Fully Convolutional Deep Neural Networks with Optimized Hyperparameters for Detection of Shockable and Non-Shockable Rhythms

Deep neural networks (DNN) are state-of-the-art machine learning algorithms that can be learned to self-extract significant features of the electrocardiogram (ECG) and can generally provide high-output diagnostic accuracy if subjected to robust training and optimization on large datasets at high computational cost. So far, limited research and optimization of DNNs in shock advisory systems is found on large ECG arrhythmia databases from out-of-hospital cardiac arrests (OHCA). The objective of this study is to optimize the hyperparameters (HPs) of deep convolutional neural networks (CNN) for detection of shockable (Sh) and nonshockable (NSh) rhythms, and to validate the best HP settings for short and long analysis durations (2–10 s). Large numbers of (Sh + NSh) ECG samples were used for training (720 + 3170) and validation (739 + 5921) from Holters and defibrillators in OHCA. An end-to-end deep CNN architecture was implemented with one-lead raw ECG input layer (5 s, 125 Hz, 2.5 uV/LSB), configurable number of 5 to 23 hidden layers and output layer with diagnostic probability p ∈ [0: Sh,1: NSh]. The hidden layers contain N convolutional blocks × 3 layers (Conv1D (filters = Fi, kernel size = Ki), max-pooling (pool size = 2), dropout (rate = 0.3)), one global max-pooling and one dense layer. Random search optimization of HPs = {N, Fi, Ki}, i = 1, … N in a large grid of N = [1, 2, … 7], Fi = [5;50], Ki = [5;100] was performed. During training, the model with maximal balanced accuracy BAC = (Sensitivity + Specificity)/2 over 400 epochs was stored. The optimization principle is based on finding the common HPs space of a few top-ranked models and prediction of a robust HP setting by their median value. The optimal models for 1–7 CNN layers were trained with different learning rates LR = [10−5; 10−2] and the best model was finally validated on 2–10 s analysis durations. A number of 4216 random search models were trained. The optimal models with more than three convolutional layers did not exhibit substantial differences in performance BAC = (99.31–99.5%). Among them, the best model was found with {N = 5, Fi = {20, 15, 15, 10, 5}, Ki = {10, 10, 10, 10, 10}, 7521 trainable parameters} with maximal validation performance for 5-s analysis (BAC = 99.5%, Se = 99.6%, Sp = 99.4%) and tolerable drop in performance (<2% points) for very short 2-s analysis (BAC = 98.2%, Se = 97.6%, Sp = 98.7%). DNN application in future-generation shock advisory systems can improve the detection performance of Sh and NSh rhythms and can considerably shorten the analysis duration complying with resuscitation guidelines for minimal hands-off pauses.

[1]  Yiming Yang,et al.  DARTS: Differentiable Architecture Search , 2018, ICLR.

[2]  Jeffrey M. Hausdorff,et al.  Physionet: Components of a New Research Resource for Complex Physiologic Signals". Circu-lation Vol , 2000 .

[3]  Md. Kamrul Hasan,et al.  A simple time domain algorithm for the detection of ventricular fibrillation in electrocardiogram , 2011, Signal Image Video Process..

[4]  Hongxing Liu,et al.  ECG Heartbeat Classification Using Convolutional Neural Networks , 2020, IEEE Access.

[5]  Jonathan Rubin,et al.  Densely connected convolutional networks for detection of atrial fibrillation from short single-lead ECG recordings. , 2018, Journal of electrocardiology.

[6]  Jonathan Rubin,et al.  Analyzing single-lead short ECG recordings using dense convolutional neural networks and feature-based post-processing to detect atrial fibrillation , 2018, Physiological measurement.

[7]  Karl Unterkofler,et al.  Detecting Ventricular Fibrillation by Time-Delay Methods , 2007, IEEE Transactions on Biomedical Engineering.

[8]  Pavel Jurák,et al.  Fast Detection of Ventricular Fibrillation and Ventricular Tachycardia in 1-Lead ECG from Three-Second Blocks , 2018, 2018 Computing in Cardiology Conference (CinC).

[9]  Manal M. Tantawi,et al.  Heartbeat Classification Using 1D Convolutional Neural Networks , 2019, AISI.

[10]  Elizabeth F. Wanner,et al.  QRS Detection in ECG Signal with Convolutional Network , 2018, CIARP.

[11]  Sergey Alekseev,et al.  ECG Segmentation by Neural Networks: Errors and Correction , 2018, 2019 International Joint Conference on Neural Networks (IJCNN).

