No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Adversarial Multi-Task Learning for Robust End-to-End ECG-based Heartbeat Classification
Abstract
In clinical practice, heart arrhythmias are manually diagnosed by a doctor, which is a time-consuming process. Furthermore, this process is error-prone due to noise from the recording equipment and biological non-idealities of patients. Thus, an automated arrhythmia classifier would be time and cost-effective as well as offer better generalization across patients. In this paper, we propose an adversarial multitask learning method to improve the generalization of heartbeat arrythmia classification. We built an end-to-end deep neural network (DNN) system consisting of three sub-networks; a generator, a heartbeat-type discriminator, and a subject (or patient) discriminator. Each of these two discriminators had its own loss function to control its impact. The generator was "friendly" to the heartbeat-type discrimination task by minimizing its loss function and "hostile" to the subject discrimination task by maximizing its loss function. The network was trained using raw ECG signals to discriminate between five types of heartbeats - normal heartbeats, right bundle branch blocks (RBBB), premature ventricular contractions (PVC), paced beats (PB) and fusion of ventricular and normal beats (FVN). The method was tested with the MIT-BIH arrhythmia dataset and achieved a 17% reduction in classification error compared to a baseline using a fully-connected DNN classifier.Clinical Relevance-This work validates that it is possible to develop a subject-independent automated heart arrhythmia detection system to assist clinicians in the diagnosis process.
In the health care and medical domain, it has been proven challenging to diagnose correctly many diseases with complicated and interferential symptoms, including arrhythmia. However, with the evolution of artificial intelligence (AI) techniques, the diagnosis and prognosis of arrhythmia became easier for the physicians and practitioners using only an electrocardiogram (ECG) examination. This review presents a synthesis of the studies conducted in the last 12 years to predict arrhythmia’s occurrence by classifying automatically different heartbeat rhythms. From a variety of research academic databases, 40 studies were selected to analyze, among which 29 of them applied deep learning methods (72.5%), 9 of them addressed the problem with machine learning methods (22.5%), and 2 of them combined both deep learning and machine learning to predict arrhythmia (5%). Indeed, the use of AI for arrhythmia diagnosis is emerging in literature, although there are some challenging issues, such as the explicability of the Deep Learning methods and the computational resources needed to achieve high performance. However, with the continuous development of cloud platforms and quantum calculation for AI, we can achieve a breakthrough in arrhythmia diagnosis.
Background:
Electrocardiogram (ECG) is one of the most common noninvasive diagnostic tools that can provide useful information regarding a patient's health status. Deep learning (DL) is an area of intense exploration that leads the way in most attempts to create powerful diagnostic models based on physiological signals.
Objective:
This study aimed to provide a systematic review of DL methods applied to ECG data for various clinical applications.
Methods:
The PubMed search engine was systematically searched by combining "deep learning" and keywords such as "ecg," "ekg," "electrocardiogram," "electrocardiography," and "electrocardiology." Irrelevant articles were excluded from the study after screening titles and abstracts, and the remaining articles were further reviewed. The reasons for article exclusion were manuscripts written in any language other than English, absence of ECG data or DL methods involved in the study, and absence of a quantitative evaluation of the proposed approaches.
Results:
We identified 230 relevant articles published between January 2020 and December 2021 and grouped them into 6 distinct medical applications, namely, blood pressure estimation, cardiovascular disease diagnosis, ECG analysis, biometric recognition, sleep analysis, and other clinical analyses. We provide a complete account of the state-of-the-art DL strategies per the field of application, as well as major ECG data sources. We also present open research problems, such as the lack of attempts to address the issue of blood pressure variability in training data sets, and point out potential gaps in the design and implementation of DL models.
Conclusions:
We expect that this review will provide insights into state-of-the-art DL methods applied to ECG data and point to future directions for research on DL to create robust models that can assist medical experts in clinical decision-making.
The electrocardiogram (ECG) is one of the simplest and oldest tools to assess the heart condition of cardiac patients. Heart diseases have emerged as one of the leading causes of death all over the world. According to the world health organization (WHO), millions of people are dying every year from heart-related diseases. A classification model that can early detect arrhythmia will be able to reduce this number by manyfold. Many researchers are working in this area and proposed many deep learning and machine Learning based models for arrhythmia classification. These models have high accuracy but require a machine with high computational power. Hence, these models are not sustainable options for the practical field. In this paper, we have proposed a 1D Convolutional Neural Network (CNN) model with high accuracy and low computational complexity. Our proposed methodology is appraised on the MIT-BIH arrhythmia dataset. We achieved overall 98.25% accuracy into five classes with an f1 score of 98.24%, precision 97.58%, and recall 96.79% which is better than previous results classifying arrhythmia. We can claim that our proposed method is better than most other existing models because of the higher accuracy with a simple architecture that can be run on an edge device with relatively low hardware configuration.
