Opening the black box: interpretability of machine learning algorithms in electrocardiography

Recent studies have suggested that cardiac abnormalities can be detected from the electrocardiogram (ECG) using deep machine learning (DL) models. However, most DL algorithms lack interpretability, since they do not provide any justification for their decisions. In this study, we designed two new frameworks to interpret the classification results of DL algorithms trained for 12-lead ECG classification. The frameworks allow us to highlight not only the ECG samples that contributed most to the classification, but also which between the P-wave, QRS complex and T-wave, hereafter simply called ‘waves’, were the most relevant for the diagnosis. The frameworks were designed to be compatible with any DL model, including the ones already trained. The frameworks were tested on a selected Deep Neural Network, trained on a publicly available dataset, to automatically classify 24 cardiac abnormalities from 12-lead ECG signals. Experimental results showed that the frameworks were able to detect the most relevant ECG waves contributing to the classification. Often the network relied on portions of the ECG which are also considered by cardiologists to detect the same cardiac abnormalities, but this was not always the case. In conclusion, the proposed frameworks may unveil whether the network relies on features which are clinically significant for the detection of cardiac abnormalities from 12-lead ECG signals, thus increasing the trust in the DL models. This article is part of the theme issue ‘Advanced computation in cardiovascular physiology: new challenges and opportunities’.

Download Full-text

Quantification of Interpatient 12-lead ECG Variabilities within a Healthy Cohort

Current Directions in Biomedical Engineering ◽

10.1515/cdbme-2020-3127 ◽

2020 ◽

Vol 6 (3) ◽

pp. 493-496

Author(s):

Claudia Nagel ◽

Nicolas Pilia ◽

Axel Loewe ◽

Olaf Dössel

Keyword(s):

Healthy Subjects ◽

Large Scale ◽

Machine Learning Algorithms ◽

P Wave ◽

Qrs Complex ◽

T Wave ◽

Healthy Cohort ◽

Electrocardiogram Ecg ◽

Morphology Changes

AbstractThe morphology of the electrocardiogram (ECG) varies among different healthy subjects due to anatomical and structural reasons, such as for example the shape of the heart geometry or the position and size of surrounding organs in the torso. Knowledge about these ECG morphology changes could be used to parameterize electrophysiological simulations of the human heart. In this work, we detected the boundaries of ECG waveforms, i.e. the P-wave, the QRS-complex and the T-wave, in 12- lead ECGs from 918 healthy subjects in the Physionet Computing in Cardiology Challenge 2020 Database with the IBT openECG toolbox. Subsequently, we obtained the onset, the peak and the offset of each P-wave, QRS-complex and T-wave in the signal. In this way, the duration of the P-wave, the QRScomplex and the T-wave, the PQ-, RR- and the QT-interval as well as the amplitudes of the P-wave, the Q-, R- and Speak and the T-wave in each lead were extracted from the 918 healthy ECGs. Their statistical distributions and correlation between each other were assessed. The highest variabilities among the 918 healthy subject were found for the RR interval and the amplitudes of the QRScomplex. The highest correlation was observed for feature pairs that represent the same feature in different leads. Especially the R-peak amplitudes showed a strong correlation across different leads. The calculated feature distributions can be used to optimize the parameters of populations of cardiac electrophysiological models. In this way, realistic in-silico generated surface ECGs can be simulated in large scale and could be used as input data for machine learning algorithms for a classification of cardiovascular diseases.

Download Full-text

Electrocardiogram Changes with Acute Alcohol Intoxication: A Systematic Review

The Open Cardiovascular Medicine Journal ◽

10.2174/1874192401812010001 ◽

2018 ◽

Vol 12 (1) ◽

pp. 1-6 ◽

Cited By ~ 4

Author(s):

Hitesh Raheja ◽

Vinod Namana ◽

Kirti Chopra ◽

Ankur Sinha ◽

Sushilkumar Satish Gupta ◽

...

