CA2522732A1 - Biomedical signal analysis techniques using wavelets - Google Patents
Biomedical signal analysis techniques using wavelets Download PDFInfo
- Publication number
- CA2522732A1 CA2522732A1 CA002522732A CA2522732A CA2522732A1 CA 2522732 A1 CA2522732 A1 CA 2522732A1 CA 002522732 A CA002522732 A CA 002522732A CA 2522732 A CA2522732 A CA 2522732A CA 2522732 A1 CA2522732 A1 CA 2522732A1
- Authority
- CA
- Canada
- Prior art keywords
- wavelets
- scale
- electrogram
- fine
- coarse
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Abandoned
Links
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7253—Details of waveform analysis characterised by using transforms
- A61B5/726—Details of waveform analysis characterised by using transforms using Wavelet transforms
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
- A61B5/346—Analysis of electrocardiograms
- A61B5/349—Detecting specific parameters of the electrocardiograph cycle
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
Abstract
The invention is directed to signal processing techniques for electrograms or other biomedical signals. In particular, the signal processing techniques make use of wavelet transformation of the electrograms. For example, an electrogram can be represented by a finite set of wavelets which comprise a decomposition of the electrogram in the scale-time domain. In accordance with the invention, wavelet analysis techniques can be used to distinguish specific phenomena in electrograms. For example, wavelet analysis can be used to distinguish between occurrences of large amplitude steep deflections, small amplitude steep deflections, and large amplitude shallow deflections of an electrogram.
Description
BIOMEDTCAL SIGNAL ANALYSIS TECHNIQUES USING WAVELETS
The invention relates to medical devices, and more particularly to signal processing of electrograms or other biomedical signals.
A wide variety of medical devices have been developed in order to monitor patient conditions and possibly deliver therapy to the patient. Iu some cases, the medical devices comprise external medical devices that are used to monitor a patient. hi other cases, the medical devices are implantable medical devices (IMDs) that are surgically implanted inside a patient for short or long term therapy. Telemetry can be used to communicate sensed information from one medical device to another medical device, e.g., from an IMD
to an external medical device.
One common example of an IMD is a pacemaker. A pacemaker typically includes one or more pacing and sensing leads for delivery of pacing pulses to a patient's heart.
Another example of an TMD is a combination pacemaker-cardioverter-defibrillator. Other examples include implantable brain stimulators, implantable gastric system stimulators, implantable nerve stimulators or muscle stimulators, implantable lower colon stimulators, implantable dnig or beneficial agent dispensers or pumps, implantable cardiac signal loops or other types of recorders or monitors, implantable gene therapy delivery devices, implantable incontinence prevention or monitoring devices, implantable insulin pumps or monitoring devices, and so on.
Medical devices, including IMDs and external medical devices, often sense and record electrograms of a patient. Electrograms refer to signals which represent recorded changes in electric potential of the patient. Examples of electxograms include electrocardiograms, i.e., recorded electrical potentials associated with a patient's heart, and electroencephalograms, i.e., recorded electrical potentials associated with a patient's brain.
Other more specific examples of electrograms include atrial electrograms, coronary sinus (CS) electrograms, esophageal electrograms, high right atrial (HRA) electrograms, His bundle electrograms, infra-atrial electrograms, intracardiac electrograms, right ventricular electrograms, right ventricular apical electrograms, sinus node electrograms, and the like.
The invention relates to medical devices, and more particularly to signal processing of electrograms or other biomedical signals.
A wide variety of medical devices have been developed in order to monitor patient conditions and possibly deliver therapy to the patient. Iu some cases, the medical devices comprise external medical devices that are used to monitor a patient. hi other cases, the medical devices are implantable medical devices (IMDs) that are surgically implanted inside a patient for short or long term therapy. Telemetry can be used to communicate sensed information from one medical device to another medical device, e.g., from an IMD
to an external medical device.
One common example of an IMD is a pacemaker. A pacemaker typically includes one or more pacing and sensing leads for delivery of pacing pulses to a patient's heart.
Another example of an TMD is a combination pacemaker-cardioverter-defibrillator. Other examples include implantable brain stimulators, implantable gastric system stimulators, implantable nerve stimulators or muscle stimulators, implantable lower colon stimulators, implantable dnig or beneficial agent dispensers or pumps, implantable cardiac signal loops or other types of recorders or monitors, implantable gene therapy delivery devices, implantable incontinence prevention or monitoring devices, implantable insulin pumps or monitoring devices, and so on.
Medical devices, including IMDs and external medical devices, often sense and record electrograms of a patient. Electrograms refer to signals which represent recorded changes in electric potential of the patient. Examples of electxograms include electrocardiograms, i.e., recorded electrical potentials associated with a patient's heart, and electroencephalograms, i.e., recorded electrical potentials associated with a patient's brain.
Other more specific examples of electrograms include atrial electrograms, coronary sinus (CS) electrograms, esophageal electrograms, high right atrial (HRA) electrograms, His bundle electrograms, infra-atrial electrograms, intracardiac electrograms, right ventricular electrograms, right ventricular apical electrograms, sinus node electrograms, and the like.
Signal processing of electrograms is a common challenge in the medical field.
In particular, it is often necessary to identify specific features of an electrogram so that medical events can be identified in the patient, such as specific depolari~ations in the patient's heart. For this reason, new and alternative techniques for processing electrograms are highly desirable. Signal processing of electrograms can be performed by any of a variety of medical devices including an IMI~ that senses the electrograms, an external medical device that senses the electrograms, or possibly an external medical device that receives sensed electrograms from another device, e.g., via telemetry.
In general, the invention is dixected to signal processing techniques for electrograms. In particular, the signal pxocessing techniques make use of wavelet transformation of the electrograms. For example, an electrogram can be represented by a finite set of wavelets which comprise a decomposition of the electrogram in the scale-time domain. hl accordance with the invention, wavelet analysis techniques can be used to distinguish specific phenomena in electrograms. For example, wavelet analysis can be used to distinguish between occurrences of large amplitude steep deflections, small amplitude steep deflections, and large amplitude shallow deflections of an electxogram.
In this manner, wavelet analysis can be useful in identification of R waves, P
waves and other electrogram events.
In one embodiment, the invention provides a method comprising converting an electrogram into a set of wavelets, the set of wavelets including wavelets at a fine scale and wavelets at a coarse scale, and identifying features of the electrogram based on whether the features appear in only the fine scale wavelets, only the coarse scale wavelets or both the fine and coarse scale wavelets.
In another embodiment, the invention provides a medical device comprising a wavelet transform unit to convert an electrogram into a set of wavelets, the set of wavelets including wavelets at a fine scale and wavelets at a coarse scale, and a wavelet analysis unit to identify features of the electrogram based on whether the features appear in only the fine scale wavelets, only the coarse scale wavelets or both the fine and coarse scale wavelets.
If implemented in software, the invention can be embodied in a computer readable medium comprising computer readable instructions that when executed convert an electrogram into a set of wavelets, the set of wavelets including wavelets at a fine scale and wavelets at a coarse scale, and identify features of the electrogram based on whether the features appear in only the fme scale wavelets, only the coarse scale wavelets or both the fine and coarse scale wavelets.
In another embodiment, the invention provides an apparatus comprising means for converting an electrogram into a set of wavelets, the set of wavelets including wavelets at a fine scale and wavelets at a coarse scale, and means for identifying features of the electrogram based on whether the features appear in only the fine scale wavelets, only the coarse scale wavelets or both the fine and coarse scale wavelets.
In another embodiment, the invention provides a method comprising converting an biomedical signal into a set of wavelets, the set of wavelets including wavelets at a fine scale and wavelets at a coarse scale, and identifrying features of the biomedical signal based on whether the features appeax in only the fine scale wavelets, only the coarse scale wavelets or both the fine and coarse scale wavelets.
The invention includes features that provide significant advances. For example, wavelet analysis can provide a relatively simple and straightforward mechanism for distinguishing among features or phenomena within electrograms. Accordingly, wavelet analysis can reduce the likelihood that the medical device will confuse the occurrence of large amplitude shallow deflections with the occurrence of large amplitude steep deflections. Similarly, implementation of the features of the present invention will enable the medical device to distinguish between the occurrence of large amplitude steep deflections and the occurrence of small amplitude steep deflections.
Moreover, the wavelet analysis techniques described herein can be well suited for implementation in IMDs, where computational resources are more limited and battery power is a concern. For example, the techniques described herein are much less computationally intensive than techniques in which correlation operations are performed to identify electrogram features.
The details of one or more embodiments of the invention are set forth in the accompanying drawings and the description below. C?ther features and aspects of the invention will be apparent from the description and drawings, and from the claims.
FIG. 1 is an exemplary block diagram of a medical device according to an embodiment of the invention.
FIGS. 2A-2I2 are graphs of an exemplary electrogram comprising an atrial electrocardiogram.
FIGS. 3-6 respectively illustrate traces of information contained in the wavelets of four different scales in relation to the electrogram illustrated in FIGS. 2A-2IO.
FIG 7 is another graph of an exemplary electrogram comprising an atrial electrocaxdiogram.
FIG. 8 is a graph illustrating scale maxima lines of various features of the electrogram of FIG 7 across various scales from coarse to fine.
FIG 9 is a flow diagxam according to an embodiment of the invention.
The invention is directed to signal processing techniques for electrograms,~or other biomedical signals, in which wavelet analysis is used. An electrogram can be represented by a finite set of wavelets which comprise a decomposition of the electrogram in the scale-time domain. In accordance with the invention, wavehet analysis techniques can be used to distinguish specific phenomena in electrograms or other biomedical signals.
In particular, wavelet analysis can be used to distinguish between occurrences of large amplitude steep deflections, small amplitude steep deflections, and large amplitude shallow deflections of an electrogram. Although many details of the invention are described in the context of ehectrograms, the techniques are equally useful for analysis of other types of biomedical signals.
Large amplitude steep deflections in an electrogram correspond to Iarge near field depolarizations, i.e., harge depolarizations traveling in close proximity to a sensing electrode. In cardiac settings, identification of large amplitude steep deflections identify primary cardiac depolarizations at the given location. Small amplitude steep deflections correspond to small neax field depolarizations, i.e., small depolarizations traveling in close proximity to the sensing electrode. In cardiac settings, identification of small amplitude steep deflections identify secondary cardiac depolarizations at the given location, e.g., commonly present during atrial fibrillation. Large amplitude shallow deflections in an electrogram correspond to primary far field depolarizations, i.e., large depolarizations not traveling in close proximity to the sensing electrode. In cardiac settings, identification of large amplitude shallow deflections identify a primary cardiac depolarization at a location that does not correspond to the sensing electrode, such as a ventricular depolarization sensed by an atrial electrode.
