Bartosz Puszkarski

Wearable and mobile electrocardiography (ECG) has rapidly expanded access to rhythm monitoring outside of clinical settings, but the single- or few-lead nature of these recordings and their susceptibility to motion artefacts complicate clinical-grade interpretation. This review synthesises recent machine learning approaches for wearable ECG across three problem domains—signal-quality assessment (SQA), single-lead interpretation, and multi-lead interpretation—and covers preprocessing pipelines, quality-handling strategies, and 18 benchmark datasets. Across studies, no single architectural paradigm consistently outperforms others when evaluated fairly: classical feature-based methods remain competitive for rhythm tasks under resource constraints, while deep discriminative models show stronger cross-database generalisation when training data are sufficient. The most consistent predictor of reported performance is dataset characteristics—particularly scale, demographic composition, and whether signals were acquired from wearable devices in free-living conditions—rather than architectural choice. Three recurring evaluation deficiencies are identified: subject-level data leakage inflating classification metrics, SQA performance reported without acceptance rates, and reconstructed ECG leads validated solely by waveform similarity, without downstream diagnostic concordance. A structured comparative summary of all the methods reviewed is provided. The analysis clarifies that future progress depends primarily on evaluation rigour, representative datasets, and reproducible preprocessing rather than on architectural novelty alone.

Cardiovascular diseases (CVDs), which remain globally one of the most common causes of death, are usually diagnosed based on the electrocardiogram (ECG) signal. To support human experts, modern deep-learning models are used for CVD classification problems as an early warning. This article proposes a novel multi-branch architecture focused on processing various modalities of the ECG signal in parallel branches, replacing typical single-input architectures. Each branch is given separate input in the form of the raw signal, domain knowledge, the wavelet transform of the signal, or the signal with drift removed. The proposed method is based on deep-learning core models that can incorporate various modern neural networks. It was thoroughly tested on N-BEATS, LSTM, and GRU neural networks. The proposed architecture allows the retention of the speed of the neural network. At the same time, the combination of independently computed branches improves model performance, which finally exceeds the performance obtained by classical single-branch architectures.

Cardiovascular diseases (CVDs), which remain globally one of the most common causes of death, are usually diagnosed based on the electrocardiogram (ECG) signal. To support human experts, modern deep-learning models are used for CVD classification problems as an early warning. This article proposes a novel multi-branch architecture focused on processing various modalities of the ECG signal in parallel branches replacing typical single-input architectures. Each branch is given separate input in the form of the raw signal, domain knowledge, the wavelet transform of the signal, or the signal with drift removed. The proposed method is based on deep-learning core models that can incorporate various modern neural networks. It was thoroughly tested on N-BEATS, LSTM, and GRU neural networks. The proposed architecture allows the retention of the speed of the neural network. At the same time, the combination of independently computed branches improves model performance, which finally exceeds the performance obtained by classical single-branch architectures.

Objective. The primary purpose of this work is to analyze the ability of N-BEATS architecture for the problem of prediction and classification of electrocardiogram (ECG) signals. To achieve this, performance comparison with various types of other SotA (state-of-the-art) recurrent neural network architectures commonly used for such problems is conducted. Approach. Four architectures (N-BEATS, LSTM, LSTM with peepholes, GRU) were tested for performance and dimension reduction problems for different number of leads (2, 3, 4, 6, 12), both in variants consisting of blended branches, allowing retaining accuracy while reducing the computational capacity needed. The analysis was performed on datasets and using metrics from Challenges in Cardiology (CinC) 2021 competition. Main results. Best results were achieved for LSTM with peepholes, then LSTM, GRU and the worst for N-BEATS (challenge metrics respectively: 0.42, 0.40, 0.39, 0.35; for times: 0.0395 s, 0.0036 s, 0.0027 s, 0.0002 s). Commonly used LSTM outperforms N-BEATS in terms of multi-label classification, data set resilience, and obtained challenge metrics. Still, N-BEATS can obtain acceptable results for 2 lead classification (metric of 0.35 for N-BEATS and 0.38 for other networks) and outperforms other solutions in terms of complexity and speed. Significance. This paper features a novel approach of using the N-BEATS, which was previously used only for forecasting, to classify ECG signals with success. While N-BEATS multi-label classification capacity is lower than LSTM, its speed obtaining results with a reduced number of leads (faster by one to two degrees of magnitude) allows for arrhythmias detection and classification while using off-the-shelf wearable devices (Holter monitors, sport bands, etc).