Keyword(s):

Cardiac Arrhythmias ◽

Alcohol Intoxication ◽

P Wave ◽

Acute Alcohol Intoxication ◽

Qrs Complex ◽

Qtc Prolongation ◽

Ecg Changes ◽

T Wave ◽

St Segment ◽

Electrocardiogram Ecg

Background: Acute alcohol intoxication has been associated with cardiac arrhythmias but the electrocardiogram (ECG) changes associated with acute alcohol intoxication are not well defined in the literature. Objective: Highlight the best evidence regarding the ECG changes associated with acute alcohol intoxication in otherwise healthy patients and the pathophysiology of the changes. Methods: A literature search was carried out; 4 studies relating to ECG changes with acute alcohol intoxication were included in this review. Results: Of the total 141 patients included in the review, 90 (63.8%) patients had P-wave prolongation, 80 (56%) patients had QTc prolongation, 19 (13.5%) patients developed T-wave abnormalities, 10 (7%) patients had QRS complex prolongation, 3 (2.12%) patients developed ST-segment depressions. Conclusion: The most common ECG changes associated with acute alcohol intoxication are (in decreasing order of frequency) P-wave and QTc prolongation, followed by T-wave abnormalities and QRS complex prolongation. Mostly, these changes are completely reversible.

Download Full-text

Research on the Values of Amplitudes of Electrocardiographic Waves Depending on Age in Kids

Bulletin of University of Agricultural Sciences and Veterinary Medicine Cluj-Napoca Veterinary Medicine ◽

10.15835/buasvmcn-vm:2020.0019 ◽

2020 ◽

Vol 77 (2) ◽

pp. 80

Author(s):

Dragos Corneliu COTOR ◽

Gabriel GAJAILA ◽

Aurel DAMIAN ◽

Ana Maria ZAGRAI ◽

Carmen PETCU ◽

...

Keyword(s):

Statistical Analysis ◽

Electric Activity ◽

P Wave ◽

Mechanical Activity ◽

Qrs Complex ◽

T Wave ◽

Electrocardiogram Ecg

The electrocardiogram (ECG) is a graphical recording of the cardiac electric activity during cardiac revolutions. This bio-current triggers and maintains the mechanical activity of the heart. Within this research, the amplitudes values of the electrocardiographic waves were determined in 6 leads: I, II, III, aVL, aVR and aVF. Thus, some electrocardiograms were recorded using limb lead in clinically healthy kids, aged 1 month, 3 months and 5 months, in order to achieve the proposed objectives. Then, the statistical analysis of the obtained results was performed using t (student) test.As a consequence of the interpretation of the obtained results, it was noticed that the limb leads can be used successfully for recording the electrocardiogram in kids because it provides an easy aspect to interpret. The highest amplitude of the electrocardiographic waves is recorded in I lead, in the case of the 1 month old kids (having the following values: 0.115 mV ± 0.010 for P wave; 0.625 mV ± 0.078 for QRS complex; 0.460 mV ± 0.045 for T wave) and in II lead (having the following values for the 3 months old kids: 0.071 mV ± 0.015 for P wave; 0.540 mV ± 0.064 for QRS complex; 0.310 mV ± 0.052 for T wave and having the following values for the 5 months old kids: 0.071 mV ± 0.015 for P wave; 0.455 mV ± 0.028 for QRS complex; 0.430 mV ± 0.026 for T wave). It also found that the lowest amplitude of electrocardiographic waves is recorded in the aVF lead, but this lead cannot be used for ECG recording in kids.

Download Full-text

Characteristics of the reindeer electrocardiogram

Rangifer ◽

10.7557/2.2.2.412 ◽

1982 ◽

Vol 2 (2) ◽

pp. 36

Author(s):

Jouni Timisjärvi ◽

Mauri Nieminen ◽

Sven Nikander

Keyword(s):

