Robust Peak Detection for Holter ECGs by Self-Organized Operational Neural Networks

Kıranyaz, Serkan; Malik, Junaid; Zahid, Muhammad Uzair; İnce, Türker; Chowdhury, Muhammad E. H.; Khandakar, Amith; Tahir, Anas; Gabbouj, Moncef

doi:10.1109/tnnls.2022.3158867

Cited by 29 publications

(20 citation statements)

References 52 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Duraj et al showed high performance on LUDB by incorporating the residual and squeeze-excitation blocks into the 1D U-Net architecture for extracting segments, such as P-waves, QRS complexes, and T-waves, regardless of the lead [ 88 ]. Gabbouj et al proposed a 1D self-organized operational neural network (ONN), evaluated on second CPSC, and achieved recall of 99.79%, precision of 98.42%, and F1-score of 99.10% [ 24 , 89 ]. However, during the training phase, it featured a high complexity in the number of multiply–accumulate operations and the number of parameters compared with 1D CNN architectures.…”

Section: Related Workmentioning

confidence: 99%

A Deep Learning Architecture Using 3D Vectorcardiogram to Detect R-Peaks in ECG with Enhanced Precision

Mehri

Calmon²,

Odille

et al. 2023

Sensors

View full text Add to dashboard Cite

Providing reliable detection of QRS complexes is key in automated analyses of electrocardiograms (ECG). Accurate and timely R-peak detections provide a basis for ECG-based diagnoses and to synchronize radiologic, electrophysiologic, or other medical devices. Compared with classical algorithms, deep learning (DL) architectures have demonstrated superior accuracy and high generalization capacity. Furthermore, they can be embedded on edge devices for real-time inference. 3D vectorcardiograms (VCG) provide a unifying framework for detecting R-peaks regardless of the acquisition strategy or number of ECG leads. In this article, a DL architecture was demonstrated to provide enhanced precision when trained and applied on 3D VCG, with no pre-processing nor post-processing steps. Experiments were conducted on four different public databases. Using the proposed approach, high F1-scores of 99.80% and 99.64% were achieved in leave-one-out cross-validation and cross-database validation protocols, respectively. False detections, measured by a precision of 99.88% or more, were significantly reduced compared with recent state-of-the-art methods tested on the same databases, without penalty in the number of missed peaks, measured by a recall of 99.39% or more. This approach can provide new applications for devices where precision, or positive predictive value, is essential, for instance cardiac magnetic resonance imaging.

show abstract

Section: Related Workmentioning

confidence: 99%

A Deep Learning Architecture Using 3D Vectorcardiogram to Detect R-Peaks in ECG with Enhanced Precision

Mehri

Calmon²,

Odille

et al. 2023

Sensors

View full text Add to dashboard Cite

show abstract

“…For example, in [5], a Holter system for monitoring the state of the heart based on the Zigbee method was developed, which has a measuring and recording device. In another work [6], a method for detecting peaks for Holter devices using self-organizing operational neural networks is presented. Also, Holter has a number of other names: outpatient ECS monitoring, longterm ECS monitoring, 24-hour ECS monitoring.…”

Section: Literature Review and Problem Statementmentioning

confidence: 99%

Development of an atrioventricular block prediction of method for portable heart monitoring system

Bekbay

Alimbayeva

Alimbayev³

et al. 2022

EEJET

View full text Add to dashboard Cite

The object of the study is a portable system that allows real-time monitoring of the state of the heart for the timely provision of medical care. The task of detecting atrioventricular (AV) blocks in the conditions of free motor activity of the patient is being solved. To develop a method for detecting AV block, models of the electrical activity of the heart were used to take into account the spatiotemporal organization of the process of spreading excitation, analyze the dynamics of the behavior of the cardiovascular systems (CVS) for any value of the period of atrial excitation, and assess the degree of fitness of the CVS. The proposed method made it possible to determine the heart rate (HR) at which the development of AV block is possible. AV block of the III degree – heart rate 304 bpm; AV block of the II degree with the loss of half of the impulses – heart rate 260 bpm; AV block II degree with loss of individual impulses – heart rate 234 bpm; AV block of the 1st degree – heart rate 200 bpm. Prediction of AV block allows assessing the degree of “training” of the patient’s heart. The obtained quantitative results are consistent with the heart rate values known to modern health care. The developed method was implemented on the basis of a portable ECG monitoring system previously developed by the authors. Tests of the portable ECG monitoring system indicate an increase in the sensitivity and specificity of diagnosing cardiac arrhythmia and confirm the achievement of the goal of this study: improving the efficiency of diagnostics and expanding the functionality of the portable ECG monitoring system