[12]  C. Deakin,et al.  Chest compression pauses during defibrillation attempts , 2016, Current opinion in critical care.

[13]  Daniel Davis,et al.  Survival after application of automatic external defibrillators before arrival of the emergency medical system: evaluation in the resuscitation outcomes consortium population of 21 million. , 2010, Journal of the American College of Cardiology.

[14]  Moncef Gabbouj,et al.  Real-Time Patient-Specific ECG Classification by 1-D Convolutional Neural Networks , 2016, IEEE Transactions on Biomedical Engineering.

[15]  Minho Choi,et al.  QRS detection method based on fully convolutional networks for capacitive electrocardiogram , 2019, Expert Syst. Appl..

[16]  Qiao Li,et al.  Ventricular Fibrillation and Tachycardia Classification Using a Machine Learning Approach , 2014, IEEE Transactions on Biomedical Engineering.

[17]  Jinsul Kim,et al.  An Automated ECG Beat Classification System Using Convolutional Neural Networks , 2016, 2016 6th International Conference on IT Convergence and Security (ICITCS).

[18]  Irena Jekova,et al.  Shock advisory system with minimal delay triggering after end of chest compressions: accuracy and gained hands-off time. , 2011, Resuscitation.

[19]  Sayan Mukhopadhyay,et al.  Deep Learning and Neural Networks , 2018 .

[20]  I Jekova,et al.  Comparison of five algorithms for the detection of ventricular fibrillation from the surface ECG. , 2000, Physiological measurement.

[21]  José Luis Rojo-Álvarez,et al.  Detection of Life-Threatening Arrhythmias Using Feature Selection and Support Vector Machines , 2014, IEEE Transactions on Biomedical Engineering.

[22]  U. Rajendra Acharya,et al.  A deep convolutional neural network model to classify heartbeats , 2017, Comput. Biol. Medicine.

[23]  J P Ornato,et al.  Automatic external defibrillators for public access defibrillation: recommendations for specifying and reporting arrhythmia analysis algorithm performance, incorporating new waveforms, and enhancing safety. A statement for health professionals from the American Heart Association Task Force on Automa , 1997, Circulation.

[24]  Azeddine Mjahad,et al.  Ventricular Fibrillation and Tachycardia detection from surface ECG using time-frequency representation images as input dataset for machine learning , 2017, Comput. Methods Programs Biomed..

[25]  Irena Jekova,et al.  Real time detection of ventricular fibrillation and tachycardia. , 2004, Physiological measurement.

[26]  U. Rajendra Acharya,et al.  Automated detection of arrhythmias using different intervals of tachycardia ECG segments with convolutional neural network , 2017, Inf. Sci..

[27]  David D. Salcido,et al.  The impact of peri-shock pause on survival from out-of-hospital shockable cardiac arrest during the Resuscitation Outcomes Consortium PRIMED trial. , 2014, Resuscitation.

[28]  Richard E. Kerber,et al.  Automatic External Defibrillators for Public Access Defibrillation: Recommendations for Specifying and Reporting Arrhythmia Analysis Algorithm Performance, Incorporating New Waveforms, and Enhancing Safety , 1997, Biomedical instrumentation & technology.

[29]  Unai Irusta,et al.  Deep Neural Networks for ECG-Based Pulse Detection during Out-of-Hospital Cardiac Arrest , 2019, Entropy.

[30]  Irena Jekova,et al.  Shock advisory tool: Detection of life-threatening cardiac arrhythmias and shock success prediction by means of a common parameter set , 2007, Biomed. Signal Process. Control..

[31]  J. Bache European Resuscitation Council Guidelines for Resuscitation , 1998 .

[32]  Minh Tuan Nguyen,et al.  Feature Learning Using Convolutional Neural Network for Cardiac Arrest Detection , 2018, 2018 International Conference on Smart Green Technology in Electrical and Information Systems (ICSGTEIS).

[33]  K. Årestedt,et al.  Sensitivity and specificity of two different automated external defibrillators. , 2017, Resuscitation.

[34]  Samarendra Dandapat,et al.  Detection of Shockable Ventricular Arrhythmia using Variational Mode Decomposition , 2016, Journal of Medical Systems.

[35]  Felipe Alonso-Atienza,et al.  Mixed convolutional and long short-term memory network for the detection of lethal ventricular arrhythmia , 2019, PloS one.

[36]  Xu-Sheng Zhang,et al.  Detecting ventricular tachycardia and fibrillation by complexity measure , 1999, IEEE Transactions on Biomedical Engineering.

