Motion and noise artifact detection for ECG data
View Patent ↗Technologies are provided herein for real-time detection of motion and noise (MN) artifacts in electrocardiogram signals recorded by electrocardiography devices. Specifically, the present disclosure provides techniques for increasing the accuracy of identifying paroxysmal atrial fibrillation (AF) rhythms, which are often measured via such devices. According to aspects of the present disclosure, a method for detecting MN artifacts in an electrocardiogram (ECG) recording includes receiving an ECG segment and decomposing the received ECG segment into a sum of intrinsic mode functions. The intrinsic mode functions associated with MN artifacts present within the ECG segment are then isolated. The method further includes determining randomness and variability characteristic values associated with the isolated intrinsic mode functions and comparing the randomness and variability characteristic values to threshold randomness and variability characteristic values. If the randomness and variability characteristic values exceed the threshold characteristic values, the ECG signal is determined to include MN artifacts.
1. A method for detecting motion and noise artifacts in an electrocardiogram (ECG) recording, comprising:
receiving an ECG segment;
decomposing the received ECG segment into a sum of intrinsic mode functions;
isolating intrinsic mode functions associated with motion and noise artifacts present within the ECG segment;
determining randomness and variability characteristic values associated with the isolated intrinsic mode functions;
comparing the randomness and variability characteristic values to threshold randomness and variability characteristic values; and
determining that the received ECG segment includes motion and noise artifacts if the randomness and variability characteristic values exceed the threshold randomness and variability characteristic values.
2. The method of claim 1 , wherein the ECG segment is recorded by an ambulatory ECG recording device.
3. The method of claim 1 , wherein decomposing the received ECG segment into a sum of intrinsic mode functions comprises performing empirical mode decomposition (EMD) on the received ECG segment.
4. The method of claim 1 , wherein isolating intrinsic mode functions associated with motion and noise artifacts present within the ECG segment further comprises:
identifying first order intrinsic mode functions;
squaring the identified first order intrinsic mode functions; and
normalizing the squared first order intrinsic mode functions.
5. The method of claim 4 , further comprising discarding squared first order intrinsic mode functions that have peaks that exceed a predefined threshold value.
6. The method of claim 4 , further comprising extracting the low level sequences from the normalized squared first order intrinsic mode functions.
7. The method of claim 4 , further comprising extracting the low noise level threshold sequences from the normalized squared first order intrinsic mode function components.
8. The method of claim 1 , wherein determining randomness and variability characteristic values associated with the first order intrinsic mode function components of the received ECG segment comprises determining a Shannon Entropy, a mean, and a root mean square of successive RR differences (RMSSD) of extracted components of the received ECG segment.
9. The method of claim 8 , wherein comparing the randomness and variability characteristics to threshold randomness and variability characteristic values comprises comparing the determined Shannon Entropy, mean, and RMSSD to threshold values of Shannon Entropy, mean, and RMSSD, respectively.
10. The method of claim 1 , wherein determining randomness and variability characteristic values associated with the first order intrinsic mode function components of the received ECG segment comprises determining a Shannon Entropy, a mean, and a root mean square of successive RR differences (RMSSD) of extracted components of the received ECG segment.
11. The method of claim 10 , wherein comparing the randomness and variability characteristics to threshold randomness and variability characteristic values comprises comparing the determined Shannon Entropy, mean, and variance to threshold values of Shannon Entropy, mean, and variance, respectively.
12. A system for detecting motion and noise artifacts in an electrocardiogram (ECG) recording, comprising:
a processor;
a memory; and
a real-time motion and noise artifact detection application, comprising:
an ECG reception module that receives an ECG segment,
an empirical mode decomposition module that decomposes the received ECG segment into a sum of intrinsic mode functions,
a signal processing module that isolates intrinsic mode functions associated with motion and noise artifacts present within the ECG segment,
at least one characteristic calculation module that determines randomness and variability characteristic values associated with the isolated intrinsic mode functions, and
a ECG determination module that compares randomness and variability characteristic values to threshold randomness and variability characteristic values and determines that the received ECG segment includes motion and noise artifacts if the randomness and variability characteristic values exceed the threshold randomness and variability characteristic values.
