Open research questions in ECG Monitoring and Analysis

What the literature leaves open

• In this pilot study, it was shown that a dual output phys- iology-informed NARX model maintained accurate BP estimation over a six-hour daily life period, when trained on half of the data of its non-physiology-informed coun- terpart. This was done by leveraging the intrinsic person- alized four-parameter BP–PPG sigmoidal layer into the neural network training process through a dual-output loss function that simultaneously estimates the BP and PPG waveforms. Future work will focus on implementing a more advanced viscoelastic model of BP-PPG relation- ship, leveraging the estimated PPG signal for model reca- libration and validating the model on a larger and more diverse population to ensure reliability across individuals.

  A Dual-Output Physiology-Informed Neural Network Architecture for Continuous Cuffless Blood Pressure Waveform Estimation: A Proof‑of‑Concept · 2026 · DOI

• Recent real-world implementation of automated echocardiographic platforms have demonstrated the feasibility of AI-driven workflows capable of performing view classification, segmentation, and guideline-based reporting within minutes, highlighting the potential of AI to improve the efficiency and scalability of echocardiographic services Such systems illustrate the broader vision of AI-enhanced echocardiography, in which AI can. support multiple steps of the imaging workflow, from image acquisition and automated measurements to interpretation, reporting, and prognostication of clinical outcomes Future developments in AI applied to MR imaging will likely evolve from isolated algorithms toward integrated multimodal diagnostic frameworks. While most current systems analyze a. single modality, emerging approaches aim to combine electrocardiographic signals, echocardiographic imaging, CMR quantification, and structured clinical data to generate more comprehensive diagnostic profiles and improve risk stratification throughout the [38,39]. course of the disease Despite these advances, important challenges remain before widespread clinical implementation can be achieved. Many currently available models have been developed using retrospective datasets and highly controlled imaging environments, highlighting the need for further prospective validation, standardized imaging acquisition protocols, and improved model interpretability. Addressing these limitations will be essential to ensure the safe and equitable evolution of AI in imaging-based evaluation of MR, and its integration into clinical workflows beyond imaging such as various biomarkers available through the electronic health records [40,41].

  Advances in artificial intelligence for the evaluation of mitral regurgitation · 2026 · DOI

• Although the results obtained in our study support meaningful conclusions that electrolyte disorders can be detected from ECG using artificial intelligence methods, it is important to repeat these findings with larger datasets in order to evaluate their applicability and achieve more effective results.

  Detection of Hypokalemia, Hyponatremia, and Hyperkalemia in Heart Failure Patients Using Artificial Intelligence Techniques via Electrocardiography · 2026 · DOI

• ECG-Fyler consistently outperformed training from scratch and other ECG foundation-model baselines, with the largest gains when labelled data were scarce, and showed strong cross-age and cross-institution generalization from paediatric pretraining to adult external validation.

  An ECG foundation model for generalizable cardiac function prediction across the lifespan · 2026 · DOI

• While artificial intelligence (AI)-based electrocardiogram (ECG) models have shown promise for AS detection, it remains unclear whether they primarily reflect conventional left ventricular hypertrophy (LVH) voltage criteria or capture additional ECG features.

  Does ECG-Based AI Detect Aortic Stenosis Beyond Conventional LVH Criteria? An Analysis of the CLIDAS Database · 2026 · DOI

• Spiking neural networks (SNNs) are promising for low-power edge inference, yet it remains unclear how class-imbalance loss design interacts with RR-interval features in directly trained quantized SNNs, and FPGA validation in this setting is largely unexplored.

  A Cascaded Quantized Spiking Neural Network for Real-Time ECG Arrhythmia Detection on Edge Hardware · 2026 · DOI

• leverages both This study proposes a CNN network integrated with the SRM the feature criterion algorithm, which extraction capabilities of CNN and the SRM algorithm's strong generalization ability for unknown data classification. Test results from the MIT-BIH Arrhythmia Database demonstrate that the proposed network achieves higher accuracy (F1 score: 87.4, accuracy: 88.2%) compared to CNN networks without the SRM algorithm. These results indicate its potential as a diagnostic tool in clinical practice. Future work includes:(1) further testing the accuracy of the proposed model on other data sets to verify the generalization of the proposed model;(2) further exploring other deep learning feature extraction methods. Future work includes: (1) further testing the accuracy of the proposed model on other data sets to verify the generalization of the proposed model, such as incorporating diverse physiological databases to assess robustness across varying patient demographics; (2) further exploring other deep learning feature extraction methods, including attention mechanisms or transformer-based approaches, to potentially improve reduce computational overhead. (3) Investigating the optimization of model hyperparameters and training strategies to enhance efficiency for real-time clinical applications, addressing potential challenges like latency and resource constraints. (4) Collaborating with medical institutions to conduct pilot studies, ensuring the model's practical utility and ethical considerations in diagnostic workflows, while also exploring integration with wearable devices for continuous monitoring.