Computerized electrocardiogram (ECG) interpretation plays a critical role in the clinical ECG workflow¹. Widely available digital ECG data and the algorithmic paradigm of deep learning² present an opportunity to substantially improve the accuracy and scalability of automated ECG analysis. However, a comprehensive evaluation of an end-to-end deep learning approach for ECG analysis across a wide variety of diagnostic classes has not been previously reported. Here, we develop a deep neural network (DNN) to classify 12 rhythm classes using 91,232 single-lead ECGs from 53,549 patients who used a single-lead ambulatory ECG monitoring device. When validated against an independent test dataset annotated by a consensus committee of board-certified practicing cardiologists, the DNN achieved an average area under the receiver operating characteristic curve (ROC) of 0.97. The average F1 score, which is the harmonic mean of the positive predictive value and sensitivity, for the DNN (0.837) exceeded that of average cardiologists (0.780). With specificity fixed at the average specificity achieved by cardiologists, the sensitivity of the DNN exceeded the average cardiologist sensitivity for all rhythm classes. These findings demonstrate that an end-to-end deep learning approach can classify a broad range of distinct arrhythmias from single-lead ECGs with high diagnostic performance similar to that of cardiologists. If confirmed in clinical settings, this approach could reduce the rate of misdiagnosed computerized ECG interpretations and improve the efficiency of expert human ECG interpretation by accurately triaging or prioritizing the most urgent conditions.
Building deep neural network acoustic models directly based on far-field speech from multiple recording environments with different acoustic properties is an increasingly popular approach to address the problem of distant speech recognition. The currently common approach to building such multi-condition (multi-domain) models is to compile available data from all different environments into a single train set, discarding information regarding the specific environment to which each utterance belongs. We propose a novel strategy for training neural network acoustic models based on adversarial training which makes use of environment labels during training. By adjusting the parameters of the initial layers of the network adversarially with respect to a domain classifier trained to recognize the recording environments, we enforce better invariance to the diversity of recording conditions. We provide a motivating study on the mechanism by which a deep network learns environmental invariance, and discuss some relations with existing approaches for improving the robustness of DNN models. The proposed multi-domain adversarial training is evaluated on an end-to-end speech recognition task based on the AMI meeting corpus, achieving a relative character error rate reduction of +3.3% with respect to a conventional multi-condition trained baseline and +25.4% with respect to a clean-trained baseline.
Objectives: Atrial fibrillation (AF) is a common heart rhythm disorder associated with deadly and debilitating consequences including heart failure, stroke, poor mental health, reduced quality of life and death. Having an automatic system that diagnoses various types of cardiac arrhythmias would assist cardiologists to initiate appropriate preventive measures and to improve the analysis of cardiac disease. To this end, this paper introduces a new approach to detect and classify automatically cardiac arrhythmias in electrocardiograms (ECG) recordings. Methods: The proposed approach used a combination of Convolution Neural Networks (CNNs) and a sequence of Long Short-Term Memory (LSTM) units, with pooling, dropout and normalization techniques to improve their accuracy. The network predicted a classification at every 18th input sample and we selected the final prediction for classification. Results were cross-validated on the Physionet Challenge 2017 training dataset, which contains 8,528 single lead ECG recordings lasting from 9s to just over 60s. Results: Using the proposed structure and no explicit feature selection, 10-fold stratified cross-validation gave an overall F-measure of 0.83.10-0.015 on the held-out test data (mean-standard deviation over all folds) and 0.80 on the hidden dataset of the Challenge entry server.
We propose a new generalization bound for domain adaptation when there are multiple source domains with labeled instances and one target domain with unlabeled instances. The new bound has an interesting interpretation and reduces to an existing bound when there is only one source domain. Compared with existing bounds, the new bound does not require expert knowledge about the target distribution, nor the optimal combination rule for multisource domains. Interestingly, our theory also leads to an efficient implementation using adversarial neural networks: we show how to interpret it as learning feature representations that are invariant to the multiple domain shifts while still being discriminative for the learning task. To this end, we propose two models, both of which we call multisource domain adversarial networks (MDANs): the first model optimizes directly our bound, while the second model is a smoothed approximation of the first one, leading to a more data-efficient and task-adaptive model. The optimization tasks of both models are minimax saddle point problems that can be optimized by adversarial training. To demonstrate the effectiveness of MDANs, we conduct extensive experiments showing superior adaptation performance on three real-world datasets: sentiment analysis, digit classification, and vehicle counting.
