relative to the torso such that the gravity vector lies approximately in the xz plane. [Diagram 2]

[0018] FIG. 2 shows two exemplary waveforms representing raw signals measured by EMG electrodes, such as the electrodes shown in FIG. 1, after undergoing two different forms of pre-processing, in accordance with an exemplary embodiment of the present invention. [Diagram 3]

[0019] FIG. 3 is a flow chart of a method 50 for quantifying respiratory muscle effort, according to an exemplary embodiment of the invention. [Figure 4]

[0020] FIG. 4 is a flow chart of a process 100 for defining intervals during a respiratory cycle c of a preprocessed EMG signal of normal breathing and a preprocessed EMG signal of a sniff where a candidate sniff may have occurred, according to an exemplary embodiment of the present invention. [Diagram 5]

[0021] FIG. 5 is a flowchart of a process 200 for defining certain characteristics of the main bump of a candidate sniff defined during process 100 using data points and time intervals based on the maximum value of the EMG signal determined in process 100, in accordance with an exemplary embodiment of the present invention. [Figure 6]

[0022] FIG. 6 is a flowchart of a process 300 for defining additional characteristics of a major bump of a candidate sniff defined during process 100 using data points and time intervals based on the EMG signal minimum determined in process 100, in accordance with an exemplary embodiment of the present invention. [Figure 7]

[0023] FIG. 7 shows a collection of example waveforms representing signals measured by EMG electrodes and accelerometers, such as those shown in FIGS. 1A and 1B, according to an example embodiment of the invention, the example waveforms having undergone various levels of pre-processing according to processes 100-600 shown in FIGS. 4-6 and 8-10. [Figure 8]

[0024] FIG. 8 is a flow chart of a process 400 for constructing a signal generally representative of the change in angle of the accelerometer shown in FIGS. 1A and 1B as a patient breathes, according to an exemplary embodiment of the invention. [Figure 9]

[0025] FIG. 9 is a flow chart of a process 500 for constructing upper and lower frequency band signals from raw accelerometer signals measured by the accelerometer referenced in FIG. 8 in accordance with an exemplary embodiment of the invention. [Figure 10]

[0026] FIG. 10 is a flowchart of a process 600 for defining characteristics of the raw accelerometer signal referenced in FIGS. 8 and 9 for classifying candidate sniffs identified during processes 100-300 as either sniffs or signal artifacts in accordance with an exemplary embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0017]

[0027] As used herein, the singular forms include plurals unless the context clearly dictates otherwise.

[0018]

[0028] As used herein, a statement that two or more parts or components are "coupled" shall mean that those parts are coupled or operate together directly or indirectly, i.e., through one or more intermediate parts or components to the extent that a link occurs.

[0019]

[0029] As used herein, the term "artifact" is intended to mean distortions on the electromyogram (EMG) signal waveform due to non-respiratory activity, including but not limited to activation of non-respiratory muscles (e.g., due to body movement, changes in electrode-skin impedance, or interference with cables connected to electrodes sensing the EMG signal).

[0020]

[0030] As used herein, the term "controller" shall mean a plurality of programmable analog and / or digital devices (including associated memory components or portions) that can store, retrieve, execute, and process data (e.g., software routines and / or information used by such software routines), including, without limitation, a field programmable gate array (FPGA), complex programmable logic device (CPLD), programmable system on chip (PSOC), application specific integrated circuit (ASIC), microprocessor, microcontroller, programmable logic controller, or any other suitable processing device or apparatus. The memory portions provide storage registers, i.e., non-transitory machine-readable media, for storage of data and program code, such as in the manner of a computer's internal storage area, and may be any one or more of various types of internal and / or external storage media, such as, but not limited to, RAM, ROM, EPROM, EEPROM, and FLASH, which may be volatile or non-volatile memory.

[0021]

[0031] As used herein, the term "machine learning model" refers to a software system that develops and builds mathematical models based on sample data, known as "training data," to make predictions or decisions without being explicitly programmed to do so; such software systems include, but are not limited to, computer software systems that develop algorithms that are trained to recognize patterns from a set of training data and then recognize patterns from the training data set within other data sets.

[0022]

[0032] As used herein, the term "number" is intended to mean one or an integer greater than one (ie, multiple).

[0023]

[0033] Directional terms used herein, such as, for example and without limitation, up, down, left, right, upper, lower, front, back, and derivatives thereof, refer to the orientation of the elements as illustrated in the drawings and do not limit the claims, unless expressly stated.

[0024]

[0034] The present invention provides a method and system for non-invasively and objectively quantifying a patient's respiratory effort, also referred to as respiratory muscle effort (RME), as described in detail herein in connection with various specific exemplary embodiments. As described in more detail herein, sniffs are important for accurate determination of RME, but sniffs and signal artifacts often appear similarly on the EMG signal waveform. Devices such as nasal cannulas or esophageal electrodes can be used to generate additional respiratory signals that can be compared to the EMG signal to distinguish sniffs from signal artifacts on the EMG waveform, but nasal cannulas can be uncomfortable for the patient, while esophageal electrodes are invasive. The method and system of the present invention improves upon the objective quantification of RME in two main ways: by utilizing an accelerometer signal in addition to the EMG signal to better distinguish sniffs from signal artifacts in the EMG signal waveform, and by using only non-invasive devices, i.e., EMG electrodes and accelerometers, to maximize patient comfort during collection of RME data.

[0025]

[0035] 1A and 1B show simplified schematic diagrams of a respiratory effort quantification system 1 consisting of a number of EMG electrodes 2, a two- or three-axis accelerometer 4, and a controller 6. According to an exemplary embodiment of the present invention, the EMG electrodes 2 are attached to a patient P to monitor the respiratory muscle activity of the patient P, while the accelerometer 4 is attached to the patient P to measure accelerations in multiple axes (i.e., x-axis, y-axis, and / or z-axis) of the body surface (i.e., chest) on which the accelerometer 4 is attached that occur simultaneously with the respiratory activity. In particular, the accelerometer 4 also measures accelerations due to artifacts that are completely unrelated to the respiratory activity. As will be explained in more detail herein with respect to the process 400 shown in FIG. 8, the acceleration signals measured by the accelerometer 4 are used to construct a signal that is roughly representative of the change in orientation of the body surface on which the accelerometer 4 is attached. The accelerometer 4 can therefore be considered to generate motion and orientation signals directly or indirectly.