Distinguishing among large amplitude steep deflections, small amplitude steep deflections, and large amplitude shallow deflections of an electrogram can be challenging.
However, wavelet analysis provides a xelatively easy mechanism to do so. In particular, wavelets can be generated f~r an electrogram at various scales or scale factors. The scale factors span from a coarse scale to a fine scale. The coarse scale wavelets generally refer to wavelets having relatively large scale factors, whereas the fine scales generally refer to wavelets having relatively small scale factors. The ternzs coarse scale and fine scale, however, are generally relative and can assume different values in different embodiments.
Electrogram features that appear in alI scales of wavelets can be identified as large amplitude steep deflections. Electrogram feaW res that appear in fine scales but not coarse scales can be identified as small amplitude steep deflections. Electrogram features that appear in coarse scales but not fine scales can be identified as large amplitude shallow deflections. For example, the features can be said to appear in a given wavelet if a wavelet coefficient for that wavelet exceeds some predetermined threshold, and the threshold may be different for different scales.
In an electrogram, the differences between large amplitude steep deflections, small amplitude steep deflections, and large amplihide shallow deflections is generally relative, and can depend on the type of electrogram being analyzed. In other words, as used in this disclosure, the terms "large" and "small" are relative terms, and the terms "steep" and "shallow" are also relative terms. By way of example, amplitudes of atrial sensed atrial signals typically are in the range of 1-4 mV and have slopes exceeding 1 Volt per second.
Atrial (unipolar) sensed ventricular signals are generally approximately 0.1 -0.3 times the amplitude of atrial events, but this can vary considerably depending on the location of the atrial electrode, the indifferent electrode, and the size of the atrial electrode. The slope of atrial (unipolar) sensed ventricular signals is typically in the range of 0.1 Volts per second.
Despite the fact that the depolarization fields decrease with distance between the source and recording site, large far field depolarization wavefxonts, e.g., ventricular depolarization recorded by an atrial electrode, can contribute significantly to recorded amplitudes of electrograms. The spatial sensitivity of recording electrodes is wide for distant sources contributing to the recorded signal. However, the bandwidth related to this distance phenomena is relatively low.
In normal homogeneously coupled atrial tissue, simultaneously recorded tznipolar electrograms at various locations are typically biphasic with little morphological variation.
Comparison of electrograrn morphology recorded simultaneously from multiple neighboring locations can reveal infoz-mation on Local intercellular coupling either during atrial fibrillation or during normal sinus rhythm.
Spatial electrode sensitivity can be reduced by taking advantage of a priori information regarding the area of sensitivity related to recorded signal bandwidth.
Wavelet transforms are particularly useful in analysis of non-stationary signals because wavelet transforms provide an alternative to the classical Short Time Fourier Transform (STFT) and Gabor transform. The wavelet transform is typically a linear operation that decomposes a signal into components that appear at different scales (or resolutions). A
mother wavelet comprises a zero average function ~I' E L2(R) (finite energy):
~~I'(t)~lt = 0 (Admissibility condition) EQUATION 1 Equation 1 can be normalized ~~~I'~yl, and centered round t=0. Then, a family of wavelets can be obtained by scaling and translation of the mother wavelet ~I' by s, and translation by u:
'IJu,s(t)= ~~'(t su) EQUATIONZ
Wavelet analysis allows the use of coarse wavelets where more precise low-frequency information is needed, and fine wavelets where high-frequency information is required. In analogy to the STET, the wavelet transform is deftned as the sum over all time of the signal multiplied by scaled, shifted versions of the wavelet function. For functions f E LZ(R) the wavelet transform at time a and scale s is defined as:
Wf(u,s)=<.f,~'t,,s >= f.f(t)~I'*(t~u)dt EQUATION3 s This type of transform satisfies energy conservation. With decrease of scale 's', the support for the wavelet decreases and the wavelet becomes more sensitive to high-frequency components of the signal, enhancing finer grain details of the signal. An increase in scale, on the other hand, provides more emphasis on the coarse structure of the signal. The result of the wavelet transform can be defined in the scale-time plane. The wavelet transform can be rewritten as a convolution product:
Wf (u,s)= ~f(t)~I'*(t u)dt= f *~I'S(u), EQUATION4 s where ~S (t) = 1 'I" * (-t ) EQUATION 5 S
The Fourier transform of ~I'S(t) is:
~I'(~) _ ~~I'* (sue) EQUATION G
It appears that 'h is the transfer function of a band-pass filter, so the convolution can compute the wavelet transform with dilated impulse response band-pass filters.
Many electrograms, including electrocardiograms, carry most important information at their singularities and sharp deflections. The wavelet transform is particularly well adapted to characterize transient phenomena or singularities, because wavelet transforms decompose signals into building blocks well localized in time and frequency.
The wavelet transform can focus on localized signal structures with a zooming procedure that r progressively reduces the scale parameter 's.' A measure of local regularity of the signal is pxovided by the decay of the wavelet transform amplitude across its scales.
Singularities can be detected by following the wavelet transform local maxima at fine scales.
~ Wf (u,s) ~<_ As"+'~z EQUATION 7 From Equation 7, one can derive:
logZ ~ Wf (u, s) ~<_ logz A + (a + 1 / 2) logz s EQUATION 8 FIG. 1 is an exemplary block diagram of a medical device 10 according to an embodiment of the invention. Medical device 10 may comprise any of a wide variety of medical devices used to analyze electrograms. For example, medical device 10 may comprise an implanted medical device (IMD) that includes various implanted electrodes (not shown) used for sensing the electrograrns. Alternatively, medical device 10 lnay C0111pr1Se all eXterIlal medical device that uses surface electrodes on a patient's skin to sense the electrograms. Also, medical device can be an internal or external device that measure electrograms via subcutaneous electrodes. In other cases, medical device 10 may comprise an external device that receives sensed electrograms from another device, e.g., via telemetry. In any case, medical device 10 performs signal analysis on electrograms using wavelet analysis techniques as described herein.
In general, medical device 10 includes a wavelet transform unit I2 and a wavelet analysis unit 14. In the example illustrated in FIG. 1, wavelet transform unit 12 and a wavelet analysis unit 14 comprise software modules executed on a digital signal processor (DSP) 16. In that case, medical device 10 also includes memory 19 to store computer readable instructions that can be executed in DSP 16 to realize wavelet transform unit 12 and a wavelet analysis unit 14. For example, memory 19 can comprise random access memory (RAM), read-only memory (ROM), non-volatile random access memory (NVRA1VI), electrically erasable programmable read-only memory (EEPROM), flash memory, or the like.
Medical device 10 can also include an analog-to-digital (AlD) converter I8 to convert an analog electrogram to digital samples that comprise a digital electrogram. DSP
16 receives the digital electrogram and invokes wavelet transform unit I2 to transform the electrogram to wavelets and invokes wavelet analysis unit 14 to analyze the wavelets.
In other embodiments, however, wavelet transform unit 12 and a wavelet analysis unit 14 comprise dedicated hardware or logic that performs the functions described herein. Also, wavelet transform unit I2 and a wavelet analysis unit 14 can be implemented as one or more processors, application specific integrated circuits (ASICs), field progrannnable gate arrays (FPGAs), various combinations of hardware and software, or the like.
Moreover, in some embodiments wavelet transform unit 12 and a wavelet analysis unit 14 can be implemented as analog logic circuits. In that case, the need for DSP 16 and A/D converter 18 could be eliminated. However, the illustrated example of FIG. 1, in which wavelet transform unit 12 and wavelet analysis unit 14 comprise software modules executed on DSP 16, axe relatively easy and cost effective from an implementation standpoint. In that case, the software modules are initially stored in memory 19 and invoked by DSP 1 G to execute the techniques described herein.
Medical device 10 receives an analog electrogram (EGM) and A/D converter I 8 converts tl~e analog electrogram to a digital electrogram, i.e., a stream of digital samples that represent the electrogram. Again, medical device receives the electrogram from oxle or more sensing electrodes of medical device 10, or receives the electrogram from another device used to sense the electrograms.
Wavelet transform unit 12 receives an electrogram (in this case in digital form), axed performs wavelet transformation on the electrogram to generate a set of wavelets which collectively include the information in the electrogram. For example, wavelet transform unit 12 performs wavelet transformation using mathematical framework similax to that outlined above. In particular, the set of wavelets can be obtained by scaling and translation of a selected mother wavelet. Wavelet transform unit 12 can comprise a set of dilated impulse response band-pass filters designed to perform the desired wavelet transformation on the electrogram. The set of wavelets generated by wavelet transform unit 12 include numerous wavelets at various scale factors. The scale factors span from a coarse scale to a fine scale.
The coarse scale wavelets provide a larger overall picture of the electrograln, but lack specific details of the electrogram. The fine scale wavelets provide a less complete picture of the electrogram, but include more detail. The whole set of wavelets include wavelets of a number of different scale factors. The scale factors associated with coarse scale wavelets can be greater than or equal to 10 multiplied by scale factors associated with wavelets in the fme scale, although the invention is not necessarily limited in that respect.
Wavelet analysis unit 14 analyzes the generated wavelets to identify features of the electrogram based on the wavelets. In accordance with the invention, wavelet analysis unit 14 can distinguish between different electrogram features based on whether the features appear on coarse wavelets, fine waveiets, or both coarse and fine wavelets.
Wavelet analysis unit 14 outputs results based on this analysis. For example, the output of wavelet analysis unit 14 can comprise an indication of identified features and the timing of the identified features within the electrogram.
More specifically, wavelet analysis unit 14 identifies electrogram feaW res comprising large amplitude steep deflections when the features appear in all scales of wavelets. Also, wavelet analysis unit 14 identifies electrogram features comprising small amplitude steep deflections when the features appear in fine scales but not coarse scales.
Moreover, wavelet analysis unit 14 identifes electxogram features comprising large amplitude shallow deflections when the features appear in coarse scales but not fine scales. In this manner, wavelet analysis unit 14 analyzes the wavclets to identify characteristics of the electrogram.
FIG 2A is a graph of an exemplary clectrogram 20 comprising an atrial electrocardiogram, i.e., an electrogram ofthe heart sensed by an atrial electrode. The %-axis is in units of time (milliseconds) and tlae Y axis is in units of amplitude (mV1200).
