Cascaded binary classifier for identifying rhythms in a single-lead electrocardiogram (ECG) signal

Patent metadata

Field	Value
Publication number	US-10750968-B2
Application number	US-201815883712-A
Country	US
Kind code	B2
Filing date	Jan 30, 2018
Priority date	Sep 19, 2017
Publication date	Aug 25, 2020
Grant date	Aug 25, 2020

Abstract

Current technologies analyze electrocardiogram (ECG) signals for a long duration, which is not always a practical scenario. Moreover the current scenarios perform a binary classification between normal and Atrial Fibrillation (AF) only, whereas there are many abnormal rhythms apart from AF. Conventional systems/methods have their own limitations and may tend to misclassify ECG signals, thereby resulting in an unbalanced multi-label classification problem. Embodiments of the present disclosure provide systems and methods that are robust and more efficient for classifying rhythms for example, normal, AF, other abnormal rhythms and noisy ECG recordings by implementing a spectrogram based noise removal that obtains clean ECG signal from an acquired single-lead ECG signal, an optimum feature selection at each layer of classification that selects optimum features from a pool of extracted features, and a multi-layer cascaded binary classifier that identifies rhythms in the clean ECG signal at each layer of the classifier.

First claim

What is claimed is:

1. A processor-implemented method, comprising:
processor implemented method, comprising:
acquiring, via one or more hardware processors, a single-lead electrocardiogram (ECG) signal that is recorded for a predefined time interval;
applying in real-time, via the one or more hardware processors, a spectrogram based noisy data removal technique on the acquired single-lead ECG signal to obtain a clean ECG signal, wherein the step of applying, via the one or more hardware processors, a spectrogram based noisy data removal technique on the acquired single-lead ECG signal comprises:
dividing the acquired single-lead ECG signal into a plurality of windows;
computing a spectrogram of each of the plurality of windows;
performing a comparison of the computed spectrogram of each of the plurality of windows with a dynamically computed threshold;
determining noise in at least a subset of the plurality of windows based on the comparison; and
extracting the clean ECG signal based on the determined noise in the at least a subset of the plurality of windows;

extracting one or more features from the clean ECG signal;
selecting, using an optimum feature selection technique, one or more optimum features from the one or more extracted features, wherein the optimum feature selection technique is at least one of a minimum redundancy maximum relevancy (mRMR) technique and a Maximal Information Coefficient (MIC) technique; and
identifying based on the selected one or more optimum features, using a binary cascade classifier, at least one of one or more normal rhythms, a first set of abnormal rhythms, and a second set of abnormal rhythms in at least one of the single-lead electrocardiogram (ECG) signal, and the clean ECG signal.

2. The processor implemented method of claim 1 , wherein noise is determined in the at least a subset of the plurality of windows when each window from the subset has power that is greater than a threshold power.

3. The processor implemented method of claim 1 , wherein the dynamically computed threshold is based on a signal to noise ratio (SNR).

4. A system comprising:
a memory storing instructions;
one or more communication interfaces; and
one or more hardware processors coupled to the memory via the one or more communication interfaces, wherein the one or more hardware processors are configured by the instructions to:
acquire a single-lead electrocardiogram (ECG) signal that is recorded for a predefined time interval;
apply in real-time, a spectrogram based noisy data removal technique on the acquired single-lead ECG signal to obtain a clean ECG signal, wherein the clean ECG signal is extracted by applying the spectrogram based noisy data removal technique on the acquired single-lead ECG signal by:
dividing the acquired single-lead ECG signal into a plurality of windows;
computing a spectrogram of each of the plurality of windows;
performing a comparison of the computed spectrogram of each of the plurality of windows with a dynamically computed threshold;
determining noise in at least a subset of the plurality of windows based on the comparison; and
extracting the clean ECG signal based on the determined noise in the at least a subset of the plurality of windows;

extract one or more features from the clean ECG signal;
select, using an optimum feature selection technique, one or more optimum features from the one or more extracted features, wherein the optimum feature selection technique is at least one of a minimum redundancy maximum relevancy (mRMR) technique and a Maximal Information Coefficient (MIC) technique; and
identify based on the selected one or more optimum features, using a binary cascade classifier, at least one of one or more normal rhythms, a first set of abnormal rhythms, and a second set of abnormal rhythms in at least one of the single-lead electrocardiogram (ECG) signal, and the clean ECG signal.

5. The system of claim 4 , wherein noise is determined in the at least a subset of the plurality of windows when each window from the subset have power that is greater than a threshold power.

6. The system of claim 4 , wherein the dynamically computed threshold is based on a signal to noise ratio (SNR).

7. One or more non-transitory machine readable information storage mediums comprising one or more instructions which when executed by one or more hardware processors causes:
acquiring, via the one or more hardware processors, a single-lead electrocardiogram (ECG) signal that is recorded for a predefined time interval;
applying in real-time, via the one or more hardware processors, a spectrogram based noisy data removal technique on the acquired single-lead ECG signal to obtain a clean ECG signal, wherein the step of applying, via the one or more hardware processors, a spectrogram based noisy data removal technique on the acquired single-lead ECG signal comprises:
dividing the acquired single-lead ECG signal into a plurality of windows;
computing a spectrogram of each of the plurality of windows;
performing a comparison of the computed spectrogram of each of the plurality of windows with a dynamically computed threshold;
determining noise in at least a subset of the plurality of windows based on the comparison; and
extracting the clean ECG signal based on the determined noise in the at least a subset of the plurality of windows;

extracting one or more features from the clean ECG signal;
selecting, using an optimum feature selection technique, one or more optimum features from the one or more extracted features, wherein the optimum feature selection technique is at least one of a minimum redundancy maximum relevancy (mRMR) technique and a Maximal Information Coefficient (MIC) technique; and
identifying based on the selected one or more optimum features, using a binary cascade classifier, at least one of one or more normal rhythms, a first set of abnormal rhythms, and a second set of abnormal rhythms in at least one of the single-lead electrocardiogram (ECG) signal, and the clean ECG signal.

8. The one or more non-transitory machine readable information storage mediums of claim 7 , wherein noise is determined in the at least a subset of the plurality of windows when each window from the subset has power that is greater than a threshold power.

9. The one or more non-transitory machine readable information storage mediums of claim 7 , wherein the dynamically computed threshold is based on a signal to noise ratio (SNR).

Assignees

• Tata Consultancy Services Ltd

Inventors

• Datta Shreyasi
• Puri Chetanya
• Mukherjee Ayan
• Banerjee Rohan
• Choudhury Anirban Dutta
• Ukil Arijit
• Bandyopadhyay Soma
• Pal Arpan
• Khandelwal Sundeep
• Singh Rituraj

Classifications

• G06F18/2148 Primary

  characterised by the process organisation or structure, e.g. boosting cascade · CPC title

• G06F18/2431

  Multiple classes · CPC title

• A61B5/349 Primary

  Detecting specific parameters of the electrocardiograph cycle · CPC title

• A61B5/366

  Detecting abnormal QRS complex, e.g. widening · CPC title

• A61B5/352

  Detecting R peaks, e.g. for synchronising diagnostic apparatus; Estimating R-R interval · CPC title