Piece-wise Matching Layer in Representation Learning for ECG Classification

Abstract

This paper proposes piece-wise matching layer as a novel layer in representation learning methods for electrocardiogram (ECG) classification. Despite the remarkable performance of representation learning methods in the analysis of time series, there are still several challenges associated with these methods ranging from the complex structures of methods, the lack of generality of solutions, the need for expert knowledge, and large-scale training datasets. We introduce the piece-wise matching layer that works based on two levels to address some of the aforementioned challenges. At the first level, a set of morphological, statistical, and frequency features and comparative forms of them are computed based on each periodic part and its neighbors. At the second level, these features are modified by predefined transformation functions based on a receptive field scenario. Several scenarios of offline processing, incremental processing, fixed sliding receptive field, and event-based triggering receptive field can be implemented based on the choice of length and mechanism of indicating the receptive field. We propose dynamic time wrapping as a mechanism that indicates a receptive field based on event triggering tactics. The proposed layer outperforms the common kernels and pooling on high-level representations of ECG classification. This layer offers several advantages including 1) it enables the learning from few instances for training and handling long quasi-periodic time series, 2) the layer followed by a simultaneous multiple feature tracking mechanism that does not require any annotated data or expert knowledge in quasi-periodic time series, and 3) the proposed framework can provide and handle several scenarios like online processing and dynamic data-driven application systems. To evaluate the performance of this method in time series analysis, we applied the proposed layer in two publicly available datasets of PhysioNet competitions in 2015 and 2017 where the input data is ECG signal. These datasets present a good example of quasiperiodic time series signals with a small number of training samples, highly imbalanced classes, long time series recordings, and non-linear and complex patterns. We compared the performance of our method against a variety of known tuned methods from expert knowledge, machine learning, deep learning methods, and the combination of them. The proposed approach improves the state of the art in two known completions 2015 and 2017 around 4% and 7% correspondingly while it does not rely on in advance knowledge of the classes or the possible places of arrhythmia.