[0026]

[0036] In an exemplary embodiment, the EMG electrode 2 is placed in the second intercostal space. The placement of the reference electrode can vary, and FIG. 1A shows two exemplary embodiments of reference electrode placement. In one exemplary embodiment, the reference electrode 2′ is placed on the clavicle, and in another exemplary embodiment, the reference electrode 2″ is placed on the sternum slightly above electrode 2. The accelerometer 4 is attached to the patient P's sternum such that the x and y axes lie in a plane tangent to the body surface, with the z axis pointing to the outside of the patient's body, i.e., perpendicular to the body surface, while the x axis points to the patient P's head. Thus, the xz plane is approximately perpendicular to the earth's surface, and the gravity vector lies approximately in the xz plane. The EMG electrodes 2 and the accelerometer 4 are shown in electrical communication with a controller 6, which is configured to receive and store the signals measured by the EMG electrodes 2 and the accelerometer 4.

[0027]

[0037] It will be appreciated that several types of controllers, e.g., various types of computers, can receive and store the signal information detected by devices such as the EMG electrodes 2 and the accelerometer 4. Thus, any type of controller 6 and any number (i.e., one or more) of controllers 6 can be used to receive and store the information transmitted by the EMG electrodes 2 and the accelerometer 4 without departing from the scope of the disclosed invention. Furthermore, in an exemplary embodiment, software including machine learning models 7 is integrated into the controller 6, as shown in FIG. 1, such that the methods and processes of the present invention can be automated.

[0028]

[0038] 2 shows an example of a first pre-processed EMG signal 10 and a second pre-processed EMG signal 11, each signal showing multiple respiratory cycles, each consisting of one significant oscillation of the EMG signal 10 or the EMG signal 11. The EMG signal 10 and the EMG signal 11 result from using two different levels of pre-processing on a raw EMG signal measured by an electrode, such as the EMG electrode 2 shown in FIG. 1A. Specifically, the EMG signal 10 is the result of heavily low-pass filtering the raw EMG signal (i.e., the EMG signal detected by multiple electrodes, such as the EMG electrode 2 in FIG. 1A), while the EMG signal 11 is the result of performing baseline removal and delay compensation on the EMG signal 10.

[0029]

[0039] The EMG signals 10 and 11 in FIG. 2 are shown pre-processed rather than in raw form to better highlight the different respiratory cycle attributes. The method and system of the present invention utilizes the pre-processed EMG signal to remove as many signal artifacts as possible while isolating and preserving normal breathing and sniff activity within the signal (the terms "normal breathing" and "sniff" are explained in more detail later in this specification). A high level of detail on such signal pre-processing is provided herein, but the present invention is not directed to the piecemeal details of such pre-processing. For simplicity of disclosure hereinafter, references to attributes of the EMG signal will refer to EMG signal 10, but it will be understood that any description of the temporal attributes of EMG signal 10 also applies to EMG signal 11 and vice versa, since both signals are derived from the same raw EMG signal and span the same time interval.

[0030]

[0040] The EMG signal 10 is divided into two parts, INS and OUT, as shown using solid and dashed lines in the waveform of Figure 2. The part of the EMG signal 10 that contains the maximum value of the main peak 12 of the respiratory cycle will be referred to as the INS part 14 of the EMG signal 10, and it will be understood that all parts of the EMG signal 10 drawn with the same dashed line as the labeled INS part 14 also represent the INS part 14 of the EMG signal 10. The part of the EMG signal 10 that is not the INS part 14 will be referred to as the OUT part 16, and it will be understood that all parts of the EMG signal 10 drawn with the same solid line as the labeled OUT part 16 also represent the OUT part 16 of the EMG signal 10. The INS part 14 of a given respiratory cycle c may also be referred to as I[c], and the OUT part of a given respiratory cycle c may also be referred to as O[c].

[0031]

[0041] In the method and system of the present invention, the exact start and end times of each INS portion 14 and OUT portion 16 need not be rigidly defined, since the defining feature of a given respiratory cycle is the inclusion of a respiratory cycle maximum in the INS portion. Although only one such maximum 12 of the main peak of a respiratory cycle in the EMG signal 10 is labeled in FIG. 2, it will be understood that all maxima occurring during the respiratory cycle shown in FIG. 2 are also main peak maxima 12. The time at which the main peak maxima 12 of a respiratory cycle occur is referred to herein as t rbg max 2, it will be appreciated that each of the OUT portions 16 includes a minimum 18 for each respiratory cycle. Although only one such minimum 18 for a respiratory cycle in the EMG signal 10 is labeled in FIG. 2, it will be appreciated that all local minima occurring during the respiratory cycle shown in FIG. 2 are also minima 18. The time at which a respiratory cycle minimum 18 occurs is referred to herein as t rbg min [c]. Typically, the INS portion 14 of the respiratory cycle in the EMG signal 10 corresponds to inspiration, since respiratory muscle activity is greatest during inspiration. However, it will be appreciated that in some patients, respiratory muscle activity is instead greatest during expiration, and in such patients the INS portion 14 of the respiratory cycle in the EMG signal corresponds to expiration.

[0032]

[0042] When a given respiratory cycle is referred to herein as a respiratory cycle c, it should be understood that the respiratory cycle immediately preceding cycle c is cycle [c-1] and the cycle immediately following cycle c is [c+1]. A respiratory cycle may be defined as either an algorithmic or a natural breathing cycle. An algorithmic respiratory cycle is denoted as ABC[c] and includes an OUT portion 16 of the respiratory cycle followed immediately by an INS portion 14 such that ABC[c]=(O[c],I[c]). A natural breathing cycle is denoted as NBC[c] and includes an INS of the respiratory cycle followed immediately by an OUT such that NBC[c]=(I[c],O[c+1]). An algorithmic respiratory cycle is so named because defining the respiratory cycle as (O[c],I[c]) rather than (I[c],O[c+1]) aids in artifact detection by real-time algorithms.

[0033]

[0043] Before proceeding with a detailed description of exemplary embodiments of the present invention, for ease of reference, several acronyms and variable names (including those variables mentioned above) and their meanings are listed below. Some of the entries in the list below have not yet been introduced, but the list may be referenced as these acronyms and variables are used throughout this disclosure.

