ECG Arrhythmia Classification Using Transfer Learning from 2- Dimensional Deep CNN Features

Abstract

Due to the recent advances in the area of deep learning, it has been demonstrated that a deep neural network, trained on a huge amount of data, can recognize cardiac arrhythmias better than cardiologists. Moreover, traditionally feature extraction was considered an integral part of ECG pattern recognition; however, recent findings have shown that deep neural networks can carry out the task of feature extraction directly from the data itself. In order to use deep neural networks for their accuracy and feature extraction, high volume of training data is required, which in the case of independent studies is not pragmatic. To arise to this challenge, in this work, the identification and classification of four ECG patterns are studied from a transfer learning perspective, transferring knowledge learned from the image classification domain to the ECG signal classification domain. It is demonstrated that feature maps learned in a deep neural network trained on great amounts of generic input images can be used as general descriptors for the ECG signal spectrograms and result in features that enable classification of arrhythmias. Overall, an accuracy of 97.23 percent is achieved in classifying near 7000 instances by ten-fold cross validation.

Classification of Arrhythmia by Using Deep Learning with 2-D ECG Spectral Image Representation

“…We also used 1-D ECG signals as input to the CNN model used in experiments and achieved a classification accuracy of 97.80%. In recent years, 2-D CNN models have also been used, by converting the 1-D ECG signals to 2-D representation, with noticeable performance [16]. Towards this end, the proposed model was based on a 2-D representation of the ECG data to efficiently apply 2-D CNN models and benefit from the flexibility of data augmentation in such methods.…”

“…The techniques presented in literature have been applied to smaller datasets; however, for the purpose of generalization, the performance should be tested on larger datasets. There are methods reported that use 2-D ECG signals [16,45]; however, to the best of our knowledge, there are not clear details on how the 1-D ECG signal is converted to 2-D images for using 2-D CNN models. Most methods have been tested on only a few types of arrhythmia and must be evaluated on all major types of arrhythmia.…”

“…The 2-D spectrogram is similar to hyper-spectral and multi-spectral images (MSI), which have diverse applications in remote sensing and clinical diagnosis, including spectral un-mixing, ground cover classification and matching, mineral exploration, medical image classification, change detection, synthetic material identification, target detection, activity recognition, and surveillance [9][10][11][12][13][14][15]. The 2-D matrix of spectrogram coefficients could be useful for extracting robust features for representation of a cardiac ECG signal [16]. This representation could allow the application of CNN architectures (designed to operate on 2-D inputs) for development of automated systems related to CVDs.…”

DeepArrNet: An Efficient Deep CNN Architecture for Automatic Arrhythmia Detection and Classification From Denoised ECG Beats

“…A number of deep learning-based approaches have also been adopted recently for arrhythmia classification [14]- [26]. In [14]- [19], 1D convolutional neural network (CNN) and in [23]- [25], recurrent neural network (RNN) and LSTM network are employed while in [20]- [22], 2D CNN is used by converting 1D ECG beats into 2D images. Most of the deep learning-based approaches are facing some common issues: (1) raw ECG data collected from patients are being directly fed to the deep neural network making the classification process complicated due to presence of various low and high-frequency noises, (2) for dealing with 1D ECG signal, data augmentation is not necessarily used and even if it is used, natural variational pattern of ECG isn't properly captured or preserved, and (3) most of the approaches use very deep CNNs with large number of parameters that not only increase the computational complexity but also lead to overfitting the model to training data.…”

ECG-signal multi-classification model based on squeeze-and-excitation residual neural networks

“…In contrast to the methods mentioned thus far, some scholars have used short-term Fourier and wavelet transforms to convert ECG data into two-dimensional (frequency, time) data and used them as input for a deep neural network. Salem et al [14] used the transformation "spectrogram" from a one-dimensional (1D) ECG signal from the MIT-BIH dataset and the European ST-T dataset to make 2D images. They also used a 161-layer DenseNet, pre-trained on millions of images, to extract abstract information and then applied an SVM for four-class classi cation (Normal Sinus/Atrial Fibrillation and Flutter/Ventricular Fibrillation/ST Segment Change).…”