[37]  Irena Jekova,et al.  Assessment of ECG frequency and morphology parameters for automatic classification of life-threatening cardiac arrhythmias , 2005, Physiological measurement.

[38]  Irena Jekova,et al.  Method for ventricular fibrillation detection in the external electrocardiogram using nonlinear prediction. , 2002, Physiological measurement.

[39]  Javier Del Ser,et al.  ECG-based Random Forest Classifier for Cardiac Arrest Rhythms , 2019, 2019 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC).

[40]  U. Irusta,et al.  Deep learning approach for a shock advise algorithm using short electrocardiogram analysis intervals , 2019, Resuscitation.

[41]  D. Atkins,et al.  Comparison of Electrocardiographic Characteristics of Adults and Children for Automated External Defibrillator Algorithms , 2014, Pediatric emergency care.

[42]  S Barro,et al.  Algorithmic sequential decision-making in the frequency domain for life threatening ventricular arrhythmias and imitative artefacts: a diagnostic system. , 1989, Journal of biomedical engineering.

[43]  Scott David Greenwald,et al.  The development and analysis of a ventricular fibrillation detector , 1986 .

[44]  Hamido Fujita,et al.  Decision support system for arrhythmia prediction using convolutional neural network structure without preprocessing , 2019, Applied Intelligence.

[45]  A. Murray,et al.  Comparison of four techniques for recognition of ventricular fibrillation from the surface ECG , 1993, Medical and Biological Engineering and Computing.

[46]  Irena Jekova,et al.  Bench study of the accuracy of a commercial AED arrhythmia analysis algorithm in the presence of electromagnetic interferences , 2009, Physiological measurement.

[47]  A. Amann,et al.  Reliability of old and new ventricular fibrillation detection algorithms for automated external defibrillators , 2005, Biomedical engineering online.

[48]  Quoc V. Le,et al.  Efficient Neural Architecture Search via Parameter Sharing , 2018, ICML.

[49]  Felipe Alonso-Atienza,et al.  Machine Learning Techniques for the Detection of Shockable Rhythms in Automated External Defibrillators , 2016, PloS one.

[50]  Yu Tsao,et al.  Noise Reduction in ECG Signals Using Fully Convolutional Denoising Autoencoders , 2019, IEEE Access.

[51]  Guigang Zhang,et al.  Deep Learning , 2016, Int. J. Semantic Comput..

[52]  U. Raghavendra,et al.  Automated identification of shockable and non-shockable life-threatening ventricular arrhythmias using convolutional neural network , 2018, Future Gener. Comput. Syst..

[53]  Masoumeh Haghpanahi,et al.  Cardiologist-level arrhythmia detection and classification in ambulatory electrocardiograms using a deep neural network , 2019, Nature Medicine.

[54]  Valérie Gay,et al.  Ventricular Tachycardia/Fibrillation Detection Algorithm for 24/7 Personal Wireless Heart Monitoring , 2007, ICOST.

[55]  Ali Bahrami Rad,et al.  SPECTRO-TEMPORAL ECG ANALYSIS FOR ATRIAL FIBRILLATION DETECTION , 2018, 2018 IEEE 28th International Workshop on Machine Learning for Signal Processing (MLSP).

[56]  T. Eftestøl,et al.  Fully automatic rhythm analysis during chest compression pauses. , 2015, Resuscitation.

[57]  Ye Li,et al.  Multiscaled Fusion of Deep Convolutional Neural Networks for Screening Atrial Fibrillation From Single Lead Short ECG Recordings , 2018, IEEE Journal of Biomedical and Health Informatics.

[58]  Sofía Ruiz de Gauna,et al.  A high-temporal resolution algorithm to discriminate shockable from nonshockable rhythms in adults and children. , 2012, Resuscitation.

[59]  Xuemei Guo,et al.  Non-invasive Fetal Electrocardiography Denoising Using Deep Convolutional Encoder-Decoder Networks , 2019, Lecture Notes in Electrical Engineering.

[60]  N. Thakor,et al.  Ventricular tachycardia and fibrillation detection by a sequential hypothesis testing algorithm , 1990, IEEE Transactions on Biomedical Engineering.

[61]  Jesús Requena-Carrión,et al.  Analysis of the robustness of spectral indices during ventricular fibrillation , 2013, Biomed. Signal Process. Control..

[62]  Daniel Jost,et al.  Comparison of Pediatric and Adult ECG Rhythm Analysis by Automated External Defibrillators During Out-of-Hospital Cardiac Arrest , 2018, 2018 Computing in Cardiology Conference (CinC).