show abstract

“…Different solutions are used for real-time monitoring on wearable devices, of which LSTM is also the most common (Amirshahi and Hashemi 2019, Saadatnejad et al 2020, Wójcikowski 2021, Gabbouj et al 2022. The same network was used as a basis for this study and was proven to be sufficient for real-time analysis on wearable devices (Amirshahi and Hashemi 2019, Wójcikowski 2021, Gabbouj et al 2022. Also, the 2020 PhysioNet/Computing in Cardiology Challenge showed that the most common solutions used for 12-lead classification were also CNNs and RNNs (Alday et al 2021).…”

Section: Introductionmentioning

confidence: 98%

“…In that field, one of the most commonly used classifiers is the Long Storm-term Memory (LSTM) network (Warrick et al 2020). Different solutions are used for real-time monitoring on wearable devices, of which LSTM is also the most common (Amirshahi and Hashemi 2019, Saadatnejad et al 2020, Wójcikowski 2021, Gabbouj et al 2022. The same network was used as a basis for this study and was proven to be sufficient for real-time analysis on wearable devices (Amirshahi and Hashemi 2019, Wójcikowski 2021, Gabbouj et al 2022.…”

Section: Introductionmentioning

confidence: 99%

Comparison of neural basis expansion analysis for interpretable time series (N-BEATS) and recurrent neural networks for heart dysfunction classification

2022

View full text Add to dashboard Cite

Objective: The primary purpose of this work is to analyze the ability of N-BEATS architecture for the problem of prediction and classification of electrocar- diogram (ECG) signals. To achieve this, performance comparison with various types of other SotA (state-of-the-art) recurrent neural network architectures commonly used for such problems is conducted. Approach: Four architectures (N-BEATS, LSTM, LSTM with peepholes, GRU) were tested for performance and dimension reduction problems for different number of leads (2, 3, 4, 6, 12), both in variants consisting of blended branches, allowing retaining ac- curacy while reducing the computational capacity needed. The analysis was performed on datasets and using metrics from Challenges in Cardiology (CinC) 2021 competition. Main results: Best results were achieved for LSTM with peepholes, then LSTM, GRU and the worst for N-BEATS (challenge metrics respectively: 0.42, 0.40, 0.39, 0.35; for times: 0.0395 s, 0.0036 s, 0.0027 s, 0.0002 s). Commonly used LSTM outperforms N- BEATS in terms of multi-label classification, data set resilience, and obtained challenge metrics. Still, N-BEATS can obtain acceptable results for 2 lead classification (metric of 0.35 for N-BEATS and 0.38 for other networks) and outperforms other solutions in terms of complexity and speed. Significance: This paper features a novel approach of using the N-BEATS, which was previously used only for forecasting, to classify ECG signals with success. While N- BEATS multi-label classification capacity is lower than LSTM, its speed obtaining results with a reduced number of leads (faster by one to two degrees of magnitude) allows for arrhythmias detection and classification while using off-the-shelf wearable devices (Holter monitors, sport bands, etc.)

show abstract

Robust Peak Detection for Holter ECGs by Self-Organized Operational Neural Networks

Cited by 29 publications

References 52 publications

A Deep Learning Architecture Using 3D Vectorcardiogram to Detect R-Peaks in ECG with Enhanced Precision

A Deep Learning Architecture Using 3D Vectorcardiogram to Detect R-Peaks in ECG with Enhanced Precision

Development of an atrioventricular block prediction of method for portable heart monitoring system

Comparison of neural basis expansion analysis for interpretable time series (N-BEATS) and recurrent neural networks for heart dysfunction classification

Contact Info

Product

Resources

About