[0034] [List of acronyms and variables] i.BC: Short for "respiratory cycle", this is part of one large oscillation that is expected to correspond either to a combination of successive inhalations and exhalations or to a large bump due to artifact in the EMG signal. ii. INS: indicates the part of the respiratory cycle that contains the maximum value of the main peak of the respiratory cycle. iii.OUT: indicates the part of the respiratory cycle that is not the INS part. iv.I[c]: indicates the INS portion of respiratory cycle c. vO[c]: indicates the OUT portion of respiratory cycle c. vi.c: Denotes an integer consecutive respiratory cycle index. vii. ABC[c]: represents the "algorithmic respiratory cycle" defined to make the artifact detection algorithm used in the present invention meaningful and to facilitate real-time processing, where each ABC[c] contains an OUT of a respiratory cycle followed immediately by an INS, such that ABC[c] = (O[c], I[c]). viii. NBC[c]: represents the “natural breathing cycle”, where each NBC[c] contains an INS of a breathing cycle followed immediately by an OUT, such that NBC[c]=(I[c], O[c+1]). ix.s[n]: represents an EMG signal preprocessed with normal respiratory parameters, EMG signals 10 and 11 shown in FIG. 2 represent two such EMG signals s[n] preprocessed with normal respiratory parameters. xv[n]: denotes the EMG signal preprocessed with sniff parameters. xi.t rbg max [c]: the time point associated with the maximum value of s[n] during I[c]. Referring to FIG. 2, point 12 represents a local maximum value of respiratory cycle c in s[n], and the time point associated with the labeled point 12 is the t rbg max [c]. xii.t rbg min [c]: The time point associated with the minimum of s[n] during O[c]. Referring to FIG. 2, point 18 represents a local minimum of respiratory cycle c in s[n], and the time point associated with the labeled point 18 is the t rbg min [c]. xiii.t snf min This is the point in time associated with the minimum value of v[n] between [c]:O[c]. xiv.t snf max [c]: The point in time associated with the maximum value of v[n] during I[c]. xv.E rbg abs: The average of the maxima of absolute EMG activity for respiratory cycles deemed artifact-free by artifact detection for normal breathing. xvi.E snf abs : shows the maximum respiratory muscle activity measured during sniffs after they were identified in the EMG signal using EMG signal features and simultaneously accelerometer signal features. xvii.E rel :E rbg abs E snf abs and constitutes a relative measure of respiratory muscle effort (RME).

[0035]

[0044] As mentioned above, the method and system of the present invention utilizes EMG signals that have been pre-processed to reduce as many signal artifacts as possible, which allows normal breathing and sniff activity to be separated from the raw EMG signal. "Normal" breathing is considered to occur when a patient is assumed to be relaxed and breathing naturally or spontaneously. In comparison, a "sniff" is much sharper, stronger, and shorter than a normal breath. It is noted that the inherent nature of a sniff (i.e., sharper, stronger, and shorter compared to the inspiration of a normal breath) and the relatively infrequent occurrence of a sniff compared to a normal breath do not immediately lend themselves to natural detection of a sniff by algorithms designed to detect respiratory cycles and artifacts during normal breathing activity.

[0036]

[0045] Thus, a set of normal breathing parameters designed to detect normal breathing activity is used to preprocess a given raw EMG signal to generate a normal breathing signal s[n], such as EMG signals 10 and 11 in Fig. 2 (such preprocessing may be referred to hereinafter as "normal breathing preprocessing"), and another set of sniff parameters designed to detect sniff activity and preserve key features of a sniff is used to preprocess a given EMG signal to generate a sniff signal v[n] (such preprocessing may be referred to hereinafter as "sniff preprocessing"). Thus, according to an exemplary embodiment of the present invention, for a given raw EMG signal, both a normal breathing preprocessed signal s[n] and a sniff preprocessed signal v[n] are generated, such that for each time point represented by a data point in signal s[n], there is a corresponding data point in signal v[n], and vice versa.

[0037]

[0046] The pre-processing of the EMG signal to generate the normal breathing signal s[n] and the sniff pre-processed signal v[n] follows some of the same steps, with the main difference between normal breathing pre-processing and sniff pre-processing being that sniff pre-processing includes significantly less net smoothing of the EMG signal. Both normal breathing pre-processing and sniff pre-processing include spike removal, scaling, a first round of high-pass filtering, optional power line interference reduction, rectification, downsampling, median filtering, and light low-pass filtering. Normal breathing processing further includes another round of low-pass filtering, another round of high-pass filtering for baseline removal, and construction of auxiliary signals for detection of respiratory cycle portions and artifacts.

[0038]

[0047] Spike removal removes very short large spikes in the raw EMG signal due to, for example but not limited to, pacemaker and sharp artifacts. Scaling simply refers to converting a signal measured in volts to microvolts (or other convenient units). A first round of high-pass filtering reduces low frequency motion artifacts, tonic activity, electrocardiogram (ECG) activity, power line interference, and sensor noise. If residual power line interference is present after the first round of high-pass filtering, a comb filter with notches at the fundamental power line frequency and its harmonics can be used to further reduce the power line interference. Rectification is the first operation to calculate the low frequency envelope of the high frequency EMG signal to construct a proxy respiratory signal. Rectification is followed by downsampling to reduce memory and processing requirements, since the frequencies present in the envelope signal are much lower than those in the raw signal. Median filtering is applied to further reduce residual spikes, especially due to ECG. Light low-pass filtering is used to further smooth the envelope signal and obtain a proxy respiratory muscle activity signal. The further smoothing by light low-pass filtering facilitates the calculation of minima and maxima in the pre-processed EMG signal.

