WO2008007236A2 - Atrial fibrillation detection - Google Patents
Atrial fibrillation detection Download PDFInfo
- Publication number
- WO2008007236A2 WO2008007236A2 PCT/IB2007/052095 IB2007052095W WO2008007236A2 WO 2008007236 A2 WO2008007236 A2 WO 2008007236A2 IB 2007052095 W IB2007052095 W IB 2007052095W WO 2008007236 A2 WO2008007236 A2 WO 2008007236A2
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- ecg
- analyzing
- classifier
- atrial fibrillation
- interval
- Prior art date
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
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
-
- 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
- A61B5/361—Detecting fibrillation
-
- 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
- A61B5/25—Bioelectric electrodes therefor
-
- 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]
- A61B5/332—Portable devices specially adapted therefor
-
- 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
- A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
Definitions
- This invention relates to the detection of atrial fibrillation, on the basis of ECG readings from electrodes. It may be used for example with readings derived from electrodes incorporated in "wearable" electrode systems.
- Atrial fibrillation is a common heart arrhythmia with a prevalence of approximately 0.4 to 1% of the general population, and its prevalence increases with age. It is responsible for the highest number of hospital admissions due to arrhythmias, and consequently, it is desirable to be able to monitor the condition of patients, using portable devices which are capable of producing reliable indications of arrhythmia, without producing false positives.
- Electrocardiograph (ECG) signals show a characteristic pattern of electrical impulses that are generated by the heart. Different waves are identifiable in the ECG signal - the P wave is from the atrial excitation and the QRS complex and T-wave are from the ventricular excitation and relaxation, respectively, as illustrated in Figure 1.
- the ST segment usually appears as a straight, level line between the QRS complex and the T wave. Elevated or depressed ST segments may mean the heart muscle is damaged or not receiving enough blood, a sign that a myocardial infarct may have occurred.
- the T wave corresponds to the period when the lower heart chambers are relaxing electrically and preparing for their next muscle contraction.
- AF is a heart rhythm which is usually characterized by QRS complexes with normal morphology and with irregular arrival times. This can be caused by a diseased atrium which disrupts the normal passage of electrical stimulus from the sinus node through the atrium to the ventricles.
- QRS complexes QRS complexes with normal morphology and with irregular arrival times. This can be caused by a diseased atrium which disrupts the normal passage of electrical stimulus from the sinus node through the atrium to the ventricles.
- Figure 2 One example of AF is depicted in Figure 2.
- AF can be either chronic or intermittent. Intermittent AF is referred to as paroxysmal AF. AF is difficult to detect, particularly if it is paroxysmal, since a sample ECG recording from paroxysmal AF subject may not contain any actual episodes of AF. It is therefore preferable to monitor paroxysmal patients on a regular basis without causing them any discomfort.
- a wearable measurement system that is incorporated in a textile has been developed [I]. If an indication of AF is detected by a suitable method, and preferably confirmed by a cardiologist, a drug administration or other suitable therapeutic intervention can be provided to manage AF treatment.
- the present invention provides a method of detecting atrial fibrillation from an ECG comprising a combination of at least two of the steps of:
- the selected measurements comprise the steps of (a) analyzing R-R interval sequences; and (c) determining the presence or absence of P-waves, as set out above.
- the selected measurements are (a) analyzing R-R intervals and (b) analyzing the signal remaining after QRST cancellation, while in a third embodiment the selected measurements are (b) canceling the QRST portion of the ECG and analyzing the resulting signal; and (c) determining the presence or absence of P-waves.
- all three steps (a), (b) and (c) may also be used in combination. AF detection can only be carried out on clean (i.e. relatively noise free) signals.
- Detecting noisy ECG segments can be done by a combination of threshold detection and the identification of high frequencies which are usually characteristic of noise.
- the initial collection of the data may, for example, be carried out using a wearable belt including three integrated dry electrodes based on carbon loaded rubber.
- This allows the device to be easily worn by a patient, and can be arranged to transmit signals wirelessly, for example by means of bluetooth, to an external PC or other portable computing device.
- the invention also extends to apparatus for use in the detection and/or monitoring of atrial fibrillation, comprising a wearable device incorporating electrodes adapted for contacting the skin, and means for transmitting detected signals to a computer system which is arranged to detect an AF condition by the method of the invention as outlined above.
- the device preferably incorporates a wireless transmission system such as
- the wearable device is integrated into an item of clothing such as a belt or shirt, so that it can be held in suitably good contact with the patient's skin.
- the computer system may be a PC or a hand-held device such as a notebook computer, PDA or "smartphone".