Portion 22 is a large amplitude steep deflection in clectrogram 20, and corresponds to large near field depolarization, i.e., a large depolarization traveling in close proximity to the sensing electrode. Thus, identification of the large amplitude steep deflection (portion 22) identifies primary cardiac depolarizations at the given location. Wavelet transform 10 unit 14 identifies portion 22 as a large amplitude steep deflection because this electrogram feature appears in both coarse and fine wavelets.
FIG. 2B is another graph of electrogram 20 comprising an atrial electrocardiogram.
The X-axis is in units of time (milliseconds) and the Y axis is in units of amplitude (mV/200). As illustrated in FIG. 2B, portion 32 is a small amplitude steep deflection in electrogram 20, and corresponds to a small near field depolarization, i.e., small depolarizations traveling in close proximity to the sensing electrode.
Identification of small amplitude steep deflections (portion 32) identifies secondary cardiac depolaxizations at the given location, e.g., commonly present during atrial fibrillation.
Wavelet transform unit 14 identifies portion 32 as a small amplitude steep deflection because this electrogram feature appears in fme wavelets but does not appear in coarse wavelets.
FIG 2C is another graph of electrogram 20 comprising an atrial electrocardiogram.
The X-axis is in units of time (milliseconds) and the Y axis is in units of amplitude (mV/200). As illustrated in FIG 2C, portion 42 is a Large amplitude shallow deflection in electrogram 20, and corresponds to a primary far field depolarization, i.e., a large depolarization not traveling in close proximity to the sensing electrode.
Identification of large amplitude shallow deflections (portion 42) identifies primary cardiac depolarizations at a location that does not correspond to the sensing electrode. Wavelet transform unit 14 identifies portion 42 as a large amplitude shallow deflection because this electrogram feature appears in coarse wavelats but does not appear in fine wavelets.
Such wavelet analysis provides a relatively simple and straightforward mechanism fox distinguishing among features or phenomena within electrograms.
Accordingly, medical device 10 is less lilcely to confuse the occurrence of large amplitude shallow deflections with the occurrence of large amplitude steep deflections.
Similarly, medical device 10 is less likely to confuse the occurrence of haxge amplitude steep deflections with the occurrence of small amplitude steep deflections, and so forth. In this manner, signal processing of electrograms can be improved and simplified. In particular, the wavelet analysis techniques described herein are well suited fox implementation in IMDs, where camputationah resources are more limited and battery power is a concern. For example, the techniques described herein are much less computationally intensive than techniques in which correlation operations are performed to identify ehectrogram features.
FIG. 2D is another graph of electrogram 20 comprising an atrial electrocardiogram. FIGS. 3-6 respectively illustrate traces of information contained in the wavelets of four different scales. In each of FIGS. 3-6, the X-axis is in units of time (milliseconds) and the Y axis is in units of amplitude (mV/200).
In particular, FIG 3 illustrates a trace 35 associated with a scale factor of 4, FIG 4 illustrates a trace 45 associated with a scale factor of 2, FIG 5 illustrates a trace 5S
associated with a scale factor of 1 and FIG 6 illustrates a trace 65 associated with a scale factor of 0.4. The wavelets associated with trace 35 are fewer than the wavelets of the other traces because trace 35 represent the coarse wavelets in this example.
In other words, every wavelet used to define trace 35 spans a larger portion of electrogram 20 than wavelets of the other ttaces 45, 55, 65.
On the other hand, the wavelets associated with trace 65 are more numerous than the wavelets of the other traces because trace 65 represents the fine wavelets in this example. In other words, every wavehet used to define trace 65 spans a smaller portion of electrogram 20 than wavelets of the other traces 35, 45, 55. Each respective trace 35, 45, 55, 65 is a general representation of the information collectively contained in all of the wavelets at the respective scale.
As can be appreciated by examining FIGS. 3-6, portion 22 (FIG 2A) of electrogram 20 is identifiable in every trace 35, 45, 55, 65. In other words, the wavelets at every scale include information indicative of the large amplitude steep deflection associated with portion 22. Portion 32 (FIG 2B) of electrogram 20, however, is identifiable only in traces 5S, 6S and is generally not present in traces 3S, 45. W other words, only the wavelets at fine scales include information indicative of the sznahl amplitude steep deflection associated with portion 32.
Portion 42 (FIG. 2B) of electrogxam 20, is identifiable only in traces 3S, 4S
and SS, but is difficult to identify in trace 65. At even finer scales, portion 42 would become completely unidentifiable. Thus, only the wavelets at more coarse scales include information indicative of the large amplitude shallow deflection associated with portion S 42.
Wavelet analysis unit 14 exploits the fact that some features arc identifiable in all scales, some features are identifiable only in coarse scales, and some features are identifiable only in fine scales. In particular, wavelet analysis unit 14 identifies large amplitude steep deflections when the features appear in all scales of wavelets, identifies small amplitude steep deflections when the features appear in fine scales but not coarse scales, and identifies large amplitude shallow deflections when the features appear in coarse scales but not fine scales. As used in this disclosure, the term "appear" generally means that the given feature can be identified in the respective scale. For example, if a given wavelet coefficient exceeds a particular predefined value, the feature can be 1 S identified and can be said to appear in the wavelet. Different predefined values can be used at different scales to determine whether a feature appears in a given wavelet, e.g., by comparing the give predefined value to the given wavelet coefficient. Other techniques, however, can also be used to determine whether a feature appears in a given wavelet.
FIG. 7 is another graph of electrogram 20 comprising an atrial electrocardiogram. In FIG.
7, the X-axis is in units of time (milliseconds) and the Y axis is in units of amplitude (mV/200). FIG 8 is a corresponding graph illustrating scale maxima lines 81, 82, 83 of identifying features of electrogram 20 across various scales from coarse to fine. In FIG 8, the X-axis is in units of time (milliseconds) and the Y axis represents different scales of wavelets from coarse to fine.
The scale maxima lines 81, 82, 83 respectively identify whether the corresponding feature at that given time instance is identifiable in the wavelets at the different scales. As can be appreciated in FIG. 8, scale maxima line 81 spans all scales from coarse to fine.
Thus, the feature associated with scale maxima line 8I is identifiable in wavelets of every scale from coarse to fine. Scale maxima line 81 corresponds to portion 22 (FIG
2A) of electrogram 20. Thus, the large amplitude steep deflection associated with portion 22 appears in all scales of wavelets from coarse to fine.
Scale maxima line 82, on the other hand, is present only in fine scales. In particular, scale maxima line 82 stops as it approaches coarser scales. Thus, the feature associated with scale maxima line 82 is identifiable only in ~xne scale wavelets and not in coarse scale wavelets. Scale maxima line 82 corresponds to portion 32 (FIG. 213) of electrogram 20.
Thus, the small amplitude steep deflection associated with portion 32 appear only in fine scale wavelets.
Scale maxima line 83 is present only in the more coarse scales. In particular, scale maxima line 83 stops as it approaches the finest scales. Thus, the feature associated with scale maxima line 83 is identifiable in coarse scale wavelets and is not identifiable in the fine scale wavelets. Scale maxima line 83 corresponds to portion 42 (FIG 2C) of electrogram 20. Thus, the large amplitude shallow deflection associated with portion 42 appears in coarse scale wavelets, but not fine scale wavelets. In accordance with the invention, wavelet analysis unit 14 exploits the phenomenon that some features are identifiable in all scales of wavelets, some features are identifiable only in coarse scales, and some feattues axe identifiable only in fine scales.
FIG 9 is a flow diagram according to an embodiment of the invention. As shown in FIG. 9, wavelet transform unit 12 performs wavelet transformation to convert an electrogram into wavelets (91). In particular, the set of wavelets can be obtained by scaling and translation of a selected mother wavelet as described above. The set of wavelets may include numerous wavelets at various scale factors. The scale factors span from a coarse scale to a fine scale.
Wavelet analysis unit 14 analyzes the generated wavelets to identify features of the electrogram based on the wavelets. In particular, wavelet analysis unit 14 determines whether features appear in various wavelets, e.g., by determining whether a given wavelet coefficient exceeds a particular value. If nothing appears in fine scale wavelets (no branch of 92) or coarse scale wavelets (no branch of 93), wavelet analysis unit 14 determines that no feature was detected (94). If an electrogram featuxe appears in fine scale wavelets (yes branch of 92), but do not appear in coarse scale wavelets (no branch of 95), wavelet analysis unit 14 identifies the feature as a small amplitude steep deflection (96). If an electrogram feature does not appear in fine scale wavelets (no branch of 92), but does appear in coarse scale wavelets (yes branch of 93), wavelet analysis unit 14 identifies the feature as a large amplitude shallow deflection (97). If an electrogram feature appears in both fme scale wavelets (yes branch of 92) and coarse scale wavelets (yes branch of 95), wavelet analysis unit 14 identifies the feature as a large amplitude steep deflection (98). In this manner, wavelet analysis unit 14~ exploits wavelet analysis to identify electrogram features.
EXAMPLE
The wavelet transform techniques described herein were applied to a set of atrial signals obtained during acute atrial fibrillation. A database of electrograms contained each of the signals detected and extracted from the various electrodes positioned at various locations on the atrial free wall. All signals were annotated for their number of deflections originating from different simultaneous activation wavefronts. A range of one to five simultaneous wavefronts were used in the test. The test range was divided into two sub-sets with respect to the level of fragmentation. The first subset contained aI1 single and double wavefronts while the second subset contained all triple, quad and penta wavefronts.
All signals in the second subset were extracted from an acute atrial fibrillation database, and an equal size random sample of signals of the first subset was taken.
Some preprocessing steps were performed on the electrograms including wavelet denoising, e.g., by transforming the electrograms to wavelets, setting low amplitude wavelet coefficients to zero, and transforming the wavelets to generate denoised electrograms. The test set contained a total of 492 signals. No information regarding spatial relatiomwas included in the test set of signals. The continuous wavelet transform was calculated for all the signals in the test set using the first derivative of a Gaussian function. This wavelet family was built starting from the Gaussian function, taking the first derivative o~
O(t) = Cne'' , where ~~O~p~~~ - 1 ~f (t) _ ~t O(t) EQUATION 10 The wavelet transformation using the first derivative of the Gaussian function O, produced local maxima over the various scale pronounced down slopes of the electrogram signal.
A bandwidth assessment of the signals included in the test set revealed the analysis scales of the wavelet transform. Linear scales where fixed from normalized values of 1 to 20, with value 1 representing the finest scale and value 20 representing the coarsest scale.