[0039]

[0048] Referring now to FIG. 3, a flow chart illustrating a method 50 for quantifying RME is shown. It should be noted that the steps of method 50 provide a general overview of the steps involved in quantifying RME, and that most of the steps of method 50 are processes that include several steps in themselves. Accordingly, more specific details of each step of method 50 are provided herein with respect to processes 100, 200, 300, 400, 500 and 600 shown in FIGS. 4-6 and 8-9. In step 51 of method 50, respiratory muscle activity is measured by EMG electrodes 2 while acceleration of the chest in the x-, y- and / or z-axis concurrent with respiratory activity is measured by accelerometer 4 to generate raw EMG and accelerometer signals, as previously described with reference to FIGS. 1A and 1B. In step 52, the raw EMG signals measured by EMG electrodes 2 are pre-processed with normal respiratory parameters to generate a normal respiratory signal s[n] and with sniff parameters to generate a sniff signal v[n]. Throughout this disclosure, the portion of the preprocessed EMG signal s[n] or v[n] that represents a potential sniff may be referred to as a "candidate sniff." In step 53, various attributes of the normal breathing signal s[n] and the sniff signal v[n] are used to identify the time interval during which the candidate sniff occurred. Further details regarding steps 51-53 of method 50 are provided herein in relation to process 100 shown in FIG.

[0040]

[0049] Continuing with reference to Figure 3, in step 54 of method 50, the raw accelerometer signal measured by accelerometer 4 is pre-processed to generate a number of signals in different frequency bands. Further details regarding step 54 are provided herein with respect to process 500 shown in Figure 9. In step 55, attributes of the pre-processed EMG signals and pre-processed accelerometer signals found in steps 52-54 are used to confirm or reject the candidate sniff as an actual sniff by analyzing attributes of the pre-processed accelerometer signals that match the candidate sniff. Further details regarding step 55 of method 50 are provided herein with respect to processes 300 and 600 shown in Figures 6 and 10, respectively.

[0041]

[0050] Final step 56 of method 50 is the last step in quantifying RME and involves the normal breathing attribute E determined during process 100 (corresponding to steps 51-53 of method 50) shown in FIG. rbg abs , the sniff attribute E determined by performing steps 51 to 55 of method 50 snf abs 4-6 and 8-10, and using the corresponding processes 100-600 shown in FIG. rbg abs and E snf abs Using the value E rel It is necessary to determine the rbg abs represents the average maximum value of normal respiratory activity considered artifact-free, while E snf abs In step 56 of method 50, attribute E rel teeth,

number

[0042]

[0051] 4-6 and 8-10 are flow charts detailing the individual steps of various processes 100-600 used to extract normal breathing and sniff features from EMG signals using both EMG and accelerometer signals. Processes 100-600 are intended to be performed for every respiratory cycle c in a given signal waveform and have been developed using classical feature extraction approaches. For example, but not by way of limitation, where reference is made to finding a minimum or maximum of a signal waveform within a given respiratory cycle c, it should be noted that said respiratory cycle c and said minimum or maximum of the signal waveform within said respiratory cycle c are identified by a trained machine learning model 7 and manually checked for correctness. It should therefore also be noted that the EMG and accelerometer waveforms referred to in the detailed description of processes 100-600 may include either or both sniffs, commanded or spontaneously performed by the patient, so long as the physician is able to record the timing of such sniffs and use knowledge of said timing to verify the validity of processes 100-600.

[0043]

[0052] 4 is a flowchart detailing steps of a process 100 for defining intervals during a given respiratory cycle c of a normal-breath pre-processed EMG signal s[n] and a corresponding sniff pre-processed EMG signal v[n] during which sniff candidates may have occurred, according to an exemplary embodiment of the present invention. As previously mentioned, the time point associated with the maximum value of s[n] during I[c] is t rbg max [c], and the time associated with the minimum value of s[n] during O[c] is t rbg min [c]. In the process 100, at time t rbg max [c] is used as an initial seed point in each respiratory cycle c to identify some reference points in v[n]. rbg min [c] is similarly used as an initial seed point to identify additional reference points in v[n], as described later in this specification with respect to process 300 shown in FIG.

[0044]

[0053] In step 101 of process 100, at time t rbg max Time interval T centered on [c] rbg max [c] is

number

number

number

[0045]

[0054] In step 105, a threshold sniff value v[n] is calculated for each respiratory cycle c in the sniff signal v[n]. η [c]

number

number

number

number

number

[0046]

[0055] In step 110, the threshold sniff value v η [c], and at time t snf max The intervals around [c] are

number

[0047]

[0056] 5 is a flow chart detailing the steps of a process 200 for defining some characteristics of the main bump of a candidate sniff using the pre-processed EMG signals s[n] and v[n] and the data points and time intervals determined in process 100. In step 201, the duration D[c] of the bump interval Tη[c] defined in step 110 of process 100 is determined as

number

number

[0048]

[0057] In step 204, the bump interval T η To obtain a second measure of the asymmetry of the waveform in [c], t L η [c] and t R η [c] (i.e., bump interval T η The skewness of v[n] over [c] is determined. Again, since the sniff typically rises faster than it decays, the bump interval T η The skewness of [c] is expected to be positive (i.e., right-skewed). In an exemplary embodiment of the invention, the bump spacing T η A typical threshold for skewness over [c] is zero.

[0049]

[0058] Intuitively, properly executed sniffs cannot occur too close together. Therefore, T η A large bump in v[n] occurring just before or just after v[c] is very likely an artifact and may affect a candidate sniff. To detect such bumps, steps 205 and 206 L η Spacing T to the left of [c] LL η [c] (i.e., bump spacing T η [c] (the interval preceding the beginning of the Rη Interval T on the right side of [c] RR η [c] (i.e., bump spacing T η [c]) following the end of

number

number

[0050]

[0059] Both average values ​​μ LL η (v) [c] and μ RR η (v) [c] is the threshold v calculated in step 105 of process 100. η [c] or less, and μ LL η (v) [c] <v η [c] and μ RR η (v) [c] <v η[c]. If there are no undesirable bumps occurring near potential sniffs (bumps are "undesirable" because they complicate sniff isolation and identification), then the value μ LL η (v) [c] and μ RR η (v) [c] can be used as an additional number to quantify the asymmetry of a candidate sniff. Again, since sniffs are expected to rise faster than they decay, the following criterion is imposed:

number

[0051]

[0060] The processes 100 and 200 shown in Figures 4 and 5 use values ​​and timestamps associated with fiducial points of s[n] and v[n] derived from the maximum of the EMG signal. However, the minimum of the EMG signal is also very relevant for quantifying RME. Thus, Figure 6 is a flowchart detailing the steps of a process 300 for defining additional features of a candidate sniff based on the minimum of the pre-processed EMG signals s[n] and v[n], using the data points and time intervals determined in processes 100 and 200. In particular, the process 300 is used to determine EMG-derived features and attributes that can be used to distinguish between sniff and non-sniff activity (e.g., artifacts). As mentioned above, the time point associated with the minimum of s[n] during O[c] is t rbg min [c], and in the process 300, at time t rbg min [c] is used as an initial seed point in each respiratory cycle c to identify some reference points in v[n].