- Fig. 1 is a simplified diagram of a typical ECG signal
- Fig. 2 is an example of an AF episode detected by the methods of the present invention
- Figs. 3a and 3b illustrate the process of noise detection
- Fig. 4 shows a decision tree algorithm
- Fig. 5 illustrates a "sliding window" technique in feature generation
- Fig. 6 illustrates a wearable measuring device.
- the data may, for example, be collected using a wearable measuring device comprising a belt 2 with three integrated dry electrodes 4, and incorporating a miniaturized ECG amplifier indicated at 6, as illustrated in Figure 6.
- the electrodes which are based on carbon-loaded rubber, are fixed into the belt using a thermal moulding process.
- the position of the belt is preferably around the chest to obtain an optimal ECG signal.
- the battery capacity preferably allows measuring for at least 7 days continuously in a typical operation mode.
- Data is transmitted to a PC via the bluetooth interface.
- Figure 2 shows typical example of AF acquired by the wearable system.
- a P wave template can be selected from the normal sinus segment for each patient. Consequently, this P wave template is compared to the P wave candidate before QRS complex. In case of an AF segment, the P wave may disappear, resulting in possible indication of AF.
- the P wave general template is generated from healthy patients.
- noisy segments are rejected, using known methods of noise detection. For example this can be done by identifying high frequency regions of the signal (normally indicative of noise) and applying threshold detection.
- the fiducial points Before the feature extraction itself, as a preliminary step of the ECG signal, the fiducial points must be detected, for example by using the modified Pan-Tompkins algorithm [8].
- the first feature group relates to the RR intervals.
- a measure of the irregularity of the RR intervals can be obtained from the RR interval transition matrix used in [2]. This matrix represents the relative frequency of transitions between intervals whose lengths are either short, regular or long.
- the second feature group is a test for the presence of a P wave.
- the P wave can be observed before QRS complex while in a case of AF, there is no P wave present.
- the P wave detection can be done, for example, using template matching in which a correlation coefficient is used as a dissimilarity measure between the P wave candidate and a template.
- a threshold must be chosen to allow acceptance of very similar beats. In this way, each QRS complex can be labeled as having/not having a preceding P wave.
- the last feature group consists of the frequency domain properties of ECG remainder obtained after QRST cancellation.
- the remainder electrocardiogram after the ventricular component has been removed represents the atrial activity component of the signal.
- Fiducial points of ventricular complexes are marked, preferably using a method based on the algorithm presented by Pan and Tompkins [6]. Basically, the average beat is aligned with the fiducial points of all dominant beat windows and subtracted.
- the absolute powers in the frequency bands of PSD spectrum extending from 10, 20, 30, 40, 60, 80Hz to 125Hz are estimated (e.g. P20 is the summation of the power found in frequency bands between 20 and 125 Hz). Ratios of high frequency (from F to 125Hz) to low frequency (extending from OHz to F Hz) are also calculated. As a percentage they are expressed as:
- the entropy of the remainder may be calculated as well.
- Feature Selection Using the above described methods may result in a large number of features.
- the features are extracted in a sliding window consisting of 30 beats, as shown in Figure 5.
- This approach results in a one-to-one correspondence between features and beats in the stream. In this way, each beat is labeled individually, rather than in groups.
- Classification It is possible to use various different classifiers to analyze the resulting waveforms. For example there are quadratic classifiers, normal densities based linear classifiers or normal densities based quadratic classifiers. There are Bayes normal classifiers, where in the first case one assumes equal covariance matrices resulting in a linear discriminant function (LDC). In the second case the co variances matrices are different for each category resulting in a quadratic discriminant function (QDC).
- LDC linear discriminant function
- classifiers are a k-nearest neighbor classifier (e.g. 3-KNN) or a neural network, such as a back propagation neural network. As an example this may comprise one hidden layer of 10 neuron units and one output neuron unit (10-ANN). Other possibilities include Support Vector Machines (SVM) or a C4.5 decision tree. Analysis of Results
- Feature extraction is preferably performed automatically using a decision tree structure. It can also be performed manually by looking at different scattered plots and statistical parameters such as the correlation matrix.
- two features as an input for classifier are selected, using automatic analysis, which comprise one feature from the R-R interval analysis and one feature from the group of P template matching (number of found P waves in the window of 30 beats long).
- the QRST cancellation implemented in a preferred embodiment of the invention subtracts the mean beat computed for the whole record [5]. When several QT beat morphologies are presented in the signal, the cancellation technique may be inadequate. Due to big differences in even interpersonal ECG morphology two or three beat templates are preferably computed using an unsupervised approach such as hierarchical clustering.
- Atrial fibrillation detection can be reliably achieved using simple features combined with a suitable classifier.
- Most algorithms requires only time or morphology information for AF classification.
- the approach of the present invention combines both methods.