Heart Rate ◽

Indirect Evidence ◽

Frontal Plane ◽

Heart Beat ◽

P Wave ◽

Qrs Complex ◽

Single Plane ◽

T Wave ◽

Wide Range ◽

Electrocardiogram Ecg

The electrocardiogram (ECG) provides reliable information about heart rate, initiation of heart beat and also, to some degree, indirect evidence on the functional state of the heart muscle. A wide range of such information is readily obtainable from conventional scalar leads, even if the records are limited to a single plane. The present investigation deals with the normal reindeer ECG in the frontal plane. The technique used is the scalar recording technique based on the Einthovenian postulates. The P wave was positive in leads II, III and aVF, negative in lead aVL and variable in leads I and aVR. The direction of the P vector was 60 to 120°. The QRS complex was variable. The most common forms of QRS complex were R and rS in leads I and aVR; R, Rs and rS in lead aVL and Qr or qR in other leads. The most common direction of the QRS vector was 240 to 300°. The T wave was variable. The duration of various intervals and deflection depended on heart rate.Elektrokardiogram på ren.Abstract in Swedish / Sammandrag: Elektrokardiogramet (EKG) ger tillförlitliga uppgifter om hjärtfrekvens, retledning och, indirekt, delvis även om hjärtmuskelns funktionell tillstånd. Största delen av denna information fås med normal skalar koppling även om registrering sker i ett plan. I detta arbete har renens normala EKG i frontalplanet undersökts. Kopplingarna har baserats på Einthovs postulat. P-vågen var riktad uppåt i koppling II, III och aVF, nedåt i koppling aVL och den varierade i koppling I och aVR. P-vektorns riktning var 60 - 120°. QRS-komplexet varierade. De vanligaste formerna var R och rS i koppling I och aVR; R, Rs och rS i koppling aVL och Qr eller qR i andra kopplingar. Vanligen var QRS-vektorns riktning 240 - 300°. T-vågen varierade. Awikelserna och intervallernas längd var beroende av hiärtfrekvenssen.Poron sydänsähkökäyrän ominaisuuksia.Abstract in Finnish / Yhteenveto: Sydänsähkökäyrästä saadaan luotettavaa tietoa sydämen syketiheydestä, sähköisestä johtumisesta ja välillisesti jossain määrin myös sydänlihaksen toiminnallisesta tilasta. Suurin osa tämänkaltaista tietoa voidaan saada tavanomaisia skalaarisia kytkentöjäkäyttäen, ja usein yhdessä tasossa tapahtuva rekisteröinti on riittävä. Tässä työssä on tutkittu porojen normaalia sydänsähkökäyrää ja sen eri poikkeamien suuntautumista frontaalitasossa, kun rekisteröinnissä on käytetty Einthovenin postulaattien mukaisia raajakytkentöjä. P aalto suuntautui ylöspäin kythkennöissä II, III ja aVF, alaspäin kytkennässä aVL ja vaihteli kytkennöissä I ja aVR. P vektorin suunta oli 60 - 120°. QRS kompleksi vaihteli. Tavallisimmat muodot olivat R ja rS kytkennöissä I ja aVR; R, Rs ja rS kytkennässä aVL ja Qr tai qR muissa kytkennöissä. Tavallisin QRS vektorin suunta oli 240 - 300°. T aalto vaihteli. Poikkeaminen ja intervallien kesto riippui sydämen syketiheydestä.

Download Full-text

HIGH-SPEED APPROACH FOR DELINEATING P AND T WAVES CHARACTERISTICS IN ELECTROCARDIOGRAM SIGNAL

Journal of Mechanics in Medicine and Biology ◽

10.1142/s0219519414300038 ◽

2014 ◽

Vol 14 (02) ◽

pp. 1430003 ◽

Cited By ~ 1

Author(s):

SAMEER K. SALIH ◽

S. A. ALJUNID ◽

SYED M. ALJUNID ◽

OTEH MASKON ◽

ABID YAHYA

Keyword(s):

High Speed ◽

P Wave ◽

Detection Accuracy ◽

Peak Time ◽

Detection Rates ◽

New Approach ◽

Ecg Signals ◽

T Wave ◽

Electrocardiogram Ecg ◽

T Waves

Identifying and delineating P and T wave characteristics are greatly important in interpreting and diagnosing electrocardiogram (ECG) signals. P and T waves with high accuracy are more difficult to delineate because of their various shapes, positions, directions and boundaries. This paper proposes a high-speed approach to delineate P and T waves in a single lead using two high-speed algorithms of high detection accuracy. This approach presents a simple, adaptive and intelligent P and T wave scan method that determines the onset, peak and end time locations within an adaptive period appointed by previous records of the QRS complex. By using a translating (rising to/from falling) interval inside the scan wave, the peak time location of P and T waves and the T wave sign (upward or downward) are determined. Continuously, this time location is considered a reference point for determining the onset and the end time locations based on a series of computed outcomes related to amplitude and slope difference. The new approach is validated by 105 annotated records from the QT database collected from seven different categories of ECG signals. Simulation results show that the average detection rates of sensitivity and positive predictivity are equal to 99.97% and 99.36% for P wave and 99.98% and 99.26% for T wave, respectively. The average time errors computed by the mean and standard deviation for the P wave onset, peak and end time locations are -3.00 ± 2.94, -0.69 ± 4.42 and 0.67 ± 4.56 ms, respectively. The values for T wave are -3.33 ± 4.96, 0.24 ± 5.36 and -0.36 ± 5.68 ms. Results demonstrate the reliability, accuracy and forcefulness of the proposed approach in delineating various categories of P and T waves.