[0052]

[0061] In step 301 of the process 300, a time interval T in s[n] for a given respiratory cycle c is calculated. rbg min [c] is

number

number

[0053]

[0062] In step 305, the maximum value (t max [c],v maxIn order to quantify the fact that the average of the minima preceding and following the amplitude A[c] of the corresponding respiratory cycle is preferably low with respect to the amplitude A[c] of the corresponding respiratory cycle, a first asymmetry feature γ rel min is defined. Thus, γ rel min teeth,

number

number

number

[0054]

[0063] FIG. 7 shows a collection of multiple waveforms 71, 72, and 77 of EMG signals with various levels of pre-processing aligned with multiple waveforms 73, 74, 75, and 76 of accelerometer signals with various levels of pre-processing according to an exemplary embodiment of the disclosed concepts, the EMG signals and accelerometer signals being measured simultaneously as shown by the EMG electrodes 2 and accelerometer 4 in FIG. 1A and FIG. 1B. The waveforms shown in FIG. 7 are used for illustrative purposes only and are not intended to limit the scope of the invention. Waveforms 71-77 show raw or processed EMG or accelerometer signals from the same time interval, but each waveform emphasizes a different aspect of respiratory muscle activity or body surface orientation changes.

[0055]

[0064] Continuing with reference to FIG. 7, waveform 71 shows the raw EMG signal, and waveform 72 shows waveform 71 after it has been subjected to pre-processing high pass filtering. Comparing waveform 72 to waveform 71, it can be seen that pre-processing of the raw EMG signal enhances respiratory muscle activity while reducing signal artifacts. As described herein with respect to process 400 shown in FIG. 8, a pre-processed accelerometer signal is shown constructed from the raw accelerometer signal that roughly represents the change in angle of the accelerometer 4 over time, regardless of whether the change in angle is due to respiratory or non-respiratory activity. Waveform 74 is a pre-processed accelerometer signal (constructed from the raw accelerometer signal) that represents both the lower and upper frequency band signals of the x-axis of the accelerometer 4 (waveforms 74A and 74B, respectively, in FIG. 7), and waveform 75 is a pre-processed accelerometer signal (also constructed from the raw accelerometer signal) that represents both the lower and upper frequency band signals of the z-axis of the accelerometer 4 (waveforms 75A and 75B, respectively, in FIG. 7). The construction of the waveforms 74, 75 is described herein with reference to the process 600 shown in FIG.

[0056]

[0065] 7, waveform 76 represents the ratio of the upper and lower frequency band signals for both the x-channel and the z-channel of accelerometer 4. Specifically, as described herein with respect to process 500 shown in FIG 9, a rectified and smoothed lower frequency band signal acc_lpf_sm is found in step 504 of process 500 (acc_lpf_sm_x is found for the x-channel and acc_lpf_sm_z is found for the z-channel) and a rectified and smoothed upper frequency band signal acc_hpf_sm is found in step 505 of process 500 (acc_hpf_sm_x is found for the x-channel and acc_hpf_sm_z is found for the z-channel). 7, waveform 76A is the ratio (acc_hpf_sm_x) / (acc_lpf_sm_x+acc_hpf_sm_x) expressed in decibels, and waveform 76B is the ratio (acc_hpf_sm_z) / (acc_lpf_sm_z+acc_hpf_sm_z) expressed in decibels. Finally, waveform 77 represents a sniff processed EMG signal v[n] in which multiple sniffs 80 have been correctly identified after the steps of method 50 and processes 100-600 have been performed for each of the respiratory cycles identified in the pre-processed EMG.

[0057]

[0066] Returning briefly to Figures 1A and 1B, as previously mentioned, the accelerometer 4 is attached to the patient P's sternum with the z-axis pointing outside the patient P's body and perpendicular to the body surface, while the x- and y-axes lie in a plane tangent to the body surface with the x-axis pointing towards the patient P's head. In this way, the xz-plane is approximately perpendicular to the Earth's surface and the gravity vector lies approximately in the xz-plane. Now referring to Figure 8, a process 400 is used to construct an accelerometer signal that is roughly representative of the change in angle of the accelerometer 4 over time, such change in angle being due to some combination of respiratory and non-respiratory activity. In a first step 401, a fourth order Butterworth anti-aliasing low pass filter is applied to the raw accelerometer signal, with a cutoff frequency of 20 Hz being used in the exemplary embodiment. In step 402, the resulting signal is decimated. In an exemplary embodiment of the invention, the signal is decimated by a factor of 16. In step 403, a second order Butterworth low pass filter is applied to further reduce high frequency noise and artifacts in the accelerometer signal, the resulting signal being referred to below as acc_lpf_403. In the exemplary embodiment, a cutoff frequency of 10 Hz is applied.

[0058]

[0067] In step 404, a linear phase finite impulse response moving average filter is applied. In an exemplary embodiment of the invention, the linear phase finite impulse response moving average filter is applied with a support of 2 seconds. Step 404 generates a vector μ[n] with x and z components as the first and second elements of the vector, respectively. In step 405, the signal acc_lpf_403 generated in step 403 is delayed by the group delay of the filter to align with the vector μ[n] to generate a vector a[n]. In step 406, the tilt angle θ[n], which roughly represents the angle resulting from the change in orientation of the body surface (i.e., the xy plane shown in FIG. 1A) on which the accelerometer 4 is attached, is found by finding the y component of the vector product of μ[n] and a[n] (i.e., the sine of the angle between μ[n] and a[n]) based on the previously established assumption that the gravity vector lies approximately in the xz plane, and by multiplying the vector product by the vectors a[n] and a[n] using the following equation: - It is estimated by normalizing by the product of the magnitudes of [n].

number

[0059]