Abstract
A method of detecting atrial fibrillation from an ECG is presented, comprising a combination of at least two of the steps of analyzing R-R interval sequences to produce a measure of the irregularity of the RR interval sequence; canceling the QRST portion of the ECG, and analyzing the resulting signal; and analyzing the ECG signal preceding the QRS complex, to determine the presence or absence of P-waves. This is followed by a step of using a classifier to classify the ECG into one of two classes, namely 'AF present' and 'AF absent', based on a range or ranges of results of the measurement steps which are determined in advance. The invention improves patient monitoring.
Description
Atrial fibrillation detection
This invention relates to the detection of atrial fibrillation, on the basis of ECG readings from electrodes. It may be used for example with readings derived from electrodes incorporated in "wearable" electrode systems.
Atrial fibrillation ("AF") is a common heart arrhythmia with a prevalence of approximately 0.4 to 1% of the general population, and its prevalence increases with age. It is responsible for the highest number of hospital admissions due to arrhythmias, and consequently, it is desirable to be able to monitor the condition of patients, using portable devices which are capable of producing reliable indications of arrhythmia, without producing false positives. Electrocardiograph (ECG) signals show a characteristic pattern of electrical impulses that are generated by the heart. Different waves are identifiable in the ECG signal - the P wave is from the atrial excitation and the QRS complex and T-wave are from the ventricular excitation and relaxation, respectively, as illustrated in Figure 1.
The ST segment usually appears as a straight, level line between the QRS complex and the T wave. Elevated or depressed ST segments may mean the heart muscle is damaged or not receiving enough blood, a sign that a myocardial infarct may have occurred.
The T wave corresponds to the period when the lower heart chambers are relaxing electrically and preparing for their next muscle contraction.
AF is a heart rhythm which is usually characterized by QRS complexes with normal morphology and with irregular arrival times. This can be caused by a diseased atrium which disrupts the normal passage of electrical stimulus from the sinus node through the atrium to the ventricles. One example of AF is depicted in Figure 2.
AF can be either chronic or intermittent. Intermittent AF is referred to as paroxysmal AF. AF is difficult to detect, particularly if it is paroxysmal, since a sample ECG recording from paroxysmal AF subject may not contain any actual episodes of AF. It is therefore preferable to monitor paroxysmal patients on a regular basis without causing them any discomfort. For this purpose a wearable measurement system that is incorporated in a textile has been developed [I]. If an indication of AF is detected by a suitable method, and
preferably confirmed by a cardiologist, a drug administration or other suitable therapeutic intervention can be provided to manage AF treatment.
AF detection is most often based upon R-R analysis. Attempts have been made to detect AF based on R-R interval sequences using a variety of statistical methods [2]. Another indicator for AF is the absence of clear P-waves before the QRS complex. In such cases, it may be possible to diagnose AF on the basis of a lack of regularly occurring P-waves [3], [4]. Another possible approach is to apply QRST cancellation [5], so as to remove the ventricular activity from the signal, and then calculate the power spectrum of the remainder ECG. Because of the difficulty of obtaining consistent results from the various different methods of detection and subsequent analysis of the detected ECGs, it would be preferable to find a more universally-applicable method of detecting and monitoring the condition. Preferably this should combine the best features of existing methods, while appropriately weighting the significance of the detected signals from each one. Accordingly, the present invention provides a method of detecting atrial fibrillation from an ECG comprising a combination of at least two of the steps of:
(a) analyzing R-R interval sequences to produce a measure of the irregularity of the RR interval sequence;
(b) canceling the QRST portion of the ECG, and analyzing the resulting signal; and
(c) analyzing the ECG signal preceding the QRS complex, to determine the presence or absence of P-waves; followed by a step of using a classifier to classify the ECG into one of two classes, namely "AF present" and "AF absent", based on a range or ranges of results of the selected measurement steps which are determined in advance.
Accordingly, in one embodiment of the invention the selected measurements comprise the steps of (a) analyzing R-R interval sequences; and (c) determining the presence or absence of P-waves, as set out above. In another embodiment the selected measurements are (a) analyzing R-R intervals and (b) analyzing the signal remaining after QRST cancellation, while in a third embodiment the selected measurements are (b) canceling the QRST portion of the ECG and analyzing the resulting signal; and (c) determining the presence or absence of P-waves. Of course it will also be appreciated that all three steps (a), (b) and (c) may also be used in combination.
