Open research questions in ECG Monitoring and Analysis

What the literature leaves open

• The lack of effective pattern localization methods for low signal-to-noise ratio conditions. The deterioration of matched filter effectiveness in such conditions. The need for a novel framework to address these challenges.

  Machine Intelligence on the Edge: Interpretable Cardiac Pattern Localization Using Reinforcement Learning · 2026 · DOI

• Studies involving a more diverse patient population and longer monitoring periods in real-world ambulatory settings are needed.

  A Deep Neural Network for Interpreting Wearable Electrocardiogram Data in Atrial Fibrillation: Prospective Observational Diagnostic Accuracy Study · 2026 · DOI

• Studies involving a more diverse patient population are needed to validate the generalizability of the approach. Longer monitoring periods in real-world ambulatory settings are needed to validate the long-term performance of the approach. Research on the use of the system in different real-world settings is needed.

  A Deep Neural Network for Interpreting Wearable Electrocardiogram Data in Atrial Fibrillation: Prospective Observational Diagnostic Accuracy Study · 2026 · DOI

• The need to address the substantial imbalance among the disease labels. The challenge of predicting echocardiographic abnormalities from ECGs and clinical variables.

  EchoGraph-FusionNet: Adaptive Disease-Graph and Mixture-of-Experts Learning for Predicting Echocardiographic Abnormalities from ECGs · 2026 · DOI

• Further validation through repeated experiments and corresponding uncertainty analysis is required. Evaluation of the model's real-time performance and computational efficiency is needed. Investigation of the model's applicability to other ECG datasets and classification tasks.

  Hybrid Parallel Twin-Tower Model Advances Arrhythmia Detection in ECG Signals: Integrating Convolutional Neural Networks and Kolmogorov–Arnold Networks · 2026 · DOI

• The current methods for arrhythmia detection have limitations, and there is a need for more accurate and transparent approaches. The existing R peak detection methods have shortcomings that need to be addressed. There is a need for interpreting the decisions of deep learning models in a form understandable to clinicians.

  Towards Transparent AI in Medicine: ECG-Based Arrhythmia Detection with Explainable Deep Learning · 2025 · DOI

• The use of fixed frequency bands is a compromise between enhanced noise immunity and ECG distortions. The limitations imposed by analog filtering need to be overcome.

  Design of High-Pass and Low-Pass Active Inverse Filters to Compensate for Distortions in RC-Filtered Electrocardiograms · 2025 · DOI

• The need to improve the denoising and detrending abilities of CNNs. The challenge of addressing the class imbalance problem in arrhythmia classification.

  Preprocessing-Free Convolutional Neural Network Model for Arrhythmia Classification Using ECG Images · 2025 · DOI

• The dataset is slightly imbalanced in terms of interference class. The study uses a limited number of participants (10 healthy volunteers). The experiment was designed for a specific target group (IRS members, specifically firefighters).

  Convolutional Neural Network for Automatic Detection of Segments Contaminated by Interference in ECG Signal · 2025 · DOI

• Future research should focus on model optimization and dataset expansion to enhance diagnostic capabilities and interpretability.

  A deep learning model based on combining surface and esophageal ECG data for diagnosis of paroxysmal supraventricular tachycardia · 2026 · DOI

• Although the evaluation of left ventricular ejection fraction (LVEF) in patients with atrial fibrillation (AF) or atrial flutter (AFL) is crucial for appropriate medical management, the prediction of reduced LVEF (<50%) with AF/AFL electrocardiograms (ECGs) lacks evidence.

  Prediction of reduced left ventricular ejection fraction using atrial fibrillation or flutter electrocardiograms: A machine-learning study · 2025 · DOI

• Here, the DNN fine-tuning for unknown domains is limited to a single layer, while the remaining DNN model remains unmodified.

  Towards Hardware Supported Domain Generalization in DNN-Based Edge Computing Devices for Health Monitoring · 2024 · DOI

• Prospective, population-based cohorts with standardised ECG acquisition and longitudinal follow-up are still scarce, especially those with adequate representation of prediabetes.

  Electrocardiographic Signatures of Dysglycaemia: Mechanistic Foundations, Digital Biomarkers, and Artificial Intelligence for Non-Invasive Diabetes Risk Stratification · 2026 · DOI

• Extending the SMF framework to other applications that require accurate pattern localization from noise-corrupted signals. Investigating the use of other reinforcement learning algorithms for filter design. Evaluating the performance of the SMF framework on larger datasets.

  Machine Intelligence on the Edge: Interpretable Cardiac Pattern Localization Using Reinforcement Learning · 2026 · DOI

• Traditional graph convolutional networks have a predefined and fixed adjacency matrix during training.

  MST-DGCN: Multi-Scale Temporal–Dynamic Graph Convolutional with Orthogonal Gate for Imbalanced Multi-Label ECG Arrhythmia Classification · 2025 · DOI

• The effective detection of segments with motion-induced and other interferences in ECG recordings obtained during long-term field monitoring remains a relevant research problem. The frequency spectrum of several types of interfering signals overlaps significantly with the ECG signal, making effective filtration impossible without losing much clinically relevant information.

  Convolutional Neural Network for Automatic Detection of Segments Contaminated by Interference in ECG Signal · 2025 · DOI

• The outcome demonstrates that embedded microcontrollers, when paired with tailored TinyML models, can perform sophisticated biomedical architecture alterations selection during analytics without infrastructure.