Download Full-text

Automated Detection of Depolarization and Repolarization of Cardiac Signal for Arrhythmia Classification

International Journal of Online and Biomedical Engineering (iJOE) ◽

10.3991/ijoe.v17i02.18955 ◽

2021 ◽

Vol 17 (02) ◽

pp. 173

Author(s):

Usha Kumari Chintalapati ◽

Md. Aqeel Manzar ◽

Tarun Varma N ◽

Reethika A ◽

Priya Samhitha B ◽

...

Keyword(s):

Early Stage ◽

Heart Diseases ◽

Heart Rhythm ◽

P Wave ◽

Qrs Complex ◽

S Wave ◽

S Waves ◽

T Wave ◽

Main Components ◽

Electrocardiogram Ecg

Irregular heartbeat results in heart diseases. Cardiac deaths are most seen across the globe. Detecting the heart problems in early stage can reduce the death rate. Electrocardiogram (ECG) is one of the most popular method for diagnosing different arrhythmias. Arrhythmia means irregular activity of heart or abnormal heart rhythm. In this paper, cardiac signal peaks P-wave, QRS complex and T-wave are detected for classifying the type of arrhythmia. These are the main components of ECG signal. P-wave is of very small duration, it is ex- plains about the atrial depolarization. The QRS complex may include combination of Q-wave, R-wave, and S-wave. But every QRS complex may not contain Q-R-S waves. It explains about ventricular depolarization. Whereas T wave is about ventricular re-polarization. S-Golay filter is used for denoising. This is used for smoothing the data which thereby, increases the precision of data without distortion of signal tendency. The patient data is collected from MIT-BIH Arrhythmia database for analysis. The simulation is done in Matlab software

Download Full-text

Data Independent Acquisition Based Bi-Directional Deep Networks for Biometric ECG Authentication

Applied Sciences ◽

10.3390/app11031125 ◽

2021 ◽

Vol 11 (3) ◽

pp. 1125

Author(s):

Htet Myet Lynn ◽

Pankoo Kim ◽

Sung Bum Pan

Keyword(s):

Short Term Memory ◽

Feature Learning ◽

Biometric Authentication ◽

Qrs Complex ◽

Data Types ◽

Ecg Signals ◽

Data Independent Acquisition ◽

Deep Networks ◽

Classification Tasks ◽

Electrocardiogram Ecg

In this report, the study of non-fiducial based approaches for Electrocardiogram(ECG) biometric authentication is examined, and several excessive techniques are proposed to perform comparative experiments for evaluating the best possible approach for all the classification tasks. Non-fiducial methods are designed to extract the discriminative information of a signal without annotating fiducial points. However, this process requires peak detection to identify a heartbeat signal. Based on recent studies that usually rely on heartbeat segmentation, QRS detection is required, and the process can be complicated for ECG signals for which the QRS complex is absent. Thus, many studies only conduct biometric authentication tasks on ECG signals with QRS complexes, and are hindered by similar limitations. To overcome this issue, we proposed a data-independent acquisition method to facilitate highly generalizable signal processing and feature learning processes. This is achieved by enhancing random segmentation to avoid complicated fiducial feature extraction, along with auto-correlation to eliminate the phase difference due to random segmentation. Subsequently, a bidirectional recurrent neural network (RNN) with long short-term memory (LSTM) deep networks is utilized to automatically learn the features associated with the signal and to perform an authentication task. The experimental results suggest that the proposed data-independent approach using a BLSTM network achieves a relatively high classification accuracy for every dataset relative to the compared techniques. Moreover, it exhibited a significantly higher accuracy rate in experiments using ECG signals without the QRS complex. The results also revealed that data-dependent methods can only perform well for specified data types and amendments of data variations, whereas the presented approach can also be considered for generalization to other quasi-periodical biometric signal-based classification tasks in future studies.

