ECG assessment based on neural networks with pretraining

Ripoll, Vicent J. Ribas; Wojdel, Anna; Romero, Enrique; Ramos, Pablo Manuel; Brugada, Josép

doi:10.1016/j.asoc.2016.08.013

Cited by 41 publications

(16 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Optimization was done using Stochastic Gradient Descent with Nesterov acceleration, a learning rate of 10 −3 , a learning rate decay of 10 −6 , and a momentum of 0.9. These are recommneded/typical values used in Stochastic Gradient Descent from the specialized literature on the topic [54, 55]. One of the advantages of our design is that the three blocks of the network are optimized end-to-end to simultaneously extract the high level feature description of the signal (convolutional block), its temporal relationships (recurrent network block), and the arrhythmia classification (classification block).…”

Section: Methodsmentioning

confidence: 99%

Mixed convolutional and long short-term memory network for the detection of lethal ventricular arrhythmia

et al. 2019

View full text Add to dashboard Cite

Early defibrillation by an automated external defibrillator (AED) is key for the survival of out-of-hospital cardiac arrest (OHCA) patients. ECG feature extraction and machine learning have been successfully used to detect ventricular fibrillation (VF) in AED shock decision algorithms. Recently, deep learning architectures based on 1D Convolutional Neural Networks (CNN) have been proposed for this task. This study introduces a deep learning architecture based on 1D-CNN layers and a Long Short-Term Memory (LSTM) network for the detection of VF. Two datasets were used, one from public repositories of Holter recordings captured at the onset of the arrhythmia, and a second from OHCA patients obtained minutes after the onset of the arrest. Data was partitioned patient-wise into training (80%) to design the classifiers, and test (20%) to report the results. The proposed architecture was compared to 1D-CNN only deep learners, and to a classical approach based on VF-detection features and a support vector machine (SVM) classifier. The algorithms were evaluated in terms of balanced accuracy (BAC), the unweighted mean of the sensitivity (Se) and specificity (Sp). The BAC, Se, and Sp of the architecture for 4-s ECG segments was 99.3%, 99.7%, and 98.9% for the public data, and 98.0%, 99.2%, and 96.7% for OHCA data. The proposed architecture outperformed all other classifiers by at least 0.3-points in BAC in the public data, and by 2.2-points in the OHCA data. The architecture met the 95% Sp and 90% Se requirements of the American Heart Association in both datasets for segment lengths as short as 3-s. This is, to the best of our knowledge, the most accurate VF detection algorithm to date, especially on OHCA data, and it would enable an accurate shock no shock diagnosis in a very short time.

show abstract

Section: Methodsmentioning

confidence: 99%

Mixed convolutional and long short-term memory network for the detection of lethal ventricular arrhythmia

et al. 2019

View full text Add to dashboard Cite

show abstract

“…Cardiovascular diseases (CVDs) have become a major contributor to human mortality [1]. According to the latest statistics produced by the World Health Organization (WHO), the mortality rate from CVDs will rise from 246 people per one million in 2015 to 264 people per one million in 2030 [2,3]. It is known that abnormal blood pressure can produce many complications for the heart, kidneys, and other vital organs, causing irreversible injury.…”

Section: Introductionmentioning

confidence: 99%

“…To eliminate the classical feature extraction phase, we introduced a transfer learning method based on the pretrained convolutional neural network (CNN). Common pretrained CNN methods include AlexNet, VGGNet, and GoogLeNet, which are trained based on the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) dataset [2]. In this paper, we propose a method of classifying hypertension using continuous wavelet transformation and GoogLeNet [32], which only needs the PPG signal to realize the identification and classification of hypertension.…”

Section: Introductionmentioning

confidence: 99%

Photoplethysmography and Deep Learning: Enhancing Hypertension Risk Stratification

et al. 2018

View full text Add to dashboard Cite

Blood pressure is a basic physiological parameter in the cardiovascular circulatory system. Long-term abnormal blood pressure will lead to various cardiovascular diseases, making the early detection and assessment of hypertension profoundly significant for the prevention and treatment of cardiovascular diseases. In this paper, we investigate whether or not deep learning can provide better results for hypertension risk stratification when compared to the classical signal processing and feature extraction methods. We tested a deep learning method for the classification and evaluation of hypertension using photoplethysmography (PPG) signals based on the continuous wavelet transform (using Morse) and pretrained convolutional neural network (using GoogLeNet). We collected 121 data recordings from the Multiparameter Intelligent Monitoring in Intensive Care (MIMIC) Database, each containing arterial blood pressure (ABP) and photoplethysmography (PPG) signals. The ABP signals were utilized to extract blood pressure category labels, and the PPG signals were used to train and test the model. According to the seventh report of the Joint National Committee, blood pressure levels are categorized as normotension (NT), prehypertension (PHT), and hypertension (HT). For the early diagnosis and assessment of HT, the timely detection of PHT and the accurate diagnosis of HT are significant. Therefore, three HT classification trials were set: NT vs. PHT, NT vs. HT, and (NT + PHT) vs. HT. The F-scores of these three classification trials were 80.52%, 92.55%, and 82.95%, respectively. The tested deep method achieved higher accuracy for hypertension risk stratification when compared to the classical signal processing and feature extraction method. Additionally, the method achieved comparable results to another approach that requires electrocardiogram and PPG signals.

show abstract

“…Neural Network (NN) algorithms have been frequently applied to solve data mining problems in biomedical applications [27][28][29], as for example to develop novel information extraction, diagnosis and clinical prediction techniques based on the electrocardiogram (ECG) signal [30][31][32][33][34][35]. On the contrary, the application of NN algorithms in APWrelated problems is not being so frequently observed.…”

Section: Introductionmentioning

confidence: 99%

Supervised learning methods for pathological arterial pulse wave differentiation: A SVM and neural networks approach

Paiva

Cardoso

Pereira

2018

International Journal of Medical Informatics

View full text Add to dashboard Cite

show abstract

ECG assessment based on neural networks with pretraining

Cited by 41 publications

References 20 publications

Mixed convolutional and long short-term memory network for the detection of lethal ventricular arrhythmia

Mixed convolutional and long short-term memory network for the detection of lethal ventricular arrhythmia

Photoplethysmography and Deep Learning: Enhancing Hypertension Risk Stratification

Supervised learning methods for pathological arterial pulse wave differentiation: A SVM and neural networks approach

Contact Info

Product

Resources

About