Automatic method to delineate or categorize an electrocardiogram

• US 10,426,364 B2
• Filed: 10/27/2015
• Issued: 10/01/2019
• Est. Priority Date: 10/27/2015
• Status: Active Grant

• Current Assignee: Koninklijke Philips N.V.
• Original Assignee: Cardiologs Technologies SAS (Koninklijke Philips N.V.)
• Inventors: Rapin, Jeremy, Li, Jia, Massias, Mathurin
• Primary Examiner(s): Layno, Carl H
• Assistant Examiner(s): Piateski, Erin M
• Application Number:
  US14/924,239
• Publication Number:
  US 20170112401A1
• Time in Patent Office: 1,435 Days
• Field of Search: 600 16
• CPC Class Codes: A61B 5/316 Modalities, i.e. specific diagnostic methods
• CPC Class Codes: A61B 5/349 Detecting specific parameters of the electrocardiograph cycle
• CPC Class Codes: A61B 5/7203 for noise prevention, reduction or removal
• CPC Class Codes: A61B 5/7264 Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems neural networks per se G06N3/00; expert systems per se G06N5/00
• CPC Class Codes: A61B 5/7267 involving training the classification device
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems
• CPC Class Codes: G16H 50/30 for calculating health indices; for individual health risk assessment

Abstract

A method for computerizing delineation and/or multi-label classification of an ECG signal, includes: applying a neural network to the ECG whereby labelling the ECG, and optionally displaying the labels according to time, optionally with the ECG signal.

Claims

1. A method for computerizing delineation and multi-label classification of an ECG signal, the ECG signal represented by a multiplicity of ECG data points, the method comprising applying a convolutional neural network to the ECG signal, wherein the convolutional neural network:

reads each one of the multiplicity of ECG data points;

analyzes temporally each one of the multiplicity of ECG data points, each one of the multiplicity of ECG data points corresponding to a time point;

assigns to each one of the multiplicity of EGG data points a score for at least two of a P-wave, a QRS complex, a T-wave, or no wave; and

allocates to each time point an absence, a single, or a multiplicity of corresponding waves based on the scores assigned to each one of the multiplicity of ECG data points.

2. The method according to claim 1, further comprising, prior to applying the convolutional neural network, denoising and removing a baseline of the ECG signal.

3. The method according to claim 1, further comprising, after applying the convolutional neural network, determining a beginning and an end for one or more waves that comprise the ECG signal.

4. The method according to claim 1, further comprising displaying labels with the ECG signal.

5. The method according to claim 1, wherein the convolutional neural network further detects one or more anomalies, and computes a score for each of the one or more anomalies.

6. The method according to claim 5, further comprising displaying a list of the one or more detected anomalies.

7. The method according claim 5, further comprising generating at least one label corresponding to the score for each of the one or more anomalies.

8. A programmed routine for use with a computer for delineating and classifying an ECG signal, the ECG signal represented by a multiplicity of ECG data points, the programmed routine comprising a convolutional neural network that:

reads each one of the multiplicity of ECG data points;

analyzes temporally each one of the multiplicity of ECG data points, each one of the multiplicity of ECG data points corresponding to a time point;

assigns to each one of the multiplicity of ECG data points a score for at least two of a P-wave, a QRS complex, a T-wave, or no wave; and

allocates to each time point an absence, a single, or a multiplicity of corresponding waves based on the scores assigned to each one of the multiplicity of ECG data points.

9. The programmed routine of claim 8, further comprising, prior to applying the convolutional neural network, denoising and removing a baseline of the ECG signal.

10. The programmed routine of claim 8, further comprising, after applying the convolutional neural network, determining a beginning and an end for one or more waves that comprise the ECG signal.

11. The programmed routine of claim 8, further comprising a display routine that displays labels with the ECG signal.

12. The programmed routine of claim 8, wherein the programmed routine, comprising the convolutional neural network further detects one or more anomalies, and computes a score for each of the one or more anomalies.

13. The programmed routine of claim 12, further comprising a display routine that displays a list of the one or more detected anomalies.

14. The programmed routine of claim 12, further comprising a label routine that generates at least one label corresponding to the score for each of the one or more anomalies.