Next the maxima were identified starting from the finest scale towards the coarse scales.
Local maxima where detected at the various scales using a threshold related to the signal standard deviation. In particular, the threshold was fixed to 0.5. ~ver the scales, chaining was achieved by allowing local maxima not to deviate not more 1 ms in both directions.
5 Violation of this criterion stopped the chaining process. Several classes of wavefronts can be identified purely from time domain information.
Large amplitude steep deflections indicating massive activation wavefronts passing underneath the recording electrode allowed chaining throughout all available scales.
Small amplitude steep deflections indicating one or more small wavefronts in the 10 recording electrode vicinity exclusively produced maxima in the finest scales. Large amplitude shallow deflections indicating massive depolarization at a remote location exclusively produced maxima only in the coarse scales.
The following MATLAB pseudo-code, below, illustrates one exemplary implementation of the invention. In particular, the pseudo-code in Table 1 can be stored 15 on a computer readable medium for execution in a DSP. In this manner, one embodiment of wavelet analysis unit 14 is realized in software executing on a DSP.
A number of embodiments of the invention have been described. However, one skilled in the art will appreciate that the invention can be practiced with embodiments other than those disclosed. For example, other types of mother wavelet functions can be used to generate the respective wavelets which are used in electrogram analysis. In addition, the invention can find application for analysis of a wide variety of different types of biomedical signals including but not limited to electrograms measured via external sensors, electrograms measured via implanted sensors, a signals measured by one or more a biomedical sensor, chronic or acute signals, or any other biomedical signal.
Also, although various techniques have been described as being implemented in software, similar techniques can be implemented in hardware, firmware, or the like.
Example hardware implementations of wavelet transform unit 12 and wavelet analysis unit 14 include implementations within an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a programmable logic device, specifically designed hardware components, one or more processors, or any combination thereof. If implemented in software, a computer readable medium stores computer readable instructions, e.g., program code, that can be executed by a processor or DSP
to carry out one of more of the techniques described above. For example, the computer readable medium can comprise random access memory (RAM), read-only memory (ROM), non-volatile random access memory (NVR.E11VVI), electrically erasable programmable read-only memory (EEPR~M), flash memory, or the like. The computer readable medimn can comprise computer readable instructions that when executed in a medical device carry out one or more of the techniques described herein. The disclosed embodiments are presented for purposes of illustration and not limitation, and the invention is limited only by the claims that follow.
In particular, it is often necessary to identify specific features of an electrogram so that medical events can be identified in the patient, such as specific depolari~ations in the patient's heart. For this reason, new and alternative techniques for processing electrograms are highly desirable. Signal processing of electrograms can be performed by any of a variety of medical devices including an IMI~ that senses the electrograms, an external medical device that senses the electrograms, or possibly an external medical device that receives sensed electrograms from another device, e.g., via telemetry.
In general, the invention is dixected to signal processing techniques for electrograms. In particular, the signal pxocessing techniques make use of wavelet transformation of the electrograms. For example, an electrogram can be represented by a finite set of wavelets which comprise a decomposition of the electrogram in the scale-time domain. hl accordance with the invention, wavelet analysis techniques can be used to distinguish specific phenomena in electrograms. For example, wavelet analysis can be used to distinguish between occurrences of large amplitude steep deflections, small amplitude steep deflections, and large amplitude shallow deflections of an electxogram.
In this manner, wavelet analysis can be useful in identification of R waves, P
waves and other electrogram events.
In one embodiment, the invention provides a method comprising converting an electrogram into a set of wavelets, the set of wavelets including wavelets at a fine scale and wavelets at a coarse scale, and identifying features of the electrogram based on whether the features appear in only the fine scale wavelets, only the coarse scale wavelets or both the fine and coarse scale wavelets.
In another embodiment, the invention provides a medical device comprising a wavelet transform unit to convert an electrogram into a set of wavelets, the set of wavelets including wavelets at a fine scale and wavelets at a coarse scale, and a wavelet analysis unit to identify features of the electrogram based on whether the features appear in only the fine scale wavelets, only the coarse scale wavelets or both the fine and coarse scale wavelets.
If implemented in software, the invention can be embodied in a computer readable medium comprising computer readable instructions that when executed convert an electrogram into a set of wavelets, the set of wavelets including wavelets at a fine scale and wavelets at a coarse scale, and identify features of the electrogram based on whether the features appear in only the fme scale wavelets, only the coarse scale wavelets or both the fine and coarse scale wavelets.
In another embodiment, the invention provides an apparatus comprising means for converting an electrogram into a set of wavelets, the set of wavelets including wavelets at a fine scale and wavelets at a coarse scale, and means for identifying features of the electrogram based on whether the features appear in only the fine scale wavelets, only the coarse scale wavelets or both the fine and coarse scale wavelets.
In another embodiment, the invention provides a method comprising converting an biomedical signal into a set of wavelets, the set of wavelets including wavelets at a fine scale and wavelets at a coarse scale, and identifrying features of the biomedical signal based on whether the features appeax in only the fine scale wavelets, only the coarse scale wavelets or both the fine and coarse scale wavelets.
The invention includes features that provide significant advances. For example, wavelet analysis can provide a relatively simple and straightforward mechanism for distinguishing among features or phenomena within electrograms. Accordingly, wavelet analysis can reduce the likelihood that the medical device will confuse the occurrence of large amplitude shallow deflections with the occurrence of large amplitude steep deflections. Similarly, implementation of the features of the present invention will enable the medical device to distinguish between the occurrence of large amplitude steep deflections and the occurrence of small amplitude steep deflections.
Moreover, the wavelet analysis techniques described herein can be well suited for implementation in IMDs, where computational resources are more limited and battery power is a concern. For example, the techniques described herein are much less computationally intensive than techniques in which correlation operations are performed to identify electrogram features.
The details of one or more embodiments of the invention are set forth in the accompanying drawings and the description below. C?ther features and aspects of the invention will be apparent from the description and drawings, and from the claims.
FIG. 1 is an exemplary block diagram of a medical device according to an embodiment of the invention.
FIGS. 2A-2I2 are graphs of an exemplary electrogram comprising an atrial electrocardiogram.
FIGS. 3-6 respectively illustrate traces of information contained in the wavelets of four different scales in relation to the electrogram illustrated in FIGS. 2A-2IO.
FIG 7 is another graph of an exemplary electrogram comprising an atrial electrocaxdiogram.
FIG. 8 is a graph illustrating scale maxima lines of various features of the electrogram of FIG 7 across various scales from coarse to fine.
FIG 9 is a flow diagxam according to an embodiment of the invention.
The invention is directed to signal processing techniques for electrograms,~or other biomedical signals, in which wavelet analysis is used. An electrogram can be represented by a finite set of wavelets which comprise a decomposition of the electrogram in the scale-time domain. In accordance with the invention, wavehet analysis techniques can be used to distinguish specific phenomena in electrograms or other biomedical signals.
In particular, wavelet analysis can be used to distinguish between occurrences of large amplitude steep deflections, small amplitude steep deflections, and large amplitude shallow deflections of an electrogram. Although many details of the invention are described in the context of ehectrograms, the techniques are equally useful for analysis of other types of biomedical signals.
Large amplitude steep deflections in an electrogram correspond to Iarge near field depolarizations, i.e., harge depolarizations traveling in close proximity to a sensing electrode. In cardiac settings, identification of large amplitude steep deflections identify primary cardiac depolarizations at the given location. Small amplitude steep deflections correspond to small neax field depolarizations, i.e., small depolarizations traveling in close proximity to the sensing electrode. In cardiac settings, identification of small amplitude steep deflections identify secondary cardiac depolarizations at the given location, e.g., commonly present during atrial fibrillation. Large amplitude shallow deflections in an electrogram correspond to primary far field depolarizations, i.e., large depolarizations not traveling in close proximity to the sensing electrode. In cardiac settings, identification of large amplitude shallow deflections identify a primary cardiac depolarization at a location that does not correspond to the sensing electrode, such as a ventricular depolarization sensed by an atrial electrode.
Distinguishing among large amplitude steep deflections, small amplitude steep deflections, and large amplitude shallow deflections of an electrogram can be challenging.
However, wavelet analysis provides a xelatively easy mechanism to do so. In particular, wavelets can be generated f~r an electrogram at various scales or scale factors. The scale factors span from a coarse scale to a fine scale. The coarse scale wavelets generally refer to wavelets having relatively large scale factors, whereas the fine scales generally refer to wavelets having relatively small scale factors. The ternzs coarse scale and fine scale, however, are generally relative and can assume different values in different embodiments.
Electrogram features that appear in alI scales of wavelets can be identified as large amplitude steep deflections. Electrogram feaW res that appear in fine scales but not coarse scales can be identified as small amplitude steep deflections. Electrogram features that appear in coarse scales but not fine scales can be identified as large amplitude shallow deflections. For example, the features can be said to appear in a given wavelet if a wavelet coefficient for that wavelet exceeds some predetermined threshold, and the threshold may be different for different scales.
In an electrogram, the differences between large amplitude steep deflections, small amplitude steep deflections, and large amplihide shallow deflections is generally relative, and can depend on the type of electrogram being analyzed. In other words, as used in this disclosure, the terms "large" and "small" are relative terms, and the terms "steep" and "shallow" are also relative terms. By way of example, amplitudes of atrial sensed atrial signals typically are in the range of 1-4 mV and have slopes exceeding 1 Volt per second.
Atrial (unipolar) sensed ventricular signals are generally approximately 0.1 -0.3 times the amplitude of atrial events, but this can vary considerably depending on the location of the atrial electrode, the indifferent electrode, and the size of the atrial electrode. The slope of atrial (unipolar) sensed ventricular signals is typically in the range of 0.1 Volts per second.
Despite the fact that the depolarization fields decrease with distance between the source and recording site, large far field depolarization wavefxonts, e.g., ventricular depolarization recorded by an atrial electrode, can contribute significantly to recorded amplitudes of electrograms. The spatial sensitivity of recording electrodes is wide for distant sources contributing to the recorded signal. However, the bandwidth related to this distance phenomena is relatively low.
In normal homogeneously coupled atrial tissue, simultaneously recorded tznipolar electrograms at various locations are typically biphasic with little morphological variation.
Comparison of electrograrn morphology recorded simultaneously from multiple neighboring locations can reveal infoz-mation on Local intercellular coupling either during atrial fibrillation or during normal sinus rhythm.
Spatial electrode sensitivity can be reduced by taking advantage of a priori information regarding the area of sensitivity related to recorded signal bandwidth.