AF detection can only be carried out on clean (i.e. relatively noise free) signals. Detecting noisy ECG segments (to then be discarded from further treatment) can be done by a combination of threshold detection and the identification of high frequencies which are usually characteristic of noise. The initial collection of the data may, for example, be carried out using a wearable belt including three integrated dry electrodes based on carbon loaded rubber. This allows the device to be easily worn by a patient, and can be arranged to transmit signals wirelessly, for example by means of bluetooth, to an external PC or other portable computing device. Accordingly, the invention also extends to apparatus for use in the detection and/or monitoring of atrial fibrillation, comprising a wearable device incorporating electrodes adapted for contacting the skin, and means for transmitting detected signals to a computer system which is arranged to detect an AF condition by the method of the invention as outlined above. The device preferably incorporates a wireless transmission system such as
"Bluetooth".
Preferably, the wearable device is integrated into an item of clothing such as a belt or shirt, so that it can be held in suitably good contact with the patient's skin. The computer system may be a PC or a hand-held device such as a notebook computer, PDA or "smartphone".
Some embodiments of the invention will now be described by way of example with reference to the accompanying drawings in which: Fig. 1 is a simplified diagram of a typical ECG signal;
Fig. 2 is an example of an AF episode detected by the methods of the present invention;
Figs. 3a and 3b illustrate the process of noise detection; Fig. 4 shows a decision tree algorithm; Fig. 5 illustrates a "sliding window" technique in feature generation; and
Fig. 6 illustrates a wearable measuring device.
The data may, for example, be collected using a wearable measuring device comprising a belt 2 with three integrated dry electrodes 4, and incorporating a miniaturized ECG amplifier indicated at 6, as illustrated in Figure 6. The electrodes, which are based on carbon-loaded rubber, are fixed into the belt using a thermal moulding process. The position of the belt is preferably around the chest to obtain an optimal ECG signal. The battery capacity preferably allows measuring for at least 7 days continuously in a typical operation mode. Data is transmitted to a PC via the bluetooth interface. Figure 2 shows typical example of AF acquired by the wearable system.
In case of paroxysmal patients, a P wave template can be selected from the normal sinus segment for each patient. Consequently, this P wave template is compared to the P wave candidate before QRS complex. In case of an AF segment, the P wave may disappear, resulting in possible indication of AF. The P wave general template is generated from healthy patients.
The patient is rested during measurement and if any significant noise is present the noisy segments are rejected, using known methods of noise detection. For example this can be done by identifying high frequency regions of the signal (normally indicative of noise) and applying threshold detection.
This process is illustrated by Figures 3a and 3b. It can be seen that the parts with saturation noise and high frequency noise have been successfully detected. Feature Extraction
Before the feature extraction itself, as a preliminary step of the ECG signal, the fiducial points must be detected, for example by using the modified Pan-Tompkins algorithm [8]. There are three important feature groups used in detection of atrial fibrillation, features using RR interval information, features using P-wave morphology, and features using QRST cancellation. In the preferred embodiment of the invention a combination of features from at least two of these groups is used.
1. The first feature group relates to the RR intervals. A measure of the irregularity of the RR intervals can be obtained from the RR interval transition matrix used in [2]. This matrix represents the relative frequency of transitions between intervals whose lengths are either short, regular or long.
2. The second feature group is a test for the presence of a P wave. In normal sinus rhythm, the P wave can be observed before QRS complex while in a case of AF, there is no P wave present. The P wave detection can be done, for example, using template matching in which a correlation coefficient is used as a dissimilarity measure between the P
wave candidate and a template. A threshold must be chosen to allow acceptance of very similar beats. In this way, each QRS complex can be labeled as having/not having a preceding P wave.
3. Finally, the last feature group consists of the frequency domain properties of ECG remainder obtained after QRST cancellation. The remainder electrocardiogram after the ventricular component has been removed represents the atrial activity component of the signal. Fiducial points of ventricular complexes are marked, preferably using a method based on the algorithm presented by Pan and Tompkins [6]. Basically, the average beat is aligned with the fiducial points of all dominant beat windows and subtracted. The absolute powers in the frequency bands of PSD spectrum extending from 10, 20, 30, 40, 60, 80Hz to 125Hz are estimated (e.g. P20 is the summation of the power found in frequency bands between 20 and 125 Hz). Ratios of high frequency (from F to 125Hz) to low frequency (extending from OHz to F Hz) are also calculated. As a percentage they are expressed as:
Since the AF wave has a random character, the entropy of the remainder may be calculated as well. Feature Selection Using the above described methods may result in a large number of features.
In order to reduce the dimension of the feature space it is possible to use a decision tree algorithm. Preferably the two most significant features are retained by looking at the first levels of the resulting decision tree. One simplified example of the decision tree process is shown in Figure 4 where two features of the R-R interval analysis and the P wave template matching are selected.
The features are extracted in a sliding window consisting of 30 beats, as shown in Figure 5. This approach results in a one-to-one correspondence between features and beats in the stream. In this way, each beat is labeled individually, rather than in groups. Classification It is possible to use various different classifiers to analyze the resulting waveforms.