13. The system of claim 12 , wherein the ECG reception module receives ECG signals from an electrocardiogram recording device.
14. The system of claim 12 , wherein the signal processing module that isolates intrinsic mode functions associated with motion and noise artifacts present within the ECG segment comprises:
an identification module that identifies first order intrinsic mode functions
a squaring module that squares the identified first order intrinsic mode functions; and
a scaling module that normalizes the squared first order intrinsic mode functions.
15. The system of claim 12 , wherein the at least characteristic calculation module comprises:
a Shannon Entropy calculation module that calculates a Shannon Entropy of the isolated intrinsic mode functions;
a mean calculation module that calculates a mean of the isolated intrinsic mode functions; and
a RMSSD calculation module that calculates a RMSSD of the isolated intrinsic mode functions.
16. The system of claim 12 , wherein the at least characteristic calculation module comprises:
a Shannon Entropy calculation module that calculates a Shannon Entropy of the isolated intrinsic mode functions;
a mean calculation module that calculates a mean of the isolated intrinsic mode functions; and
a variance calculation module that calculates a variance of the isolated intrinsic mode functions.
17. A computer-readable non-transitory storage medium, having computer-executable instructions stored thereon, which when executed by a processor, cause the computer to:
receive an ECG segment;
decompose the received ECG segment into a sum of intrinsic mode functions;
isolate intrinsic mode functions associated with motion and noise artifacts present within the ECG segment;
determine randomness and variability characteristic values associated with the isolated intrinsic mode functions;
compare the randomness and variability characteristic values to threshold randomness and variability characteristic values; and
determine that the received ECG segment includes motion and noise artifacts if the randomness and variability characteristic values exceed the threshold randomness and variability characteristic values.
18. The computer-readable non-transitory storage medium of claim 17 , wherein decomposing the received ECG segment into a sum of intrinsic mode functions comprises performing empirical mode decomposition (EMD) on the received ECG segment.
19. The computer-readable non-transitory storage medium of claim 17 , wherein isolating intrinsic mode functions associated with motion and noise artifacts present within the ECG segment further comprises:
identifying first order intrinsic mode functions;
squaring the identified first order intrinsic mode functions; and
normalizing the squared first order intrinsic mode functions.
20. The computer-readable non-transitory storage medium of claim 19 , further comprising discarding squared first order intrinsic mode functions that have peaks that exceed a predefined threshold value.
21. The computer-readable non-transitory storage medium of claim 19 , further comprising extracting the low level sequences from the normalized squared first order intrinsic mode functions.
22. The computer-readable non-transitory storage medium of claim 19 , further comprising extracting the low noise level threshold sequences from the normalized squared first order intrinsic mode function components.
23. The computer-readable non-transitory storage medium of claim 17 , wherein determining randomness and variability characteristic values associated with the first order intrinsic mode function components of the received ECG segment comprises determining a Shannon Entropy, a mean, and a root mean square of successive RR differences (RMSSD) of extracted components of the received ECG segment.
24. The computer-readable non-transitory storage medium of claim 23 , wherein comparing the randomness and variability characteristics to threshold randomness and variability characteristic values comprises comparing the determined Shannon Entropy, mean, and RMSSD to threshold values of Shannon Entropy, mean, and RMSSD, respectively.
25. The computer-readable non-transitory storage medium of claim 17 , wherein determining randomness and variability characteristic values associated with the first order intrinsic mode function components of the received ECG segment comprises determining a Shannon Entropy, a mean, and a root mean square of successive RR differences (RMSSD) of extracted components of the received ECG segment.
26. The computer-readable non-transitory storage medium of claim 25 , wherein comparing the randomness and variability characteristics to threshold randomness and variability characteristic values comprises comparing the determined Shannon Entropy, mean, and variance to threshold values of Shannon Entropy, mean, and variance, respectively.