Wavelet transforms are particularly useful in analysis of non-stationary signals because wavelet transforms provide an alternative to the classical Short Time Fourier Transform (STFT) and Gabor transform. The wavelet transform is typically a linear operation that decomposes a signal into components that appear at different scales (or resolutions). A
mother wavelet comprises a zero average function ~I' E L2(R) (finite energy):
~~I'(t)~lt = 0 (Admissibility condition) EQUATION 1 Equation 1 can be normalized ~~~I'~yl, and centered round t=0. Then, a family of wavelets can be obtained by scaling and translation of the mother wavelet ~I' by s, and translation by u:
'IJu,s(t)= ~~'(t su) EQUATIONZ
Wavelet analysis allows the use of coarse wavelets where more precise low-frequency information is needed, and fine wavelets where high-frequency information is required. In analogy to the STET, the wavelet transform is deftned as the sum over all time of the signal multiplied by scaled, shifted versions of the wavelet function. For functions f E LZ(R) the wavelet transform at time a and scale s is defined as:
Wf(u,s)=<.f,~'t,,s >= f.f(t)~I'*(t~u)dt EQUATION3 s This type of transform satisfies energy conservation. With decrease of scale 's', the support for the wavelet decreases and the wavelet becomes more sensitive to high-frequency components of the signal, enhancing finer grain details of the signal. An increase in scale, on the other hand, provides more emphasis on the coarse structure of the signal. The result of the wavelet transform can be defined in the scale-time plane. The wavelet transform can be rewritten as a convolution product:
Wf (u,s)= ~f(t)~I'*(t u)dt= f *~I'S(u), EQUATION4 s where ~S (t) = 1 'I" * (-t ) EQUATION 5 S
The Fourier transform of ~I'S(t) is:
~I'(~) _ ~~I'* (sue) EQUATION G
It appears that 'h is the transfer function of a band-pass filter, so the convolution can compute the wavelet transform with dilated impulse response band-pass filters.
Many electrograms, including electrocardiograms, carry most important information at their singularities and sharp deflections. The wavelet transform is particularly well adapted to characterize transient phenomena or singularities, because wavelet transforms decompose signals into building blocks well localized in time and frequency.
The wavelet transform can focus on localized signal structures with a zooming procedure that r progressively reduces the scale parameter 's.' A measure of local regularity of the signal is pxovided by the decay of the wavelet transform amplitude across its scales.
Singularities can be detected by following the wavelet transform local maxima at fine scales.
~ Wf (u,s) ~<_ As"+'~z EQUATION 7 From Equation 7, one can derive:
logZ ~ Wf (u, s) ~<_ logz A + (a + 1 / 2) logz s EQUATION 8 FIG. 1 is an exemplary block diagram of a medical device 10 according to an embodiment of the invention. Medical device 10 may comprise any of a wide variety of medical devices used to analyze electrograms. For example, medical device 10 may comprise an implanted medical device (IMD) that includes various implanted electrodes (not shown) used for sensing the electrograrns. Alternatively, medical device 10 lnay C0111pr1Se all eXterIlal medical device that uses surface electrodes on a patient's skin to sense the electrograms. Also, medical device can be an internal or external device that measure electrograms via subcutaneous electrodes. In other cases, medical device 10 may comprise an external device that receives sensed electrograms from another device, e.g., via telemetry. In any case, medical device 10 performs signal analysis on electrograms using wavelet analysis techniques as described herein.
In general, medical device 10 includes a wavelet transform unit I2 and a wavelet analysis unit 14. In the example illustrated in FIG. 1, wavelet transform unit 12 and a wavelet analysis unit 14 comprise software modules executed on a digital signal processor (DSP) 16. In that case, medical device 10 also includes memory 19 to store computer readable instructions that can be executed in DSP 16 to realize wavelet transform unit 12 and a wavelet analysis unit 14. For example, memory 19 can comprise random access memory (RAM), read-only memory (ROM), non-volatile random access memory (NVRA1VI), electrically erasable programmable read-only memory (EEPROM), flash memory, or the like.
Medical device 10 can also include an analog-to-digital (AlD) converter I8 to convert an analog electrogram to digital samples that comprise a digital electrogram. DSP
16 receives the digital electrogram and invokes wavelet transform unit I2 to transform the electrogram to wavelets and invokes wavelet analysis unit 14 to analyze the wavelets.
In other embodiments, however, wavelet transform unit 12 and a wavelet analysis unit 14 comprise dedicated hardware or logic that performs the functions described herein. Also, wavelet transform unit I2 and a wavelet analysis unit 14 can be implemented as one or more processors, application specific integrated circuits (ASICs), field progrannnable gate arrays (FPGAs), various combinations of hardware and software, or the like.
Moreover, in some embodiments wavelet transform unit 12 and a wavelet analysis unit 14 can be implemented as analog logic circuits. In that case, the need for DSP 16 and A/D converter 18 could be eliminated. However, the illustrated example of FIG. 1, in which wavelet transform unit 12 and wavelet analysis unit 14 comprise software modules executed on DSP 16, axe relatively easy and cost effective from an implementation standpoint. In that case, the software modules are initially stored in memory 19 and invoked by DSP 1 G to execute the techniques described herein.
Medical device 10 receives an analog electrogram (EGM) and A/D converter I 8 converts tl~e analog electrogram to a digital electrogram, i.e., a stream of digital samples that represent the electrogram. Again, medical device receives the electrogram from oxle or more sensing electrodes of medical device 10, or receives the electrogram from another device used to sense the electrograms.
Wavelet transform unit 12 receives an electrogram (in this case in digital form), axed performs wavelet transformation on the electrogram to generate a set of wavelets which collectively include the information in the electrogram. For example, wavelet transform unit 12 performs wavelet transformation using mathematical framework similax to that outlined above. In particular, the set of wavelets can be obtained by scaling and translation of a selected mother wavelet. Wavelet transform unit 12 can comprise a set of dilated impulse response band-pass filters designed to perform the desired wavelet transformation on the electrogram. The set of wavelets generated by wavelet transform unit 12 include numerous wavelets at various scale factors. The scale factors span from a coarse scale to a fine scale.
The coarse scale wavelets provide a larger overall picture of the electrograln, but lack specific details of the electrogram. The fine scale wavelets provide a less complete picture of the electrogram, but include more detail. The whole set of wavelets include wavelets of a number of different scale factors. The scale factors associated with coarse scale wavelets can be greater than or equal to 10 multiplied by scale factors associated with wavelets in the fme scale, although the invention is not necessarily limited in that respect.
Wavelet analysis unit 14 analyzes the generated wavelets to identify features of the electrogram based on the wavelets. In accordance with the invention, wavelet analysis unit 14 can distinguish between different electrogram features based on whether the features appear on coarse wavelets, fine waveiets, or both coarse and fine wavelets.
Wavelet analysis unit 14 outputs results based on this analysis. For example, the output of wavelet analysis unit 14 can comprise an indication of identified features and the timing of the identified features within the electrogram.
More specifically, wavelet analysis unit 14 identifies electrogram feaW res comprising large amplitude steep deflections when the features appear in all scales of wavelets. Also, wavelet analysis unit 14 identifies electrogram features comprising small amplitude steep deflections when the features appear in fine scales but not coarse scales.
Moreover, wavelet analysis unit 14 identifes electxogram features comprising large amplitude shallow deflections when the features appear in coarse scales but not fine scales. In this manner, wavelet analysis unit 14 analyzes the wavclets to identify characteristics of the electrogram.
FIG 2A is a graph of an exemplary clectrogram 20 comprising an atrial electrocardiogram, i.e., an electrogram ofthe heart sensed by an atrial electrode. The %-axis is in units of time (milliseconds) and tlae Y axis is in units of amplitude (mV1200).
Portion 22 is a large amplitude steep deflection in clectrogram 20, and corresponds to large near field depolarization, i.e., a large depolarization traveling in close proximity to the sensing electrode. Thus, identification of the large amplitude steep deflection (portion 22) identifies primary cardiac depolarizations at the given location. Wavelet transform 10 unit 14 identifies portion 22 as a large amplitude steep deflection because this electrogram feature appears in both coarse and fine wavelets.
FIG. 2B is another graph of electrogram 20 comprising an atrial electrocardiogram.
The X-axis is in units of time (milliseconds) and the Y axis is in units of amplitude (mV/200). As illustrated in FIG. 2B, portion 32 is a small amplitude steep deflection in electrogram 20, and corresponds to a small near field depolarization, i.e., small depolarizations traveling in close proximity to the sensing electrode.
Identification of small amplitude steep deflections (portion 32) identifies secondary cardiac depolaxizations at the given location, e.g., commonly present during atrial fibrillation.
Wavelet transform unit 14 identifies portion 32 as a small amplitude steep deflection because this electrogram feature appears in fme wavelets but does not appear in coarse wavelets.
FIG 2C is another graph of electrogram 20 comprising an atrial electrocardiogram.
The X-axis is in units of time (milliseconds) and the Y axis is in units of amplitude (mV/200). As illustrated in FIG 2C, portion 42 is a Large amplitude shallow deflection in electrogram 20, and corresponds to a primary far field depolarization, i.e., a large depolarization not traveling in close proximity to the sensing electrode.
Identification of large amplitude shallow deflections (portion 42) identifies primary cardiac depolarizations at a location that does not correspond to the sensing electrode. Wavelet transform unit 14 identifies portion 42 as a large amplitude shallow deflection because this electrogram feature appears in coarse wavelats but does not appear in fine wavelets.
Such wavelet analysis provides a relatively simple and straightforward mechanism fox distinguishing among features or phenomena within electrograms.
Accordingly, medical device 10 is less lilcely to confuse the occurrence of large amplitude shallow deflections with the occurrence of large amplitude steep deflections.
Similarly, medical device 10 is less likely to confuse the occurrence of haxge amplitude steep deflections with the occurrence of small amplitude steep deflections, and so forth. In this manner, signal processing of electrograms can be improved and simplified. In particular, the wavelet analysis techniques described herein are well suited fox implementation in IMDs, where camputationah resources are more limited and battery power is a concern. For example, the techniques described herein are much less computationally intensive than techniques in which correlation operations are performed to identify ehectrogram features.
FIG. 2D is another graph of electrogram 20 comprising an atrial electrocardiogram. FIGS. 3-6 respectively illustrate traces of information contained in the wavelets of four different scales. In each of FIGS. 3-6, the X-axis is in units of time (milliseconds) and the Y axis is in units of amplitude (mV/200).