  Edge-Based Tinyml Framework for Intelligent Cardiac Drug Response Monitoring Using Embedded Hardware Software Co-Design · 2026 · DOI

• Investigate whether continuous estimates provide incremental value for risk stratification beyond binary disease probabilities, - Explore the potential advantage of the multitask formulation in clinical screening where patients frequently have multiple coexisting…

  EchoGraph-FusionNet: Adaptive Disease-Graph and Mixture-of-Experts Learning for Predicting Echocardiographic Abnormalities from ECGs · 2026 · DOI

• The limitations of traditional models in generalization pose significant challenges to classification accuracy. The diversity of ECG signals and data imbalance pose significant challenges to classification accuracy.

  Hybrid Parallel Twin-Tower Model Advances Arrhythmia Detection in ECG Signals: Integrating Convolutional Neural Networks and Kolmogorov–Arnold Networks · 2026 · DOI

• Acquiring multi-lead ECG signals for a more comprehensive joint analysis, - Developing a more efficient and accurate MI detection and localization model, - Combining various data types and leveraging deep learning…

  A Multi-Domain Feature Fusion CNN for Myocardial Infarction Detection and Localization · 2025 · DOI

• Traditional MI detection and localization techniques rely on single-domain features, which proves challenging for accurate detection and localization. The inability of single-domain features to fully capture the complexity and variability in cardiac electrical activity.

  A Multi-Domain Feature Fusion CNN for Myocardial Infarction Detection and Localization · 2025 · DOI

• Further research is needed to explore the application of the proposed filters in different ECG acquisition systems, - The development of more advanced filters that can adapt to different noise environments and ECG signal characteristics, - Investigation of the effects of the proposed filters on the diagnosis of various cardiovascular…

  Design of High-Pass and Low-Pass Active Inverse Filters to Compensate for Distortions in RC-Filtered Electrocardiograms · 2025 · DOI

• Enhancing the denoising and detrending abilities of CNNs by incorporating residual blocks and attention mechanisms, - Investigating the effects of noise and trends on the model’s…

  Preprocessing-Free Convolutional Neural Network Model for Arrhythmia Classification Using ECG Images · 2025 · DOI

• Conventional cloud-based health monitoring solutions suffer from limitations such as high energy usage, communication delays, and privacy concerns. Conventional monitoring approaches rely heavily on periodic clinical visits or cloud-based telemedicine systems, which introduce latency and increase energy consumption.

  Edge-Based Tinyml Framework for Intelligent Cardiac Drug Response Monitoring Using Embedded Hardware Software Co-Design · 2026 · DOI

• The sample size was limited to 13 standard 12-lead ECGs, - The study was an exploratory evaluation, - The number of cases per category was small, which would yield unstable and potentially misleading estimates, - No application programming interface (API) parameters were modified, - No system prompts beyond the standardized evaluation prompt…

  Artificial Intelligence for Biomedical Diagnostics: Diagnostic Accuracy and Reliability of Multimodal Large Language Models in Electrocardiogram Interpretation · 2026 · DOI

Most-cited papers in ECG Monitoring and Analysis

• An artificial intelligence-enabled ECG algorithm for the identification of patients with atrial fibrillation during sinus rhythm: a retrospective analysis of outcome prediction · The Lancet · 2019 · 1,510 citations
• Automatic diagnosis of the 12-lead ECG using a deep neural network · Nature Communications · 2020 · 937 citations
• Deep Learning for ECG Analysis: Benchmarks and Insights from PTB-XL · IEEE Journal of Biomedical and Health Informatics · 2020 · 479 citations
• Deep learning in ECG diagnosis: A review · Knowledge-Based Systems · 2021 · 421 citations
• Detection of Hypertrophic Cardiomyopathy Using a Convolutional Neural Network-Enabled Electrocardiogram · Journal of the American College of Cardiology · 2020 · 395 citations
• Automated Atrial Fibrillation Detection using a Hybrid CNN-LSTM Network on Imbalanced ECG Datasets · Biomedical Signal Processing and Control · 2020 · 319 citations
• Review of noise removal techniques in ECG signals · IET Signal Processing · 2020 · 313 citations
• Application of artificial intelligence to the electrocardiogram · European Heart Journal · 2021 · 306 citations
• Prediction of mortality from 12-lead electrocardiogram voltage data using a deep neural network · Nature Medicine · 2020 · 306 citations
• Classification of 12-lead ECGs: the PhysioNet/Computing in Cardiology Challenge 2020 · Physiological Measurement · 2020 · 295 citations

Most recent work

• Artificial Intelligence–Enhanced Electrocardiogram Models for Detection of Left Ventricular Dysfunction · JACC Advances · 2026
• Teaching multimodal LLMs to comprehend 12-lead electrocardiographic images · npj Digital Medicine · 2026
• Artificial Intelligence Electrocardiogram and Left Ventricular Systolic Dysfunction in Kenya · JAMA Cardiology · 2026
• Multisite, External Validation of an AI-Enabled ECG Algorithm for Detection of Low Ejection Fraction · JACC Advances · 2026
• Performance of Artificial Intelligence–Powered ECG Analysis in Suspected ST-Segment Elevation Myocardial Infarction · JACC Advances · 2026
• AI-Enhanced Electrocardiogram for Detection of Occlusive Myocardial Infarction in High-Risk Non–ST-Segment Elevation Acute Coronary Syndrome · JACC Advances · 2026
• An ECG biomarker for sudden cardiac death discovered with deep learning · Nature · 2026
• Wearable device derived electrocardiographic age and its association with atrial fibrillation · npj Digital Medicine · 2026
• Federated learning with homomorphic encryption for secure real time ECG anomaly detection: A multi institutional privacy preserving framework · Biomedical Signal Processing and Control · 2026
• A stability-aware weighted ensemble feature selection framework for structural heart disease detection using multidimensional magnetocardiogram features · Biomedical Signal Processing and Control · 2026

Related topics in Medicine