For example there are quadratic classifiers, normal densities based linear classifiers or normal densities based quadratic classifiers. There are Bayes normal classifiers, where in the first case one assumes equal covariance matrices resulting in a linear discriminant function (LDC). In the second case the co variances matrices are different for each category resulting in a quadratic discriminant function (QDC). Other alternative classifiers are a k-nearest neighbor classifier (e.g. 3-KNN) or a neural network, such as a back propagation neural network. As an example this may comprise one hidden layer of 10 neuron units and one output neuron unit (10-ANN). Other possibilities include Support Vector Machines (SVM) or a C4.5 decision tree. Analysis of Results
Feature extraction is preferably performed automatically using a decision tree structure. It can also be performed manually by looking at different scattered plots and statistical parameters such as the correlation matrix. In one preferred embodiment of the invention, two features as an input for classifier are selected, using automatic analysis, which comprise one feature from the R-R interval analysis and one feature from the group of P template matching (number of found P waves in the window of 30 beats long). The QRST cancellation implemented in a preferred embodiment of the invention subtracts the mean beat computed for the whole record [5]. When several QT beat morphologies are presented in the signal, the cancellation technique may be inadequate. Due to big differences in even interpersonal ECG morphology two or three beat templates are preferably computed using an unsupervised approach such as hierarchical clustering.
In this way atrial fibrillation detection can be reliably achieved using simple features combined with a suitable classifier. Most algorithms requires only time or morphology information for AF classification. The approach of the present invention combines both methods.
REFERENCES:
1. J Muhlsteff, O Such, and R Schmidt, "Wearable approach for continuous ecg - aanndd aaccttiivviittyy ppaattiieenntt--mmoonniittoorriiing," in 26th Annual International Conference of the IEEE EMBS, vol. 1, 2004, pp. 2184-2187.
2. G Moody and R Mark, "A new method for detecting atrial fibrillation using r-r intervals," in Computers in Cardiology, vol. 10, 1983 pp. 227-230.
3. P Stafford, P Denbigh, and R Vincent, "Frequency analysis of the p wave: comparative techniques," in Pace, vol. 18, 1995, pp. 261-270.
4. P Stafford and J Kolvekar, S Cooper, "Signal averaged p wave compared with standard electrocardiography or echocardiography for prediction of atrial fibrillation after coronary bypass grafting," in Heart, vol. 77, 1997, pp. 417-422.
5. J Slocum, A Sahakian, and S Swiryn, "Diagnosis of atrial fibrillation from surface electrocardiograms based on computer-detected atrial activity." Journal of Electrocardiology, vol. 25, no. 1, pp. 1-8, 1992.
6. J Pan and W Tompkins, "A real-time qrs detection algorithm," IEEE Transactions on Biomedical Engineering, vol. 32. 230-236, 1985.
Claims
1. A method of detecting atrial fibrillation from an ECG comprising a combination of at least two of the steps of:
(a) analyzing R-R interval sequences to produce a measure of the irregularity of the RR interval sequence; (b) canceling the QRST portion of the ECG, and analyzing the resulting signal; and
(c) analyzing the ECG signal preceding the QRS complex, to determine the presence or absence of P-waves; followed by a step of using a classifier to classify the ECG into one of two classes, namely "AF present" and "AF absent", based on a range or ranges of results of the measurement steps which are determined in advance.
2. A method of detecting atrial fibrillation according to claim 1, in which two measurement steps are selected, comprising: (a) analyzing R-R interval sequences; and
(c) determining the presence or absence of P-waves.
3. A method of detecting atrial fibrillation according to claim 1, in which two measurement steps are selected, comprising: (a) analyzing R-R interval sequences; and
(b) canceling the QRST portion of the ECG and analyzing the resulting signal.
4. A method of detecting atrial fibrillation according to claim 1 in which two measurement steps are selected, comprising: (b) canceling the QRST portion of the ECG and analyzing the resulting signal; and
(c) determining the presence or absence of P-waves.
5. A method according to any one of claims 1 , 2 or 4 in which the power spectrum of the resulting signal is analyzed after cancellation of the QRST portion.
6. A method according to any one of claims 1 to 5 in which the measurement features for classification are selected in advance using a decision tree algorithm.
7. A method according to any one of the preceding claims in which the classifier comprises a quadratic classifier, a normal densities based quadratic classifier, a normal densities based linear classifier, or a k-nearest neighbor classifier.
8. A method according to one of claims 1 to 6 in which the classifier comprises a neural network, a C4.5 decision tree, or a Support Vector machine (SVM).
9. A method according to any one of claims 1 to 3 in which the R-R feature used as input to the classifier comprises a value derived from an RR interval transition matrix representing the relative frequency of transitions between intervals whose lengths are either short, regular or long.