In particular, FIG 3 illustrates a trace 35 associated with a scale factor of 4, FIG 4 illustrates a trace 45 associated with a scale factor of 2, FIG 5 illustrates a trace 5S
associated with a scale factor of 1 and FIG 6 illustrates a trace 65 associated with a scale factor of 0.4. The wavelets associated with trace 35 are fewer than the wavelets of the other traces because trace 35 represent the coarse wavelets in this example.
In other words, every wavelet used to define trace 35 spans a larger portion of electrogram 20 than wavelets of the other ttaces 45, 55, 65.
On the other hand, the wavelets associated with trace 65 are more numerous than the wavelets of the other traces because trace 65 represents the fine wavelets in this example. In other words, every wavehet used to define trace 65 spans a smaller portion of electrogram 20 than wavelets of the other traces 35, 45, 55. Each respective trace 35, 45, 55, 65 is a general representation of the information collectively contained in all of the wavelets at the respective scale.
As can be appreciated by examining FIGS. 3-6, portion 22 (FIG 2A) of electrogram 20 is identifiable in every trace 35, 45, 55, 65. In other words, the wavelets at every scale include information indicative of the large amplitude steep deflection associated with portion 22. Portion 32 (FIG 2B) of electrogram 20, however, is identifiable only in traces 5S, 6S and is generally not present in traces 3S, 45. W other words, only the wavelets at fine scales include information indicative of the sznahl amplitude steep deflection associated with portion 32.
Portion 42 (FIG. 2B) of electrogxam 20, is identifiable only in traces 3S, 4S
and SS, but is difficult to identify in trace 65. At even finer scales, portion 42 would become completely unidentifiable. Thus, only the wavelets at more coarse scales include information indicative of the large amplitude shallow deflection associated with portion S 42.
Wavelet analysis unit 14 exploits the fact that some features arc identifiable in all scales, some features are identifiable only in coarse scales, and some features are identifiable only in fine scales. In particular, wavelet analysis unit 14 identifies large amplitude steep deflections when the features appear in all scales of wavelets, identifies small amplitude steep deflections when the features appear in fine scales but not coarse scales, and identifies large amplitude shallow deflections when the features appear in coarse scales but not fine scales. As used in this disclosure, the term "appear" generally means that the given feature can be identified in the respective scale. For example, if a given wavelet coefficient exceeds a particular predefined value, the feature can be 1 S identified and can be said to appear in the wavelet. Different predefined values can be used at different scales to determine whether a feature appears in a given wavelet, e.g., by comparing the give predefined value to the given wavelet coefficient. Other techniques, however, can also be used to determine whether a feature appears in a given wavelet.
FIG. 7 is another graph of electrogram 20 comprising an atrial electrocardiogram. In FIG.
7, the X-axis is in units of time (milliseconds) and the Y axis is in units of amplitude (mV/200). FIG 8 is a corresponding graph illustrating scale maxima lines 81, 82, 83 of identifying features of electrogram 20 across various scales from coarse to fine. In FIG 8, the X-axis is in units of time (milliseconds) and the Y axis represents different scales of wavelets from coarse to fine.
The scale maxima lines 81, 82, 83 respectively identify whether the corresponding feature at that given time instance is identifiable in the wavelets at the different scales. As can be appreciated in FIG. 8, scale maxima line 81 spans all scales from coarse to fine.
Thus, the feature associated with scale maxima line 8I is identifiable in wavelets of every scale from coarse to fine. Scale maxima line 81 corresponds to portion 22 (FIG
2A) of electrogram 20. Thus, the large amplitude steep deflection associated with portion 22 appears in all scales of wavelets from coarse to fine.
Scale maxima line 82, on the other hand, is present only in fine scales. In particular, scale maxima line 82 stops as it approaches coarser scales. Thus, the feature associated with scale maxima line 82 is identifiable only in ~xne scale wavelets and not in coarse scale wavelets. Scale maxima line 82 corresponds to portion 32 (FIG. 213) of electrogram 20.
Thus, the small amplitude steep deflection associated with portion 32 appear only in fine scale wavelets.
Scale maxima line 83 is present only in the more coarse scales. In particular, scale maxima line 83 stops as it approaches the finest scales. Thus, the feature associated with scale maxima line 83 is identifiable in coarse scale wavelets and is not identifiable in the fine scale wavelets. Scale maxima line 83 corresponds to portion 42 (FIG 2C) of electrogram 20. Thus, the large amplitude shallow deflection associated with portion 42 appears in coarse scale wavelets, but not fine scale wavelets. In accordance with the invention, wavelet analysis unit 14 exploits the phenomenon that some features are identifiable in all scales of wavelets, some features are identifiable only in coarse scales, and some feattues axe identifiable only in fine scales.
FIG 9 is a flow diagram according to an embodiment of the invention. As shown in FIG. 9, wavelet transform unit 12 performs wavelet transformation to convert an electrogram into wavelets (91). In particular, the set of wavelets can be obtained by scaling and translation of a selected mother wavelet as described above. The set of wavelets may include numerous wavelets at various scale factors. The scale factors span from a coarse scale to a fine scale.
Wavelet analysis unit 14 analyzes the generated wavelets to identify features of the electrogram based on the wavelets. In particular, wavelet analysis unit 14 determines whether features appear in various wavelets, e.g., by determining whether a given wavelet coefficient exceeds a particular value. If nothing appears in fine scale wavelets (no branch of 92) or coarse scale wavelets (no branch of 93), wavelet analysis unit 14 determines that no feature was detected (94). If an electrogram featuxe appears in fine scale wavelets (yes branch of 92), but do not appear in coarse scale wavelets (no branch of 95), wavelet analysis unit 14 identifies the feature as a small amplitude steep deflection (96). If an electrogram feature does not appear in fine scale wavelets (no branch of 92), but does appear in coarse scale wavelets (yes branch of 93), wavelet analysis unit 14 identifies the feature as a large amplitude shallow deflection (97). If an electrogram feature appears in both fme scale wavelets (yes branch of 92) and coarse scale wavelets (yes branch of 95), wavelet analysis unit 14 identifies the feature as a large amplitude steep deflection (98). In this manner, wavelet analysis unit 14~ exploits wavelet analysis to identify electrogram features.
EXAMPLE
The wavelet transform techniques described herein were applied to a set of atrial signals obtained during acute atrial fibrillation. A database of electrograms contained each of the signals detected and extracted from the various electrodes positioned at various locations on the atrial free wall. All signals were annotated for their number of deflections originating from different simultaneous activation wavefronts. A range of one to five simultaneous wavefronts were used in the test. The test range was divided into two sub-sets with respect to the level of fragmentation. The first subset contained aI1 single and double wavefronts while the second subset contained all triple, quad and penta wavefronts.
All signals in the second subset were extracted from an acute atrial fibrillation database, and an equal size random sample of signals of the first subset was taken.
Some preprocessing steps were performed on the electrograms including wavelet denoising, e.g., by transforming the electrograms to wavelets, setting low amplitude wavelet coefficients to zero, and transforming the wavelets to generate denoised electrograms. The test set contained a total of 492 signals. No information regarding spatial relatiomwas included in the test set of signals. The continuous wavelet transform was calculated for all the signals in the test set using the first derivative of a Gaussian function. This wavelet family was built starting from the Gaussian function, taking the first derivative o~
O(t) = Cne'' , where ~~O~p~~~ - 1 ~f (t) _ ~t O(t) EQUATION 10 The wavelet transformation using the first derivative of the Gaussian function O, produced local maxima over the various scale pronounced down slopes of the electrogram signal.
A bandwidth assessment of the signals included in the test set revealed the analysis scales of the wavelet transform. Linear scales where fixed from normalized values of 1 to 20, with value 1 representing the finest scale and value 20 representing the coarsest scale.
Next the maxima were identified starting from the finest scale towards the coarse scales.
Local maxima where detected at the various scales using a threshold related to the signal standard deviation. In particular, the threshold was fixed to 0.5. ~ver the scales, chaining was achieved by allowing local maxima not to deviate not more 1 ms in both directions.
5 Violation of this criterion stopped the chaining process. Several classes of wavefronts can be identified purely from time domain information.
Large amplitude steep deflections indicating massive activation wavefronts passing underneath the recording electrode allowed chaining throughout all available scales.
Small amplitude steep deflections indicating one or more small wavefronts in the 10 recording electrode vicinity exclusively produced maxima in the finest scales. Large amplitude shallow deflections indicating massive depolarization at a remote location exclusively produced maxima only in the coarse scales.
The following MATLAB pseudo-code, below, illustrates one exemplary implementation of the invention. In particular, the pseudo-code in Table 1 can be stored 15 on a computer readable medium for execution in a DSP. In this manner, one embodiment of wavelet analysis unit 14 is realized in software executing on a DSP.
A number of embodiments of the invention have been described. However, one skilled in the art will appreciate that the invention can be practiced with embodiments other than those disclosed. For example, other types of mother wavelet functions can be used to generate the respective wavelets which are used in electrogram analysis. In addition, the invention can find application for analysis of a wide variety of different types of biomedical signals including but not limited to electrograms measured via external sensors, electrograms measured via implanted sensors, a signals measured by one or more a biomedical sensor, chronic or acute signals, or any other biomedical signal.
Also, although various techniques have been described as being implemented in software, similar techniques can be implemented in hardware, firmware, or the like.
Example hardware implementations of wavelet transform unit 12 and wavelet analysis unit 14 include implementations within an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a programmable logic device, specifically designed hardware components, one or more processors, or any combination thereof. If implemented in software, a computer readable medium stores computer readable instructions, e.g., program code, that can be executed by a processor or DSP
to carry out one of more of the techniques described above. For example, the computer readable medium can comprise random access memory (RAM), read-only memory (ROM), non-volatile random access memory (NVR.E11VVI), electrically erasable programmable read-only memory (EEPR~M), flash memory, or the like. The computer readable medimn can comprise computer readable instructions that when executed in a medical device carry out one or more of the techniques described herein. The disclosed embodiments are presented for purposes of illustration and not limitation, and the invention is limited only by the claims that follow.
Claims (30)
1. An implantable medical device comprising:
a wavelet transform unit to convert an electrogram into a set of wavelets, the set of wavelets including wavelets at a fine scale and wavelets at a coarse scale;
and a wavelet analysis unit to identify features of the electrogram based on whether the features appear in only the fine scale wavelets, only the coarse scale wavelets or both the fine and coarse scale wavelets.
a wavelet transform unit to convert an electrogram into a set of wavelets, the set of wavelets including wavelets at a fine scale and wavelets at a coarse scale;
and a wavelet analysis unit to identify features of the electrogram based on whether the features appear in only the fine scale wavelets, only the coarse scale wavelets or both the fine and coarse scale wavelets.
