Automatic method to delineate or categorize an electrocardiogram
Disclosed is a method for computerizing delineation and/or multi-label classification of an ECG signal, including: applying a neural network to the ECG, labelling the ECG, and optionally displaying the labels according to time with the ECG signal.
1. A system for detecting anomalies in an electrocardiogram (ECG) signal corresponding to a patient, the system comprising at least one server and at least one processor configured to execute instructions to:
receive the ECG signal sampled from the patient over a plurality of time points;
determine delineation scores at each time point of the plurality of time points of at least a portion of the ECG signal, the delineation scores corresponding to multiple wave types at each time point;
determine anomaly scores based on the at least a portion of an ECG signal and the delineation scores, the anomaly scores corresponding to multiple anomaly types;
determine that an anomaly corresponding to one of the multiple anomaly types is present in the ECG signal based on at least one of the anomaly scores corresponding to a first time point of the plurality of time points; and
generate information for display comprising a graphical representation of the at least a portion of the ECG signal and a visual indication of the anomaly corresponding to the graphical representation of the at least a portion of the ECG signal at the first time point of the plurality of time points.
2. The system of claim 1 , wherein determining that the anomaly is present in the ECG signal comprises determining that atrial fibrillation is present in the ECG signal.
3. The system of claim 1 , wherein the visual indication of the anomaly is aligned with a location of the anomaly on the graphical representation of the at least a portion of the ECG signal.
4. The system of claim 1 , wherein the at least one processor is further configured to execute instructions to:
determine that a second anomaly corresponding to a second one of the multiple anomaly types and different from the first anomaly is present in the ECG signal based on at least a second one of the anomaly scores corresponding to a second time point of the plurality of time points; and
generate information for display comprising a second visual indication of the second anomaly corresponding to the graphical representation of the at least a portion of the ECG signal at the second time point of the plurality of time points.
5. The system of claim 1 , wherein generating information for display comprises generating delineation information highlighting on the graphical representation of the at least a portion of the ECG signal at least one wave type of the multiple wave types.
6. The system of claim 1 , wherein the at least one processor is further configured to execute instructions to cause display of the graphical representation of the at least a portion of the ECG signal and the visual indication of the anomaly.
7. The system of claim 1 , wherein the at least one processor is further configured to execute instructions to determine a threshold value indicative of the presence of the anomaly.
8. The system of claim 7 , wherein determining that the anomaly is present in the ECG signal comprises comparing the anomaly scores to the threshold value and determining that the at least one of the anomaly scores satisfies the threshold value.
9. The system of claim 1 , wherein determining delineation scores comprises applying the at least a portion of the ECG signal to a first neural network.
10. The system of claim 9 , wherein determining anomaly scores comprises applying the at least a portion of the ECG signal and the delineation scores to a second neural network.
11. A computerized method for detecting anomalies in an electrocardiogram (ECG) signal corresponding to a patient, the computerized method comprising:
receiving an ECG signal sampled from a patient over a plurality of time points;
determining delineation scores at each time point of the plurality of time points of at least a portion of the ECG signal, the delineation scores corresponding to multiple wave types at each time point;
determining anomaly scores based on the at least a portion of an ECG signal and the delineation scores, the anomaly scores corresponding to multiple anomaly types;
determining that an anomaly corresponding to one of the multiple anomaly types is present in the ECG signal based on at least one of the anomaly scores corresponding to a first time point of the plurality of time points; and
generating information for display comprising a graphical representation of the at least a portion of the ECG signal and a visual indication of the anomaly corresponding to the graphical representation of the at least a portion of the ECG signal at the first time point of the plurality of time points.
12. The computerized method of claim 11 , wherein determining that the anomaly is present in the ECG signal comprises determining that atrial fibrillation is present in the ECG signal.
13. The computerized method of claim 11 , wherein the visual indication of the anomaly is aligned with a location of the anomaly on the graphical representation of the at least a portion of the ECG signal.