10. A method according to any one of claims 1, 2 or 4 in which the P-wave feature used as an input to the classifier comprises a value derived using template matching which gives a measurement of dissimilarity measure between the actual P-wave and a template.
11. Apparatus for detecting and/or monitoring atrial fibrillation, comprising a wearable device incorporating electrodes for collecting signals via the skin of a patient, and means for transmitting detected signals to a computer system, the computer system being programmed to analyze the detected signals by a method in accordance with any preceding claim.
12. Apparatus according to claim 11 in which the signal transmitting means comprises a wireless system.
13. Apparatus according to claim 11 or claim 12 in which the wearable device comprises a belt, shirt or harness.
14. Apparatus according to any one of claims 11 to 13 in which the computer system comprises a mobile device which can also be worn, or carried, by the patient.
15. Apparatus according to claim 14 in which the mobile device is adapted to send signals to an external monitoring means.
16. A method of treating a subject known to be prone to atrial fibrillation, comprising: regularly monitoring the incidence of AF using the detection method of any one of claims 1 to 10 and a wearable device according to any one of claims 11 to 15, and controlling therapeutic interventions to the subject in accordance with the results of the monitoring procedure.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP06115092.6 | 2006-06-07 | ||
EP06115092 | 2006-06-07 |
Publications (2)
Publication Number | Publication Date |
---|---|
WO2008007236A2 true WO2008007236A2 (en) | 2008-01-17 |
WO2008007236A3 WO2008007236A3 (en) | 2008-04-24 |
Family
ID=38923608
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/IB2007/052095 WO2008007236A2 (en) | 2006-06-07 | 2007-06-05 | Atrial fibrillation detection |
Country Status (1)
Country | Link |
---|---|
WO (1) | WO2008007236A2 (en) |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2009090581A1 (en) * | 2008-01-14 | 2009-07-23 | Koninklijke Philips Electronics N.V. | Atrial fibrillation monitoring |
US8666483B2 (en) | 2007-10-24 | 2014-03-04 | Siemens Medical Solutions Usa, Inc. | System for cardiac medical condition detection and characterization |
WO2014042618A1 (en) * | 2012-09-11 | 2014-03-20 | Draeger Medical Systems, Inc. | A system and method for detecting a characteristic in an ecg waveform |
US8744559B2 (en) | 2011-08-11 | 2014-06-03 | Richard P. Houben | Methods, systems and devices for detecting atrial fibrillation |
WO2014169595A1 (en) * | 2013-04-18 | 2014-10-23 | 深圳市科曼医疗设备有限公司 | Method and system for arrhythmia analysis |
US9277956B2 (en) | 2011-11-09 | 2016-03-08 | Siemens Medical Solutions Usa, Inc. | System for automatic medical ablation control |
JP2017139483A (en) * | 2010-10-29 | 2017-08-10 | ローレンス リバモア ナショナル セキュリティー, エルエルシー | Method and system for compact efficient laser architecture |
US9775535B2 (en) | 2013-11-08 | 2017-10-03 | Spangler Scientific Llc | Non-invasive prediction of risk for sudden cardiac death |
EP3248542A1 (en) * | 2016-05-27 | 2017-11-29 | Comarch Healthcare Spólka Akcyjna | Method for automatic detection of atrial fibrillation and flutter |
WO2018162957A1 (en) * | 2017-03-10 | 2018-09-13 | Qatar University | Personalized ecg monitoring for early detection of cardiac abnormalities |
CN109691994A (en) * | 2019-01-31 | 2019-04-30 | 英菲泰克(天津)科技有限公司 | A kind of rhythm of the heart analysis method based on electrocardiogram |
CN110680305A (en) * | 2019-10-08 | 2020-01-14 | 深圳邦健生物医疗设备股份有限公司 | Method, device and computer equipment for determining position of migration lead |
CN110895669A (en) * | 2018-09-13 | 2020-03-20 | 大连大学 | Method for constructing atrial fibrillation prediction decision tree |
CN110960207A (en) * | 2019-12-16 | 2020-04-07 | 成都天奥电子股份有限公司 | Tree model-based atrial fibrillation detection method, device, equipment and storage medium |
WO2021196872A1 (en) * | 2020-03-31 | 2021-10-07 | 京东方科技集团股份有限公司 | Measurement method and apparatus for periodic information of biological signal, and electronic device |
US11571162B2 (en) | 2019-05-09 | 2023-02-07 | Tata Consultancy Services Limited | Recurrent neural network architecture based classification of atrial fibrillation using single lead ECG |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20010056245A1 (en) * | 2000-03-15 | 2001-12-27 | Mlynash Michael D. | QRST subtraction using an adaptive template for analysis of TU wave obscured atrial activity |