[0068] Referring now to FIG. 9, a process 500 is used to construct upper and lower frequency bands of the accelerometer signal to identify transient features of measured respiratory activity and movement artifacts in the raw accelerometer signal. It should be noted that the process 400 and the process 500 used for pre-processing the raw accelerometer signal are performed in parallel. In step 501 of the process 500, the baseline is removed from the accelerometer signal by applying a third order Butterworth high pass filter, in which case a cut-off frequency of 0.5 Hz is used in the exemplary embodiment of the disclosed concept. In step 502, a third order Butterworth low pass filter is applied to extract the lower frequency band signal acc_lpf_502, in which case a cut-off frequency of 50 Hz is used in the exemplary embodiment. Similarly, in step 503, a third order Butterworth high pass filter is applied to extract the upper frequency band signal acc_hpf_503, in which case a cut-off frequency of 150 Hz is used in the exemplary embodiment. In steps 504 and 505, the lower frequency band signal acc_lpf_502 found in step 502 and the upper frequency band signal acc_hpf_503 found in step 503 are rectified and smoothed using known signal processing techniques to extract the power line signal from the lower and upper frequency bands, respectively. The rectified and smoothed lower frequency band signal generated in step 504 is denoted as acc_lpf_sm (generated from acc_lpf_502), and the rectified and smoothed upper frequency band signal generated in step 505 is denoted as acc_hpf_sm (generated from acc_hpf_503). As shown in FIG. 9, steps 502 and 504 may be performed in parallel with steps 503 and 505.

[0060]

[0069] It will be appreciated that the steps of process 500 are performed on both the x-channel and z-channel of the accelerometer 4, such that pre-processing of the x-channel signal of the accelerometer 4 using process 500 produces rectified and smoothed lower and upper frequency band signals acc_lpf_sm_x and acc_hpf_sm_x, while pre-processing of the z-channel signal of the accelerometer 4 using process 500 produces rectified and smoothed signals acc_lpf_sm_z and acc_hpf_sm_z. Referring again to FIG. 7 in conjunction with FIG. 9, signal 74A is a waveform of the x-channel lower frequency band signal found in step 504, and signal 74B is a waveform of the x-channel upper frequency band signal found in step 505. Similarly, signal 75A is a waveform of the z-channel lower frequency band signal found in step 504, and signal 75B is a waveform of the z-channel upper frequency band signal found in step 505.

[0061]

[0070] 10, there is shown a flow chart illustrating a process 600 for defining respiratory features of accelerometer signals with respect to the various pre-processed signals found during processes 100-500. As will be understood, the results of process 500 are used in process 600, as will be described in more detail below, and thus process 600 is performed on the results of process 500 with respect to both the x-channel and the z-channel. All features found during process 600 are determined based on the bump spacing T η [c]. In step 601, the smoothed lower frequency band signal acc_lpf_sm generated in step 504 and the smoothed upper frequency band signal acc_hpf_sm generated in step 505 are summed. In step 602, the ratio of the smoothed upper frequency band signal generated in step 505 to the sum signal generated in step 601 is found.

[0062]

[0071] 7 in conjunction with FIG. 10, waveform 74A (the lower frequency band signal of the x-channel of accelerometer 4) and waveform 74B (the upper frequency band signal of the x-channel of accelerometer 4) are summed in step 601 for the x-channel. Similarly, waveform 75A (the lower frequency band signal of the z-channel of accelerometer 4) and waveform 75B (the upper frequency band signal of the z-channel of accelerometer 4) are summed in step 601 for the z-channel. Waveform 76A is obtained by performing step 602 for the x-channel, since waveform 76A is obtained by determining the ratio of waveform 74B to the sum of waveforms 74A and 74B found in step 601 for the x-channel. Similarly, waveform 76B is generated by performing step 602 for the z-channel, since waveform 76B is obtained by determining the ratio of waveform 75B to the sum of waveforms 75A and 75B found in step 601 for the z-channel.

[0063]

[0072] According to an exemplary embodiment of the present invention, for a sniff candidate to qualify as a sniff and not an artifact, the ratio calculated in step 602 relative to the sniff candidate's time interval must be less than a predetermined value. This requirement is based on the observation that spectrograms of accelerometer data show that the main power of a sniff usually occurs in the low frequency region, while the entire frequency band typically contains strong content during artifact activity. Thus, in step 603, if the ratio calculated in step 602 is indeed less than the predetermined value, the candidate sniff remains a candidate sniff.

[0064]

[0073] Experimental observations further indicate that if the upper frequency band signal acc_hpf_sm generated in step 505 crosses a predefined threshold value more than a predefined number of times during the time interval of a candidate sniff, the corresponding bump interval T η [c] usually indicates artifacts. The above threshold is set at 100% of the acceleration due to gravity (9.8m / s 2 In an exemplary embodiment, the threshold is 0.1 g (i.e., 0.98 m / s2 ), and the predetermined number of crossings is selected to be 7. Thus, in step 604, if the high frequency band signal acc_hpf_sm crosses the threshold (e.g., 0.1g) more than the predetermined number of crossings (e.g., 7), the candidate sniff is identified as an artifact, otherwise the candidate sniff maintains its status as a candidate sniff.

[0065]

[0074] Finally, in step 605, the variance or standard deviation of both the low-pass filtered and high-pass filtered signals acc_lpf_502 and acc_hpf_503 found in steps 502 and 503, respectively, during process 500 is determined. If either the variance or standard deviation found in step 605 exceeds a predetermined value, the candidate sniff activity is deemed to correspond to an artifact rather than a sniff. Conversely, if the variance or standard deviation found in step 605 is below a predetermined value, the candidate sniff is deemed to be a sniff rather than a signal artifact. Waveform 77 in FIG. 7 is an example of a sniff-processed EMG signal v[n] in which multiple sniffs 80 have been correctly identified after processes 100-600 have been performed. Referring again to method 50 shown in FIG. 3, E rel The final step 56 of quantifying RME by finding , is performed after steps 100-600 have been performed for a desired number of respiratory cycles recorded by the EMG electrodes 2 and accelerometer 4 and all sniffs in that respiratory cycle have been identified.