We introduce a new representation learning approach for domain adaptation, in
which data at training and test time come from similar but different
distributions. Our approach is directlyinspired by the theory on domain
adaptation suggesting that, for effective domain transfer to be achieved,
predictions must be made based on features that cannot discriminate between the
training (source) and test (target) domains.
The approach implements this idea in the context of neural network
architectures that are trained on labeled data from the source domain and
unlabeled data from the target domain (no labeled target-domain data is
necessary). As the training progresses, the approach promotes the emergence of
features that are (i) discriminative for the main learning task on the source
domain and (ii) indiscriminate with respect to the shift between the domains.
We show that this adaptation behaviour can be achieved in almost any
feed-forward model by augmenting it with few standard layers and a new gradient
reversal layer. The resulting augmented architecture can be trained using
standard backpropagation and stochastic gradient descent, and can thus be
implemented with little effort using any of the deep learning packages.
We demonstrate the success of our approach for two distinct classification
problems (document sentiment analysis and image classification), where
state-of-the-art domain adaptation performance on standard benchmarks is
achieved. We also validate the approach for descriptor learning task in the
context of person re-identification application.
Supervised and interpatient classification of heart beats is primordial in many applications requiring long-term monitoring of the cardiac function. Several classification models able to cope with the strong class unbalance and a large variety of feature sets have been proposed for this task. In practice, over 200 features are often considered, and the features retained in the final model are either chosen using domain knowledge or an exhaustive search in the feature sets without evaluating the relevance of each individual feature included in the classifier. As a consequence, the results obtained by these models can be suboptimal and difficult to interpret. In this work, feature selection techniques are considered to extract optimal feature subsets for state-of-the-art ECG classification models. The performances are evaluated on real ambulatory recordings and compared to previously reported feature choices using the same models. Results indicate that a small number of individual features actually serve the classification and that better performances can be achieved by removing useless features.
Multitask Learning is an approach to inductive transfer that improves learning for one task by using the information contained in the training signals of other related tasks. It does this by learning tasks in parallel while using a shared representation; what is learned for each task can help other tasks be learned better. In this thesis we demonstrate multitask learning for a dozen problems. We explain how multitask learning works and show that there are many opportunities for multitask learning in real domains. We show that in some cases features that would normally be used as inputs work better if used as multitask outputs instead. We present suggestions for how to get the most out of multitask learning in artificial neural nets, present an algorithm for multitask learning with case based methods like k nearest neighbor and kernel regression, and sketch an algorithm for multitask learning in decision trees. Multitask learning improves generalization performance, can be applied in many different kinds of domains, and can be used with different learning algorithms. We conjecture there will be many opportunities for its use on real world problems. Thesis (Master's).
This paper proposes deep learning methods with signal alignment that facilitate the end-to-end classification of raw electrocardiogram (ECG) signals into heartbeat types, i.e., normal beat or different types of arrhythmias. Time-domain sample points are extracted from raw ECG signals, and consecutive vectors are extracted from a sliding time-window covering these sample points. Each of these vectors comprises the consecutive sample points of a complete heartbeat cycle, which includes not only the QRS complex but also the P and T waves. Unlike existing heartbeat classification methods in which medical doctors extract handcrafted features from raw ECG signals, the proposed end-to-end method leverages a deep neural network (DNN) for both feature extraction and classification based on aligned heartbeats. This strategy not only obviates the need to handcraft the features but also produces optimized ECG representation for heartbeat classification. Evaluations on the MIT-BIH arrhythmia database show that at the same specificity, the proposed patient-independent classifier can detect supraventricular- and ventricular-ectopic beats at a sensitivity that is at least 10% higher than current state-of-the-art methods. More importantly, there is a wide range of operating points in which both the sensitivity and specificity of the proposed classifier are higher than those achieved by state-of-the-art classifiers. The proposed classifier can also perform comparable to patient-specific classifiers, but at the same time enjoys the advantage of patient independency.