14. The computerized method of claim 11 , further comprising:
determining that a second anomaly corresponding to a second one of the multiple anomaly types and different from the first anomaly is present in the ECG signal based on at least a second one of the anomaly scores corresponding to a second time point of the plurality of time points; and
generating information for display comprising a second visual indication of the second anomaly corresponding to the graphical representation of the at least a portion of the ECG signal at the second time point of the plurality of time points.
15. The computerized method of claim 11 , wherein generating information for display comprises generating delineation information highlighting on the graphical representation of the at least a portion of the ECG signal at least one wave type of the multiple wave types.
16. The computerized method of claim 11 , further comprising causing display of the graphical representation of the at least a portion of the ECG signal and the visual indication of the anomaly.
17. The computerized method of claim 11 , further comprising determining a threshold value indicative of the presence of the anomaly.
18. The computerized method of claim 17 , wherein determining that the anomaly is present in the ECG signal comprises comparing the anomaly scores to the threshold value and determining that the at least one of the anomaly scores satisfies the threshold value.
19. The computerized method of claim 11 , wherein determining delineation scores comprises applying the at least a portion of the ECG signal to a first neural network.
20. The computerized method of claim 19 , wherein determining anomaly scores comprises applying the at least a portion of the ECG signal and the delineation scores to a second neural network.
21. A programmed routine for use with a computerized system for detecting abnormalities in an electrocardiogram (ECG) signal obtained from a patient, the programmed routine comprising instructions that when executed:
receive an ECG signal sampled from a patient over a plurality of time points;
determine delineation scores at each time point of the plurality of time points of at least a portion of the ECG signal, the delineation scores corresponding to multiple wave types at each time point;
determine anomaly scores based on the at least a portion of an ECG signal and the delineation scores, the anomaly scores corresponding to multiple anomaly types;
determine that an anomaly corresponding to one of the multiple anomaly types is present in the ECG signal based on at least one of the anomaly scores corresponding to a first time point of the plurality of time points; and
generate information for display comprising a graphical representation of the at least a portion of the ECG signal and a visual indication of the anomaly corresponding to the graphical representation of the at least a portion of the ECG signal at the first time point of the plurality of time points.
22. The programmed routine of claim 21 , wherein determining that the anomaly is present in the ECG signal comprises determining that atrial fibrillation is present in the ECG signal.
23. The programmed routine of claim 21 , wherein the visual indication of the anomaly is aligned with a location of the anomaly on the graphical representation of the at least a portion of the ECG signal.
24. The programmed routine of claim 21 , wherein the programmed routine further comprises instructions that when executed:
determine that a second anomaly corresponding to a second one of the multiple anomaly types and different from the first anomaly is present in the ECG signal based on at least a second one of the anomaly scores corresponding to a second time point of the plurality of time points; and
generate information for display comprising a second visual indication of the second anomaly corresponding to the graphical representation of the at least a portion of the ECG signal at the second time point of the plurality of time points.
25. The programmed routine of claim 21 , wherein generating information for display comprises generating delineation information highlighting on the graphical representation of the at least a portion of the ECG signal at least one wave type of the multiple wave types.
26. The programmed routine of claim 21 , wherein the programmed routine further comprises instructions that when executed cause display of the graphical representation of the at least a portion of the ECG signal and the visual indication of the anomaly.
27. The programmed routine of claim 21 , wherein the programmed routine further comprises instructions that when executed determine a threshold value indicative of the presence of the anomaly.
28. The programmed routine of claim 27 , wherein determining that the anomaly is present in the ECG signal comprises comparing the anomaly scores to the threshold value and determining that the at least one of the anomaly scores satisfies the threshold value.
29. The programmed routine of claim 21 , wherein determining delineation scores comprises applying the at least a portion of the ECG signal to a first neural network.
30. The programmed routine of claim 29 , wherein determining anomaly scores comprises applying the at least a portion of the ECG signal and the delineation scores to a second neural network.