[0066]

[0075] In an exemplary embodiment of the invention, a true sniff is expected to qualify as a sniff under the assurance that all three criteria of steps 603, 604 and 605 are also met in addition to the criteria previously described for the EMG-derived features described with respect to processes 100-300. However, if a candidate sniff exhibits a tendency towards artifacts under one or two criteria of steps 603, 604 and 605, but not under all three, the predetermined values ​​and thresholds used throughout processes 100-600 may be adjusted without departing from the scope of the disclosed concepts. All so-called "thresholds" and "predetermined values" referred to herein are suggested values ​​and are not intended to limit the scope of the invention. In particular, if a method for pre-processing of EMG signals other than the methods previously described herein is used, it will be understood that at least some of the thresholds and predetermined values ​​proposed herein will likely need to be adjusted to account for differences in the magnitude and / or frequency of the pre-processed signal components as compared to those described herein, although the methods and systems described herein will still be applicable even with such adjustments to the thresholds and predetermined values.

[0067]

[0076] As previously described, the methods and processes 50 and 100-600 described herein were developed using classical engineering and feature extraction approaches. Previously, non-invasive sniff detection was difficult to automate, resulting in inaccurate distinctions between sniffs and signal artifacts in EMG signals due to the similarities exhibited by sniffs and signal artifacts in EMG signals. However, the present invention, and in particular the use of accelerometer signal data in addition to EMG signal data and features by processes 400-600, allows for the automation of accurate, non-invasive distinctions between sniffs and signal artifacts.

[0068]

[0077] In the claims, reference signs placed between parentheses shall not be interpreted as limiting the claim. The word "comprises" or "including" does not exclude the presence of elements or steps other than those listed in the claim. In a device claim enumerating several means, several of these means may be embodied by one and the same item of hardware. The word "a" or "an" does not exclude the presence of a plurality of such elements. In any device claim enumerating several means, several of these means may be embodied by one and the same item of hardware. The mere fact that certain elements are recited in mutually different dependent claims does not indicate that these elements cannot be used in combination.

[0069]

[0078] Although the present invention has been described in detail for purposes of illustration based on what are presently considered to be the most practical and preferred embodiments, it should be understood that such details are for illustrative purposes only and that the present invention is not limited to the disclosed embodiments, but on the contrary, is intended to cover modifications and equivalent arrangements within the spirit and scope of the appended claims. For example, it should be understood that the present invention contemplates that, to the extent possible, one or more features of any embodiment can be combined with one or more features of any other embodiment.

Claims

Claim 1 A method for quantifying a patient's respiratory effort during the respiratory interval, comprising: measuring the respiratory muscle activity of the patient using a plurality of electromyogram (EMG) electrodes; measuring the acceleration at a plurality of axes of the patient's chest using an accelerometer; receiving, using a controller, the raw EMG signal measured by the EMG electrodes and the raw accelerometer signal measured by the accelerometer; generating a plurality of preprocessed EMG signals by preprocessing the raw EMG signal using the controller; generating a plurality of preprocessed accelerometer signals by preprocessing the raw accelerometer signal using the controller; identifying, using the controller, a part of the plurality of preprocessed EMG signals as a candidate sniff; determining, using the controller, a plurality of EMG-derived features from the plurality of preprocessed EMG signals related to the time interval of the candidate sniff; determining, using the controller, a plurality of accelerometer signal features from the plurality of preprocessed accelerometer signals related to the time interval of the candidate sniff; comparing, using the controller, the plurality of EMG-derived features and accelerometer signal features with a plurality of sniff detection thresholds; classifying, using the controller, the candidate sniff as either a confirmed sniff or a signal artifact based on the comparison; quantifying, using the controller, the patient's respiratory muscle effort by comparing a plurality of attributes of the plurality of preprocessed EMG signals with a plurality of attributes of the confirmed sniff A method for quantifying a patient's respiratory effort, comprising the steps above. Claim 2 identifying, using the controller, a local normal respiration EMG maximum value related to normal respiration in the plurality of preprocessed EMG signals; determining, using the controller, an average of the local normal respiration EMG maximum values; identifying, using the controller, a maximum sniff value in the plurality of preprocessed EMG signals related to the confirmed sniff The method further comprises: wherein the step of quantifying the respiratory muscle effort comprises comparing the average of the local normal respiration EMG maximum values with the maximum sniff value. The method according to claim 1. Claim 3 Using the controller, generating a normal breathing EMG signal by preprocessing the raw EMG signal to emphasize normal breathing activities in the raw EMG signal and minimize artifacts; Using the controller, generating a sniff EMG signal by preprocessing the raw EMG signal to emphasize sniffing activities in the raw EMG signal; Using the controller, identifying a local normal breathing EMG maximum value in the normal breathing EMG signal; Using the controller, identifying a local normal breathing EMG minimum value in the normal breathing EMG signal; Using the controller, identifying a maximum sniff value in the sniff EMG signal; The method according to claim 1 or claim 2, further comprising:

4. The method according to claim 1, 2 or 3, wherein the step of quantifying the patient's respiratory muscle effort comprises finding a ratio of an average of local EMG maximum values to the maximum sniff value.

5. Using the controller, for each candidate sniff, identifying a bump of the candidate sniff such that all values of the sniff EMG signal therein are greater than or equal to a predetermined threshold sniff value; Using the controller, identifying a midpoint in each bump, the midpoint being a median value such that an area under a curve of a left half of the bump is equal to an area under a curve of a right half of the bump; Using the controller, calculating an offset value by linearly interpolating between a local sniff EMG minimum value immediately before the bump and a local sniff EMG minimum value immediately after the bump at the midpoint of each bump; Using the controller, determining an amplitude of the bump by finding a difference between a maximum sniff EMG value within the bump and the offset value; When a plurality of predetermined amplitude conditions indicate artifact activities, classifying the candidate sniff as an artifact; The method according to claim 3, further comprising:

6. Using the controller, determining a first asymmetry feature of the bump by finding a ratio of an average of a local sniff EMG minimum value immediately before the bump and a local sniff EMG minimum value immediately after the bump to the amplitude; Using the controller, finding a first difference between a local sniff EMG minimum value immediately before the bump and a local sniff EMG minimum value immediately after the bump, finding a second difference between a maximum sniff value within the bump and the local sniff EMG minimum value immediately before the bump, finding a third difference between the maximum sniff value within the bump and the local sniff EMG minimum value immediately after the bump, and determining a second asymmetry feature of the bump by finding a ratio of the first difference to the smaller of the second difference and the third difference; Using the controller, determining a third asymmetry feature of the bump by determining a skewness of the bump; Classifying the candidate sniff as an artifact if a plurality of predetermined asymmetry conditions indicate artifact activity; The method according to claim 5, further comprising: **Claim 7** Using the controller to low-pass filter, rectify, and smooth the raw accelerometer signal to generate a lower frequency band power signal; Using the controller to high-pass filter, rectify, and smooth the raw accelerometer signal to generate an upper frequency band power signal; Using the controller to sum the lower frequency band power signal and the upper frequency band power signal to generate a summed frequency band power signal; Determining a first ratio of the upper frequency band power signal to the summed frequency band power signal with respect to a first axis of the accelerometer during a time interval associated with each of the candidate sniffs; Determining a second ratio of the upper frequency band power signal to the summed frequency band power signal with respect to a second axis of the accelerometer during a time interval associated with each of the candidate sniffs; Comparing, by the controller, the first ratio and the second ratio to a predetermined frequency band ratio; Identifying as an artifact any of the candidate sniffs for which the first ratio and the second ratio exceed the predetermined frequency band ratio; The method according to any one of claims 1 to 6, further comprising: **Claim 8** Using the controller to high-pass filter, rectify, and smooth the raw accelerometer signal to generate an upper high frequency band power signal; Using the controller, determining the number of times the upper high-frequency band power signal crosses a predetermined threshold during a time interval associated with each of the candidate sniffs; Using the controller, identifying any of the candidate sniffs for which the number of times the upper frequency band power signal crosses the predetermined threshold during the time interval associated with the candidate sniff exceeds a predetermined number of crossings as an artifact; The method according to any one of claims 1 to 6, further comprising:

9. Using the controller, low-pass filtering the raw accelerometer signal to generate a lower frequency band signal; Using the controller, high-pass filtering the raw accelerometer signal to generate an upper frequency band signal; Using the controller, determining the standard deviation of the lower frequency band signal and the upper frequency band signal; Using the controller, identifying any of the candidate sniffs for which the standard deviation exceeds a predetermined value as an artifact; The method according to any one of claims 1 to 6, further comprising:

10. A system for quantifying a patient's respiratory effort during a respiratory interval, comprising: A plurality of electromyogram (EMG) electrodes for measuring the patient's respiratory muscle activity; An accelerometer for measuring the acceleration in a plurality of axes of the patient's chest; A controller And having The controller receives a raw EMG signal measured by the EMG electrodes and a raw accelerometer signal measured by the accelerometer; The controller generates a plurality of preprocessed EMG signals by preprocessing the raw EMG signal; The controller generates a plurality of preprocessed accelerometer signals by preprocessing the raw accelerometer signal; The controller identifies a part of the plurality of preprocessed EMG signals as candidate sniffs; The controller determines a plurality of EMG-derived features from the plurality of preprocessed EMG signals associated with the time interval of the candidate sniffs; The controller determines a plurality of accelerometer signal features from the plurality of preprocessed accelerometer signals associated with the time interval of the candidate sniffs; The controller compares the plurality of EMG-derived features and accelerometer signal features with a plurality of sniff detection thresholds; The controller classifies the candidate sniff as a confirmed sniff or a signal artifact based on a comparison of the candidate sniff with the plurality of sniff detection thresholds of the plurality of EMG-derived features and accelerometer signal features. The controller quantifies the respiratory muscle effort of the patient by comparing a plurality of attributes of the plurality of preprocessed EMG signals with a plurality of attributes of the confirmed sniff. System. **Claim 11** The controller further identifies local normal respiration EMG maxima related to normal respiration in the plurality of preprocessed EMG signals, determines an average of the local normal respiration EMG maxima, identifies a maximum sniff value in the plurality of preprocessed EMG signals related to the confirmed sniff, quantifies the respiratory muscle effort by comparing the average of the local normal respiration EMG maxima with the maximum sniff value. The system according to claim 10. **Claim 12** The controller further generates a normal respiration EMG signal by preprocessing the raw EMG signal to emphasize normal respiratory activity in the EMG signal and minimize artifacts; generates a sniff EMG signal by preprocessing the raw EMG signal to emphasize sniff activity in the EMG signal; identifies local normal respiration EMG maxima in the normal respiration EMG signal; and identifies a maximum sniff value in the sniff EMG signal; The comparison of the average of the local normal respiration EMG maxima with the maximum sniff value includes finding a ratio of the average of the local normal respiration EMG maxima to the maximum sniff value. The system according to claim 10 or 11. **Claim 13** The controller further for each candidate sniff, identifies a bump of the candidate sniff such that all values of the sniff EMG signal therein are greater than or equal to a predetermined threshold sniff value, identifies a midpoint in each bump, where the midpoint is a median value such that the area under the curve of the left half of the bump is equal to the area under the curve of the right half of the bump, determines an offset value by linearly interpolating between a local sniff EMG minimum immediately before the bump and a local sniff EMG minimum immediately after the bump at the midpoint of each bump, determines the amplitude of the bump by finding the difference between the maximum sniff EMG value within the bump and the offset value. If a plurality of predetermined amplitude conditions indicate artifact activity, classify the candidate sniff as an artifact. The system according to claim 12. **Claim 14** The controller further: Performs high-pass filtering, rectification, and smoothing on the raw accelerometer signal to generate an upper high-frequency band power signal. Determines the number of times the upper high-frequency band power signal crosses a predetermined threshold during a time interval associated with each of the candidate sniffs. Identifies as an artifact any of the candidate sniffs for which the number of times the upper frequency band power signal crosses the predetermined threshold during the time interval associated with the candidate sniff exceeds a predetermined number of crossings. The system according to any one of claims 10 to 13. **Claim 15** The controller further: Performs low-pass filtering on the raw accelerometer signal to generate a lower frequency band signal. Performs high-pass filtering on the raw accelerometer signal to generate an upper frequency band signal. Determines the standard deviation of the lower frequency band signal and the upper frequency band signal. Identifies as an artifact any of the candidate sniffs for which the standard deviation exceeds a predetermined value. The system according to any one of claims 10 to 13.

Images