We propose a novel adversarial multi-task learning scheme, aiming at actively curtailing the inter-talker feature variability while maximizing its senone discriminability so as to enhance the performance of a deep neural network (DNN) based ASR system. We call the scheme speaker-invariant training (SIT). In SIT, a DNN acoustic model and a speaker classifier network are jointly optimized to minimize the senone (tied triphone state) classification loss, and simultaneously mini-maximize the speaker classification loss. A speaker-invariant and senone-discriminative deep feature is learned through this adversarial multi-task learning. With SIT, a canonical DNN acoustic model with significantly reduced variance in its output probabilities is learned with no explicit speaker-independent (SI) transformations or speaker-specific representations used in training or testing. Evaluated on the CHiME-3 dataset, the SIT achieves 4.99% relative word error rate (WER) improvement over the conventional SI acoustic model. With additional unsupervised speaker adaptation, the speaker-adapted (SA) SIT model achieves 4.86% relative WER gain over the SA SI acoustic model.
We note that common implementations of adaptive gradient algorithms, such as Adam, limit the potential benefit of weight decay regularization, because the weights do not decay multiplicatively (as would be expected for standard weight decay) but by an additive constant factor. We propose a simple way to resolve this issue by decoupling weight decay and the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam, and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). We also demonstrate that longer optimization runs require smaller weight decay values for optimal results and introduce a normalized variant of weight decay to reduce this dependence. Finally, we propose a version of Adam with warm restarts (AdamWR) that has strong anytime performance while achieving state-of-the-art results on CIFAR-10 and ImageNet32x32. Our source code is available at https://github.com/loshchil/AdamW-and-SGDW
We propose a new framework for estimating generative models via an adversarial process, in which we simultaneously train two models: a generative model G that captures the data distribution, and a discriminative model D that estimates the probability that a sample came from the training data rather than G. The training procedure for G is to maximize the probability of D making a mistake. This framework corresponds to a minimax two-player game. In the space of arbitrary functions G and D, a unique solution exists, with G recovering the training data distribution and D equal to 1/2 everywhere. In the case where G and D are defined by multilayer perceptrons, the entire system can be trained with backpropagation. There is no need for any Markov chains or unrolled approximate inference networks during either training or generation of samples. Experiments demonstrate the potential of the framework through qualitative and quantitative evaluation of the generated samples.
We introduce a representation learning approach for domain adaptation, in which data at training and test time come from similar but different distributions. Our approach is directly inspired by the theory on domain adaptation suggesting that, for effective domain transfer to be achieved, predictions must be made based on features that cannot discriminate between the training (source) and test (target) domains. The approach implements this idea in the context of neural network architectures that are trained on labeled data from the source domain and unlabeled data from the target domain (no labeled target-domain data is necessary). As the training progresses, the approach promotes the emergence of features that are (i) discriminative for the main learning task on the source domain and (ii) indiscriminate with respect to the shift between the domains. We show that this adaptation behavior can be achieved in almost any feed-forward model by augmenting it with few standard layers and a new Gradient Reversal Layer. The resulting augmented architecture can be trained using standard backpropagation, and can thus be implemented with little effort using any of the deep learning packages. We demonstrate the success of our approach for image classification, where state-of-the-art domain adaptation performance on standard benchmarks is achieved. We also validate the approach for descriptor learning task in the context of person re-identification application.
An electrocardiogram (ECG) measures the electric activity of the heart and has been widely used for detecting heart diseases due to its simplicity and non-invasive nature. By analyzing the electrical signal of each heartbeat, i.e., the combination of action impulse waveforms produced by different specialized cardiac tissues found in the heart, it is possible to detect some of its abnormalities. In the last decades, several works were developed to produce automatic ECG-based heartbeat classification methods. In this work, we survey the current state-of-the-art methods of ECG-based automated abnormalities heartbeat classification by presenting the ECG signal preprocessing, the heartbeat segmentation techniques, the feature description methods and the learning algorithms used. In addition, we describe some of the databases used for evaluation of methods indicated by a well-know standard developed by the Association for the Advancement of Medical Instrumentation (AAMI) and described in ANSI/AAMI EC57:1998/(R)2008 (ANSI/AAMI, 2008). Finally, we discuss limitations and drawbacks of the methods in the literature presenting concluding remarks and future challenges, and also we propose an evaluation process workflow to guide authors in future works.