2. The implantable medical device of claim 1, wherein the electrogram comprises a digital electrogram, the medical device further comprising an analog-to-digital converter to convert an analog electrogram to the digital electrogram, wherein the wavelet transform unit converts the digital electrogram into the set of wavelet and the wavelet analysis unit distinguishes features of the digital electrogram.
3. The implantable medical device of claim 1, further comprising a digital signal processor, wherein the wavelet transform unit and the wavelet analysis unit comprise software modules executed by the digital signal processor.
4. The implantable medical device of claim 1, further comprising one or more electrodes to sense the electrogram.
5. The implantable medical device of claim 1, wherein the medical device comprises one of an external medical device and an implanted medical device.
6. The implantable medical device of claim 1, wherein the medical device receives the electrogram from another medical device via telemetry.
7. The implantable medical device of claim 1, wherein wavelets in the coarse scale have a scale factor greater than or equal to 10 multiplied by a scale factor associated with wavelets in the fine scale.
8. The implantable medical device of claim 1, wherein the wavelet analysis unit identifies a particular feature in the electrogram as a large amplitude steep deflection when the particular feature appears in the fine and coarse scale wavelets.
9. The implantable medical device of claim 1, wherein the wavelet analysis unit identifies a particular feature in the electrogram as a small amplitude steep deflection when the particular feature appears in the fine scales wavelets but not the coarse scale wavelets.
10. The implantable medical device of claim 1, wherein the wavelet analysis unit identifies a particular feature in the electrogram as a large amplitude shallow deflection when the particular feature appears in the coarse scale wavelets but not the fine scale wavelets.
11. The implantable medical device of claim 1, wherein the electrogram comprises an atrial electrocardiogram, wherein the wavelet analysis unit:
identifies a first feature in the atrial electrocardiogram as an indication of primary near field depolarization when the first feature appears in the fine and coarse scale wavelets;
identifies a second feature in the atrial electrocardiogram as a secondary near field depolarization when the second feature appears in the fine scales wavelets but not the coarse scale wavelets; and identifies a third feature in the atrial electrocardiogram as a far field primary depolarization when the third feature appears in the coarse scale wavelets but not the fine scale wavelets.
identifies a first feature in the atrial electrocardiogram as an indication of primary near field depolarization when the first feature appears in the fine and coarse scale wavelets;
identifies a second feature in the atrial electrocardiogram as a secondary near field depolarization when the second feature appears in the fine scales wavelets but not the coarse scale wavelets; and identifies a third feature in the atrial electrocardiogram as a far field primary depolarization when the third feature appears in the coarse scale wavelets but not the fine scale wavelets.
12. An implantable medical device comprising:
a wavelet transform unit to convert a biomedical signal into a set of wavelets, the set of wavelets including wavelets at a fine scale and wavelets at a coarse scale; and a wavelet analysis unit to identify features of the biomedical signal based on whether the features appear in only the fine scale wavelets, only the coarse scale wavelets or both the fine and coarse scale wavelets.
a wavelet transform unit to convert a biomedical signal into a set of wavelets, the set of wavelets including wavelets at a fine scale and wavelets at a coarse scale; and a wavelet analysis unit to identify features of the biomedical signal based on whether the features appear in only the fine scale wavelets, only the coarse scale wavelets or both the fine and coarse scale wavelets.
13. A method comprising:
converting an electrogram into a set of wavelets, the set of wavelets including wavelets at a fine scale and wavelets at a coarse scale; and identifying features of the electrogram based on whether the features appear in only the fine scale wavelets, only the coarse scale wavelets or both the fine and coarse scale wavelets.
converting an electrogram into a set of wavelets, the set of wavelets including wavelets at a fine scale and wavelets at a coarse scale; and identifying features of the electrogram based on whether the features appear in only the fine scale wavelets, only the coarse scale wavelets or both the fine and coarse scale wavelets.
14. The method of claim 13, wherein wavelets in the coarse scale have scale factors greater than or equal to 10 multiplied by a scale factor associated with wavelets in the fine scale.
15. The method of claim 13, further comprising identifying a particular feature in the electrogram as a large amplitude steep deflection when the particular feature appears in the fine and coarse scale wavelets.
16. The method of claim 13, further comprising identifying a particular feature in the electrogram as a small amplitude steep deflection when the particular feature appears in the fine scales wavelets but not the coarse scale wavelets.
17. The method of claim 13, further comprising identifying a particular feature in the electrogram as a large amplitude shallow deflection when the particular feature appears in the coarse scale wavelets but not the fine scale wavelets.
18. The method of claim 13, wherein the electrogram comprises an atrial electrocardiogram, and identifying features further comprises:
identifying a first feature in the atrial electrocardiogram as an indication of primary near field depolarization when the first feature appears in the fine and coarse scale wavelets;
identifying a second feature in the atrial electrocardiogram as a secondary near field depolarization when the second feature appears in the fine scales wavelets but not the coarse scale wavelets; and identifying a third feature in the atrial electrocardiogram as a far field primary depolarization when the third feature appears in the coarse scale wavelets but not the fine scale wavelets.
identifying a first feature in the atrial electrocardiogram as an indication of primary near field depolarization when the first feature appears in the fine and coarse scale wavelets;
identifying a second feature in the atrial electrocardiogram as a secondary near field depolarization when the second feature appears in the fine scales wavelets but not the coarse scale wavelets; and identifying a third feature in the atrial electrocardiogram as a far field primary depolarization when the third feature appears in the coarse scale wavelets but not the fine scale wavelets.
19. A computer readable medium comprising computer readable instructions that when executed:
convert an electrogram into a set of wavelets, the set of wavelets including wavelets at a fine scale and wavelets at a coarse scale; and identify features of the electrogram based on whether the features appear in only the fine scale wavelets, only the coarse scale wavelets or both the fine and coarse scale wavelets.
convert an electrogram into a set of wavelets, the set of wavelets including wavelets at a fine scale and wavelets at a coarse scale; and identify features of the electrogram based on whether the features appear in only the fine scale wavelets, only the coarse scale wavelets or both the fine and coarse scale wavelets.
20. The computer readable medium of claim 19, wherein wavelets in the coarse scale have a scale factor greater than or equal to 10 multiplied by a scale factor associated with wavelets in the fine scale.
21. The computer readable medium of claim 19, further comprising computer readable instructions that when executed identify a particular feature in the electrogram as a large amplitude steep deflection when the particular feature appears in the fine and coarse scale wavelets.
22. The computer readable medium of claim 19, further comprising computer readable instructions that when executed identify a particular feature in the electrogram as a small amplitude steep deflection when the particular feature appears in the fine scales wavelets but not the coarse scale wavelets.
23. The computer readable medium of claim 19, further comprising computer readable instructions that when executed identify a particular feature in the electrogram as a large amplitude shallow deflection when the particular feature appears in the coarse scale wavelets but not the fine scale wavelets.
24. The computer readable medium of claim 19, wherein the electrogram comprises an atrial electrocardiogram, further comprising computer readable instructions that when executed:
identify a first feature in the atrial electrocardiogram as an indication of primary near field depolarization when the first feature appears in the fine and coarse scale wavelets;
identify a second feature in the atrial electrocardiogram as a secondary near field depolarization when the second feature appears in the fine scales wavelets but not the coarse scale wavelets; and identify a third feature in the atrial electrocardiogram as a far field primary depolarization when the third feature appears in the coarse scale wavelets but not the fine scale wavelets.
identify a first feature in the atrial electrocardiogram as an indication of primary near field depolarization when the first feature appears in the fine and coarse scale wavelets;
identify a second feature in the atrial electrocardiogram as a secondary near field depolarization when the second feature appears in the fine scales wavelets but not the coarse scale wavelets; and identify a third feature in the atrial electrocardiogram as a far field primary depolarization when the third feature appears in the coarse scale wavelets but not the fine scale wavelets.
25. An apparatus comprising:
means for converting an electrogram into a set of wavelets, the set of wavelets including wavelets at a fine scale and wavelets at a coarse scale; and means for identifying features of the electrogram based on whether the features appear in only the fine scale wavelets, only the coarse scale wavelets or both the fine and coarse scale wavelets.
means for converting an electrogram into a set of wavelets, the set of wavelets including wavelets at a fine scale and wavelets at a coarse scale; and means for identifying features of the electrogram based on whether the features appear in only the fine scale wavelets, only the coarse scale wavelets or both the fine and coarse scale wavelets.
26. The apparatus of claim 25, further comprising means for identifying a particular feature in the electrogram as a large amplitude steep deflection when the particular feature appears in the fine and coarse scale wavelets.
27. The apparatus of claim 25, further comprising means for identifying a particular feature in the electrogram as a small amplitude steep deflection when the particular feature appears in the fine scales wavelets but not the coarse scale wavelets.
28. The apparatus of claim 25, further comprising means for identifying a particular feature in the electrogram as a large amplitude shallow deflection when the particular feature appears in the coarse scale wavelets but not the fine scale wavelets.
29. A method comprising:
converting an biomedical signal into a set of wavelets, the set of wavelets including wavelets at a fine scale and wavelets at a coarse scale; and identifying features of the biomedical signal based on whether the features appear in only the fine scale wavelets, only the coarse scale wavelets or both the fine and coarse scale wavelets.
converting an biomedical signal into a set of wavelets, the set of wavelets including wavelets at a fine scale and wavelets at a coarse scale; and identifying features of the biomedical signal based on whether the features appear in only the fine scale wavelets, only the coarse scale wavelets or both the fine and coarse scale wavelets.
30. The method of claim 29, wherein the biomedical signal comprises a signal selected from the group consisting of: an electrogram measured via external sensors, an electrogram measured via implanted sensors, and a signal measured by a biomedical sensor.