EP1219237A2 (en) * | 2000-12-28 | 2002-07-03 | GE Medical Systems Information Technologies, Inc. | Atrial fibrillation detection method and apparatus |
US20040068196A1 (en) * | 2002-10-02 | 2004-04-08 | Louis Massicotte | Method and apparatus for trend detection in an electrocardiogram monitoring signal |
WO2005006209A1 (en) * | 2003-07-11 | 2005-01-20 | University College Dublin National University Of Ireland, Dublin | Method and apparatus for detecting paroxysmal atria fibrillation |
-
2007
- 2007-06-05 WO PCT/IB2007/052095 patent/WO2008007236A2/en active Application Filing
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20010056245A1 (en) * | 2000-03-15 | 2001-12-27 | Mlynash Michael D. | QRST subtraction using an adaptive template for analysis of TU wave obscured atrial activity |
EP1219237A2 (en) * | 2000-12-28 | 2002-07-03 | GE Medical Systems Information Technologies, Inc. | Atrial fibrillation detection method and apparatus |
US20040068196A1 (en) * | 2002-10-02 | 2004-04-08 | Louis Massicotte | Method and apparatus for trend detection in an electrocardiogram monitoring signal |
WO2005006209A1 (en) * | 2003-07-11 | 2005-01-20 | University College Dublin National University Of Ireland, Dublin | Method and apparatus for detecting paroxysmal atria fibrillation |
Non-Patent Citations (4)
Title |
---|
DE CHAZAL P ET AL INSTITUTE OF ELECTRICAL AND ELECTRONICS ENGINEERS: "Automated assessment of atrial fibrillation" COMPUTERS IN CARDIOLOGY 2001. ROTTERDAM, THE NETHERLANDS, SEPT. 23 - 26, 2001, NEW YORK, NY : IEEE, US, 23 September 2001 (2001-09-23), pages 117-120, XP010573230 ISBN: 0-7803-7266-2 * |
GRAJA S ET AL: "SVM classification of patients prone to atrial fibrillation" INTELLIGENT SIGNAL PROCESSING, 2005 IEEE INTERNATIONAL WORKSHOP ON SEPT. 1, 2005, PISCATAWAY, NJ, USA,IEEE, 1 September 2005 (2005-09-01), pages 370-374, XP010852368 ISBN: 0-7803-9030-X * |
SLOCUM J ET AL: "DIAGNOSIS OF ATRIAL FIBRILLATION FROM SURFACE ELECTROCARDIOGRAMS BASED ON COMPUTER-DETECTED ATRIAL ACTIVITY" JOURNAL OF ELECTROCARDIOLOGY, vol. 25, no. 1, 1992, pages 1-8, XP002466507 ISSN: 0022-0736 cited in the application * |
SUMMERS, CRAMP, CARSON: "Non-AI Decision Making" 2000, CRC PRESS LLC , BOCA RATON , XP002466595 in "The Biomedical Engineering Handbook: Second Edition", Ed. Joseph D. Bronzino. the whole document * |
Cited By (22)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8666483B2 (en) | 2007-10-24 | 2014-03-04 | Siemens Medical Solutions Usa, Inc. | System for cardiac medical condition detection and characterization |
US8560058B2 (en) | 2008-01-14 | 2013-10-15 | Koninklijke Philips N.V. | Real time atrial fibrillation monitoring |
WO2009090581A1 (en) * | 2008-01-14 | 2009-07-23 | Koninklijke Philips Electronics N.V. | Atrial fibrillation monitoring |
JP2017139483A (en) * | 2010-10-29 | 2017-08-10 | ローレンス リバモア ナショナル セキュリティー, エルエルシー | Method and system for compact efficient laser architecture |
US8744559B2 (en) | 2011-08-11 | 2014-06-03 | Richard P. Houben | Methods, systems and devices for detecting atrial fibrillation |
US9277956B2 (en) | 2011-11-09 | 2016-03-08 | Siemens Medical Solutions Usa, Inc. | System for automatic medical ablation control |
WO2014042618A1 (en) * | 2012-09-11 | 2014-03-20 | Draeger Medical Systems, Inc. | A system and method for detecting a characteristic in an ecg waveform |
US9724007B2 (en) | 2012-09-11 | 2017-08-08 | Draeger Medical Systems, Inc. | System and method for detecting a characteristic in an ECG waveform |
WO2014169595A1 (en) * | 2013-04-18 | 2014-10-23 | 深圳市科曼医疗设备有限公司 | Method and system for arrhythmia analysis |
US9775535B2 (en) | 2013-11-08 | 2017-10-03 | Spangler Scientific Llc | Non-invasive prediction of risk for sudden cardiac death |
US11045135B2 (en) | 2013-11-08 | 2021-06-29 | Spangler Scientific Llc | Non-invasive prediction of risk for sudden cardiac death |
US11839497B2 (en) | 2013-11-08 | 2023-12-12 | Spangler Scientific Llc | Non-invasive prediction of risk for sudden cardiac death |
US10226196B2 (en) | 2013-11-08 | 2019-03-12 | Spangler Scientific Llc | Non-invasive prediction of risk for sudden cardiac death |