Download Full-text

Literature Review on Sleep APNEA Analysis by Machine Learning Algorithms Using ECG Signals

Journal of Physics Conference Series ◽

10.1088/1742-6596/1937/1/012054 ◽

2021 ◽

Vol 1937 (1) ◽

pp. 012054

Author(s):

V. Ankitha ◽

P. Manimegalai ◽

P. Subha Hency Jose ◽

P. Raji

Keyword(s):

Machine Learning ◽

Sleep Apnea ◽

Literature Review ◽

Learning Algorithms ◽

Machine Learning Algorithms ◽

Ecg Signals

Download Full-text

Classification of Sleep Apnea Using ECG Signals With Machine Learning Techniques

Advances in Medical Technologies and Clinical Practice - Advancing the Investigation and Treatment of Sleep Disorders Using AI ◽

10.4018/978-1-7998-8018-9.ch010 ◽

2021 ◽

pp. 184-203

Author(s):

Karthik R. ◽

Ifrah Alam ◽

Bandaru Umamadhuri ◽

Bharath K. P. ◽

Rajesh Kumar M.

Keyword(s):

Machine Learning ◽

Sleep Apnea ◽

Soft Tissues ◽

Amplitude Distribution ◽

Feature Reduction ◽

Machine Learning Algorithms ◽

Supervised Machine Learning ◽

Machine Learning Techniques ◽

Feed Forward Neural Network ◽

Ecg Signals

In this chapter, the authors use various signal processing techniques to analyze and gain insights on how ECG signals for patients suffering from sleep apnea (sleep apnea or obstructive sleep apnea occurs when the muscles that support the soft tissues in the throat, such as tongue and soft palate, relax temporarily) disease vary with respect to a normal person's ECG. The work has three stages: firstly, to identify waves, complexes, morphology in an ECG which reflect the presence of the disease; second, feature extraction techniques to extract features of ECG such as duration of the wave, amplitude distribution, and morphology classes; and third, detailed clustering (unsupervised) algorithm analysis of the extracted features with efficient feature reduction methodologies such as PCA and LDA. Finally, the authors use supervised machine learning algorithms (SVM, naive Bayes classifier, feed forward neural network, and decision tree) to distinguish between ECG signals with sleep apnea and normal ECG signals.

Download Full-text

ECG SIGNAL GENERATION AND HEART RATE VARIABILITY SIGNAL EXTRACTION: SIGNAL PROCESSING, FEATURES DETECTION, AND THEIR CORRELATION WITH CARDIAC DISEASES

Journal of Mechanics in Medicine and Biology ◽

10.1142/s021951941240012x ◽

2012 ◽

Vol 12 (04) ◽

pp. 1240012 ◽

Cited By ~ 4

Author(s):

GOUTHAM SWAPNA ◽

DHANJOO N. GHISTA ◽

ROSHAN JOY MARTIS ◽

ALVIN P. C. ANG ◽

SUBBHURAAM VINITHA SREE

Keyword(s):

Heart Rate ◽

Heart Rate Variability ◽

Sinus Bradycardia ◽

Premature Ventricular Contraction ◽

P Wave ◽

Ecg Signal ◽

Qrs Complex ◽

T Wave ◽

St Segment ◽

Cardiac Disorders

The sum total of millions of cardiac cell depolarization potentials can be represented by an electrocardiogram (ECG). Inspection of the P–QRS–T wave allows for the identification of the cardiac bioelectrical health and disorders of a subject. In order to extract the important features of the ECG signal, the detection of the P wave, QRS complex, and ST segment is essential. Therefore, abnormalities of these ECG parameters are associated with cardiac disorders. In this work, an introduction to the genesis of the ECG is given, followed by a depiction of some abnormal ECG patterns and rhythms (associated with P–QRS–T wave parameters), which have come to be empirically correlated with cardiac disorders (such as sinus bradycardia, premature ventricular contraction, bundle-branch block, atrial flutter, and atrial fibrillation). We employed algorithms for ECG pattern analysis, for the accurate detection of the P wave, QRS complex, and ST segment of the ECG signal. We then catagorited and tabulated these cardiac disorders in terms of heart rate, PR interval, QRS width, and P wave amplitude. Finally, we discussed the characteristics and different methods (and their measures) of analyting the heart rate variability (HRV) signal, derived from the ECG waveform. The HRV signals are characterised in terms of these measures, then fed into classifiers for grouping into categories (for normal subjects and for disorders such as cardiac disorders and diabetes) for carrying out diagnosis.

Download Full-text