Deep neural nets with a large number of parameters are very powerful machine learning systems. However, overfitting is a serious problem in such networks. Large networks are also slow to use, making it difficult to deal with overfitting by combining the predictions of many different large neural nets at test time. Dropout is a technique for addressing this problem. The key idea is to randomly drop units (along with their connections) from the neural network during training. This prevents units from co-adapting too much. During training, dropout samples from an exponential number of different "thinned" networks. At test time, it is easy to approximate the effect of averaging the predictions of all these thinned networks by simply using a single unthinned network that has smaller weights. This significantly reduces overfitting and gives major improvements over other regularization methods. We show that dropout improves the performance of neural networks on supervised learning tasks in vision, speech recognition, document classification and computational biology, obtaining state-of-the-art results on many benchmark data sets. © 2014 Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever and Ruslan Salakhutdinov.
Top-performing deep architectures are trained on massive amounts of labeled
data. In the absence of labeled data for a certain task, domain adaptation
often provides an attractive option given that labeled data of similar nature
but from a different domain (e.g. synthetic images) are available. Here, we
propose a new approach to domain adaptation in deep architectures that can be
trained on large amount of labeled data from the source domain and large amount
of unlabeled data from the target domain (no labeled target-domain data is
necessary).
As the training progresses, the approach promotes the emergence of "deep"
features that are (i) discriminative for the main learning task on the source
domain and (ii) invariant with respect to the shift between the domains. We
show that this adaptation behaviour can be achieved in almost any feed-forward
model by augmenting it with few standard layers and a simple new gradient
reversal layer. The resulting augmented architecture can be trained using
standard backpropagation. Overall the whole approach can be implemented with
little effort using any of the deep-learning packages.
Principal component analysis (PCA) and wavelet transform (WT) are two powerful techniques for feature extraction. In addition, fuzzy c-means clustering (FCM) is among considerable techniques for data reduction. In other words, the aim of using FCM is to decrease the number of segments by grouping similar segments in training data. In this paper, four different structures, FCM-NN, PCA-NN, FCM-PCA-NN and WT-NN, are formed by using these two techniques and fuzzy c-means clustering. In addition, FCM-PCA-NN is the new method proposed in this paper for classification of ECG. This paper presents a comparative study of the classification accuracy of ECG signals by using these four structures for computationally efficient early diagnosis. Neural network used in this study is a well-known neural network architecture named as multi-layered perceptron (MLP) with backpropagation training algorithm. The ECG signals taken from MIT-BIH ECG database, are used in training to classify 10 different arrhythmias. These are normal sinus rhythm, sinus bradycardia, ventricular tachycardia, sinus arrhythmia, atrial premature contraction, paced beat, right bundle branch block, left bundle branch block, atrial fibrillation and atrial flutter. Before testing, the proposed structures are trained by backpropagation algorithm. All of the structures are tested by using experimental ECG records of 92 patients (40 male and 52 female, average age is 39.75 ± 19.06). The test results suggest that FCM-PCA-NN structure can generalize better than PCA-NN and is faster than NN, FCM-NN and WT-NN.
This paper illustrates the use of combined neural network model to guide model selection for classification of electrocardiogram (ECG) beats. The ECG signals were decomposed into time-frequency representations using discrete wavelet transform and statistical features were calculated to depict their distribution. The first level networks were implemented for ECG beats classification using the statistical features as inputs. To improve diagnostic accuracy, the second level networks were trained using the outputs of the first level networks as input data. Four types of ECG beats (normal beat, congestive heart failure beat, ventricular tachyarrhythmia beat, atrial fibrillation beat) obtained from the Physiobank database were classified with the accuracy of 96.94% by the combined neural network. The combined neural network model achieved accuracy rates which were higher than that of the stand-alone neural network model.
The MIT-BIH Arrhythmia Database was the first generally available set of standard test material for evaluation of arrhythmia detectors, and it has been used for that purpose as well as for basic research into cardiac dynamics at about 500 sites worldwide since 1980. It has lived a far longer life than any of its creators ever expected. Together with the American Heart Association Database, it played an interesting role in stimulating manufacturers of arrhythmia analyzers to compete on the basis of objectively measurable performance, and much of the current appreciation of the value of common databases, both for basic research and for medical device development and evaluation, can be attributed to this experience. In this article, we briefly review the history of the database, describe its contents, discuss what we have learned about database design and construction, and take a look at some of the later projects that have been stimulated by both the successes and the limitations of the MIT-BIH Arrhythmia Database.
Heart disease and stroke statistics 2018 at-aglance
- association
A. H. Association, "Heart disease and stroke statistics 2018 at-aglance," 2017.
MIT-BIH Database | Kaggle
- mondejar