Applications Claiming Priority (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US10/414,758 | 2003-04-16 | ||
US10/414,758 US7082327B2 (en) | 2003-04-16 | 2003-04-16 | Biomedical signal analysis techniques using wavelets |
PCT/US2004/011749 WO2004095307A2 (en) | 2003-04-16 | 2004-04-16 | Biomedical signal analysis techniques using wavelets |
Publications (1)
Publication Number | Publication Date |
---|---|
CA2522732A1 true CA2522732A1 (en) | 2004-11-04 |
Family
ID=33158766
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CA002522732A Abandoned CA2522732A1 (en) | 2003-04-16 | 2004-04-16 | Biomedical signal analysis techniques using wavelets |
Country Status (5)
Country | Link |
---|---|
US (1) | US7082327B2 (en) |
EP (1) | EP1614049B1 (en) |
JP (1) | JP4668175B2 (en) |
CA (1) | CA2522732A1 (en) |
WO (1) | WO2004095307A2 (en) |
Families Citing this family (27)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1296591B1 (en) | 2000-04-17 | 2018-11-14 | Adidas AG | Systems for ambulatory monitoring of physiological signs |
US7809433B2 (en) * | 2005-08-09 | 2010-10-05 | Adidas Ag | Method and system for limiting interference in electroencephalographic signals |
US7412103B2 (en) * | 2003-10-20 | 2008-08-12 | Lawrence Livermore National Security, Llc | 3D wavelet-based filter and method |
US9492084B2 (en) | 2004-06-18 | 2016-11-15 | Adidas Ag | Systems and methods for monitoring subjects in potential physiological distress |
US9504410B2 (en) | 2005-09-21 | 2016-11-29 | Adidas Ag | Band-like garment for physiological monitoring |
US7412282B2 (en) * | 2005-01-26 | 2008-08-12 | Medtronic, Inc. | Algorithms for detecting cardiac arrhythmia and methods and apparatuses utilizing the algorithms |
US7567835B2 (en) * | 2005-04-18 | 2009-07-28 | Medtronic, Inc. | Method and apparatus for identifying oversensing using far-field intracardiac electrograms and marker channels |
EP1887933B1 (en) | 2005-05-20 | 2017-01-18 | Adidas AG | Methods and systems for determining dynamic hyperinflation |
US8033996B2 (en) | 2005-07-26 | 2011-10-11 | Adidas Ag | Computer interfaces including physiologically guided avatars |
US8762733B2 (en) | 2006-01-30 | 2014-06-24 | Adidas Ag | System and method for identity confirmation using physiologic biometrics to determine a physiologic fingerprint |
WO2007092543A2 (en) * | 2006-02-06 | 2007-08-16 | The Board Of Trustees Of The Leland Stanford Junior University | Non-invasive cardiac monitor and methods of using continuously recorded cardiac data |
US8475387B2 (en) | 2006-06-20 | 2013-07-02 | Adidas Ag | Automatic and ambulatory monitoring of congestive heart failure patients |
US9833184B2 (en) | 2006-10-27 | 2017-12-05 | Adidas Ag | Identification of emotional states using physiological responses |
US8306290B2 (en) | 2007-04-20 | 2012-11-06 | Sierra Scientific Instruments, Llc | Diagnostic system for display of high-resolution physiological data of multiple properties |
US8684920B2 (en) * | 2008-07-07 | 2014-04-01 | Medtronic, Inc. | Classification of severity of medical condition by wavelet based multi-resolution analysis |
US8768469B2 (en) | 2008-08-08 | 2014-07-01 | Enteromedics Inc. | Systems for regulation of blood pressure and heart rate |
SG188867A1 (en) | 2009-09-11 | 2013-04-30 | Agency Science Tech & Res | Electrocardiogram signal processing system |
WO2011143490A2 (en) | 2010-05-12 | 2011-11-17 | Irhythm Technologies, Inc. | Device features and design elements for long-term adhesion |
JP5699403B2 (en) * | 2010-08-18 | 2015-04-08 | 学校法人 芝浦工業大学 | Irregular pulse train signal processing method and signal processing apparatus |
WO2013112935A1 (en) * | 2012-01-25 | 2013-08-01 | The Regents Of The University Of California | Systems and methods for automatic segment selection for multi-dimensional biomedical signals |
TWI499404B (en) * | 2012-11-19 | 2015-09-11 | Univ Nat Chiao Tung | Sleep apnea detection system and method |
AU2014209376B2 (en) | 2013-01-24 | 2017-03-16 | Irhythm Technologies, Inc. | Physiological monitoring device |
EP2983593B1 (en) | 2013-04-08 | 2021-11-10 | Irhythm Technologies, Inc. | Skin abrader |
EP4218580A1 (en) | 2014-10-31 | 2023-08-02 | Irhythm Technologies, Inc. | Wireless physiological monitoring device and systems |
KR20230119036A (en) | 2020-02-12 | 2023-08-14 | 아이리듬 테크놀로지스, 아이엔씨 | Non-invasive cardiac monitor and methods of using recorded cardiac data to infer a physiological characteristic of a patient |
US11350865B2 (en) | 2020-08-06 | 2022-06-07 | Irhythm Technologies, Inc. | Wearable device with bridge portion |
CA3188343A1 (en) | 2020-08-06 | 2022-02-10 | Jeff ABERCROMBIE | Electrical components for physiological monitoring device |
Family Cites Families (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5439483A (en) * | 1993-10-21 | 1995-08-08 | Ventritex, Inc. | Method of quantifying cardiac fibrillation using wavelet transform |
JPH0928687A (en) * | 1995-07-14 | 1997-02-04 | Toyota Central Res & Dev Lab Inc | Processing electrocargiographic signal and apparatus therefor |
US5755739A (en) * | 1996-12-04 | 1998-05-26 | Medtronic, Inc. | Adaptive and morphological system for discriminating P-waves and R-waves inside the human body |
US5778881A (en) * | 1996-12-04 | 1998-07-14 | Medtronic, Inc. | Method and apparatus for discriminating P and R waves |
JPH10216096A (en) * | 1997-02-04 | 1998-08-18 | Matsushita Electric Ind Co Ltd | Biological signal analyzing device |
US6324421B1 (en) * | 1999-03-29 | 2001-11-27 | Medtronic, Inc. | Axis shift analysis of electrocardiogram signal parameters especially applicable for multivector analysis by implantable medical devices, and use of same |
JP2003531656A (en) * | 1999-05-01 | 2003-10-28 | ザ コート オブ ネーピア ユニバーシティー | How to analyze medical signals |
WO2000069517A1 (en) * | 1999-05-12 | 2000-11-23 | Medtronic, Inc. | Monitoring apparatus using wavelet transforms for the analysis of heart rhythms |
US6440082B1 (en) * | 1999-09-30 | 2002-08-27 | Medtronic Physio-Control Manufacturing Corp. | Method and apparatus for using heart sounds to determine the presence of a pulse |
JP2001218747A (en) * | 2000-02-10 | 2001-08-14 | Tokyo Denki Univ | Electrocardiogram analysis system and computer readable recording medium having program therefor recorded thereon |
US6438419B1 (en) * | 2000-09-28 | 2002-08-20 | The University Of Pittsburgh | Method and apparatus employing a scaling exponent for selectively defibrillating a patient |
EP1429839A2 (en) * | 2001-08-28 | 2004-06-23 | Medtronic, Inc. | Implantable medical device for treating cardiac mechanical dysfunction by electrical stimulation |
US6810283B2 (en) * | 2001-09-13 | 2004-10-26 | Medtronic, Inc. | Multiple templates for filtering of far field R-waves |
-
2003
- 2003-04-16 US US10/414,758 patent/US7082327B2/en not_active Expired - Fee Related
-
2004
- 2004-04-16 EP EP04750204A patent/EP1614049B1/en not_active Expired - Lifetime
- 2004-04-16 JP JP2006510114A patent/JP4668175B2/en not_active Expired - Fee Related
- 2004-04-16 CA CA002522732A patent/CA2522732A1/en not_active Abandoned
- 2004-04-16 WO PCT/US2004/011749 patent/WO2004095307A2/en active Application Filing
Also Published As
Publication number | Publication date |
---|---|
EP1614049B1 (en) | 2012-02-22 |
WO2004095307A2 (en) | 2004-11-04 |
US20040210147A1 (en) | 2004-10-21 |
EP1614049A2 (en) | 2006-01-11 |
JP4668175B2 (en) | 2011-04-13 |
US7082327B2 (en) | 2006-07-25 |
JP2006523517A (en) | 2006-10-19 |
WO2004095307A3 (en) | 2005-06-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US7082327B2 (en) | Biomedical signal analysis techniques using wavelets | |
EP1613209B1 (en) | Biomedical signal denoising techniques | |
US6393316B1 (en) | Method and apparatus for detection and treatment of cardiac arrhythmias | |
Banerjee et al. | Delineation of ECG characteristic features using multiresolution wavelet analysis method | |
US5000189A (en) | Method and system for monitoring electrocardiographic signals and detecting a pathological cardiac arrhythmia such as ventricular tachycardia | |
DE60308725T2 (en) | SYSTEMS FOR THE DISTINCTION OF POLYMORPHES FROM MONOMOTH TACHYARRHYTHOMES | |
Sivannarayana et al. | Biorthogonal wavelet transforms for ECG parameters estimation | |
US8521276B2 (en) | Use of significant point methodology to prevent inappropriate therapy | |
EP0653224A2 (en) | A method of classifying heart rhythms by means of electrogram morphology | |
US20070270912A1 (en) | Method and apparatus for discriminating cardiac signals in a medical device based on wavelet decomposition analysis | |
JP2007522536A (en) | Dynamic blind signal separation | |
KR101048763B1 (en) | Apparauts and method for detecting signal | |
Sharma et al. | ECG Analysis-Based Cardiac Disease Prediction Using Signal Feature Selection with Extraction Based on AI Techniques | |
Kumar et al. | Robust multiresolution wavelet analysis and window search based approach for electrocardiogram features delineation | |
Zubair et al. | Electromyography noise suppression in electrocardiogram signal using modified garrote threshold shrinkage function | |
Koyrakh et al. | Wavelet transform based algorithms for EGM morphology discrimination for implantable ICDs | |
de Jesus Kozakevicius et al. | Adaptive ECG filtering and QRS detection using orthogonal wavelet transform | |
Simon et al. | Impact of sampling rate reduction on automatic ECG delineation | |
Verma et al. | Comparative study of qrs complex detection by threshold technique | |
Brooks et al. | Temporal and spatial analysis of potential maps via multiresolution decompositions | |
Rathore et al. | Detection of QRS Complex in ECG Signal using Wavelet Transform and Thresholding Technique | |
Sumathi | Cardiovascular Pharmacology: Open Access | |
Ropella et al. | Quantitative Descriptions of Cardiac Arrhythmias | |
Chawla et al. | Modeling and feature extraction of ECG using independent component analysis | |
Hashim et al. | Novel ECG analysis with application to atrial fibrillation detection |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
FZDE | Discontinued |