EP3248542A1 (en) * | 2016-05-27 | 2017-11-29 | Comarch Healthcare Spólka Akcyjna | Method for automatic detection of atrial fibrillation and flutter |
US10856763B2 (en) | 2017-03-10 | 2020-12-08 | Qatar University | Personalized ECG monitoring for early detection of cardiac abnormalities |
WO2018162957A1 (en) * | 2017-03-10 | 2018-09-13 | Qatar University | Personalized ecg monitoring for early detection of cardiac abnormalities |
CN110895669A (en) * | 2018-09-13 | 2020-03-20 | 大连大学 | Method for constructing atrial fibrillation prediction decision tree |
CN109691994A (en) * | 2019-01-31 | 2019-04-30 | 英菲泰克(天津)科技有限公司 | A kind of rhythm of the heart analysis method based on electrocardiogram |
US11571162B2 (en) | 2019-05-09 | 2023-02-07 | Tata Consultancy Services Limited | Recurrent neural network architecture based classification of atrial fibrillation using single lead ECG |
CN110680305A (en) * | 2019-10-08 | 2020-01-14 | 深圳邦健生物医疗设备股份有限公司 | Method, device and computer equipment for determining position of migration lead |
CN110960207A (en) * | 2019-12-16 | 2020-04-07 | 成都天奥电子股份有限公司 | Tree model-based atrial fibrillation detection method, device, equipment and storage medium |
WO2021196872A1 (en) * | 2020-03-31 | 2021-10-07 | 京东方科技集团股份有限公司 | Measurement method and apparatus for periodic information of biological signal, and electronic device |
Also Published As
Publication number | Publication date |
---|---|
WO2008007236A3 (en) | 2008-04-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2008007236A2 (en) | Atrial fibrillation detection | |
Satija et al. | A review of signal processing techniques for electrocardiogram signal quality assessment | |
Babaeizadeh et al. | Improvements in atrial fibrillation detection for real-time monitoring | |
Ge et al. | Cardiac arrhythmia classification using autoregressive modeling | |
Chiu et al. | Using correlation coefficient in ECG waveform for arrhythmia detection | |
US6171256B1 (en) | Method and apparatus for detecting a condition associated with acute cardiac ischemia | |
Faezipour et al. | A patient-adaptive profiling scheme for ECG beat classification | |
JP4386235B2 (en) | Method and apparatus for sequential comparison of electrocardiograms | |
JP7429371B2 (en) | Method and system for quantifying and removing asynchronous noise in biophysical signals | |
de Chazal et al. | Automatic classification of ECG beats using waveform shape and heart beat interval features | |
WO2014042618A1 (en) | A system and method for detecting a characteristic in an ecg waveform | |
JP2009160410A (en) | Method of fetal and maternal ecg identification across multiple epochs | |
Kumar et al. | Detection and classification of ECG noises using decomposition on mixed codebook for quality analysis | |
Altan et al. | A multistage deep belief networks application on arrhythmia classification | |
Henzel et al. | Atrial fibrillation episodes detection based on classification of heart rate derived features | |
CN110751131B (en) | Arrhythmia detection device | |
Demirel et al. | Energy-efficient real-time heart monitoring on edge–fog–cloud internet of medical things | |
Zhang et al. | Detection of atrial fibrillation from variable-duration ECG signal based on time-adaptive densely network and feature enhancement strategy | |
Lee et al. | A real-time abnormal beat detection method using a template cluster for the ECG diagnosis of IoT devices | |
Xie et al. | A signal quality assessment–based ECG waveform delineation method used for wearable monitoring systems | |
Bhoi et al. | QRS Complex Detection and Analysis of Cardiovascular Abnormalities: A Review. | |
Mohammadzadeh-Asl et al. | Neural network based arrhythmia classification using heart rate variability signal | |
Kumar et al. | Classification of normal and abnormal ECG signals using support vector machine and Fourier decomposition method | |
Castro et al. | Data quality assessment of capacitively-coupled ECG signals | |
Mendez et al. | Automatic detection of sleep macrostructure based on bed sensors |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
NENP | Non-entry into the national phase |
Ref country code: RU |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 07825795 Country of ref document: EP Kind code of ref document: A2 |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 07825795 Country of ref document: EP Kind code of ref document: A2 |