  ECG Signal Classification Method Based on Structural Risk Minimization · 2026 · DOI

• COMPETENCY IN CLINICAL CARE AND TRANSLATIONAL OUTLOOK: Prospective clinical PROCEDURAL SKILLS: An integrated risk prediction trials evaluating portable, patient-operated ECG device approach combining pre-existing cardiovascular risk, that integrate clinical risk factors with automated ECG symptoms, and portable ECG analysis can accurately analysis are needed to determine its impact on patient stratify patients with chest pain for ACS. Integrating it behavior, symptom-to-decision time, and downstream into portable self-assessment tool with a patientclinical outcomes in ACS. operated, cable-free ECG device potentially can help shorten symptom-to treatment times in ACSs. R E F E R E N C E S 1. Wechkunanukul K, Grantham H, Clark RA. Global review of delay time in seeking medical care for chest pain: an integrative literature review. Aust Crit Care. 2017;30:13–20. 2. Moser DK, Kimble LP, Alberts MJ, et al. Reducing delay in seeking treatment by patients with acute coronary syndrome and stroke: a scientific statement from the American Heart Association Council on Cardiovascular Nursing and Stroke Council. J Cardiovasc Nurs. 2007;22:326– 343. 3. Goldberg RJ, Spencer FA, Fox KA, et al.

  Acute Coronary Syndrome Risk Prediction Using Portable Cable-Free ECG Device Combined With Clinical Risk Assessment · 2026 · DOI

• Weaknesses include no availability of external validation sets with OMI culprit localization. However, we have included three external valida- tion sets that together capture different geographical regions, machine types, years, and labels for OMI, STEMI/NSTEMI, and LBBB. Certain subgroups remain an issue; perimyocarditis still12 raises some confusion for the model, and discrimination of nOMI in the LBBB subgroup is poor but better than chance (Supplementary Table 4, Supplementary Fig. 12). The challenge in separating between OMIs with occlusions in LCX and RCA is not unexpected; the origin of the posterior descending artery is more often the RCA than the LCX (Fig. 1A+B). This discrimination is challenging also for humans using ECGs14. The number of cases in some of the outcome classes is modest (Supplementary Table 1), with quite few ECGs to learn from in training and few ECGs in evaluation leading to some uncertainty. Still, the results across outcome classes are encouraging, but some outcome classes such as the LCX culprits would likely benefit from a larger sample size in a future similar model. The CODE-II dataset lacks paired coronary angiography, so it should be noted that acute coronary occlusion cannot be confirmed for those with a STEMI label.

  A deep learning ECG model for identification and localization of occlusion myocardial infarction · 2026 · DOI

• Despite the effectiveness of the proposed solution, there is still much room for improvement and further advancements in the field Systems for Real-Time ECG Monitoring Research can be conducted to design systems for detecting arrhythmias in real time by monitoring the ECG data in an ongoing fashion and alerting users to any abnormality detected in heart rhythms. Personalized Arrhythmia Predictions Heart rhythms can vary among different patients. It would be interesting to investigate how personalized machine learning algorithms can help increase prediction accuracy Combining Smartwatch Volume 14 Issue 5, May 2026 www.ijser.in Licensed Under Creative Commons Attribution CC BY Paper ID: SE26508173840DOI: https://dx.doi.org/10.70729/SE265081738407 of 8 International Journal of Scientific Engineering and Research (IJSER) ISSN (Online): 2347-3878 SJIF (2025): 8.036 [16] M. G. e. al., “Hybrid Deep Learning Framework for ECG Arrhythmia Classification,” Knowledge-Based Systems, vol. 302, no. --, p. 111234, 2025. - [17] D. K. e. al., “Hybrid CNN-Transformer Model for ECG Arrhythmia Classification,” Scientific Reports, vol. 15, no. 1, p. 92582, 2025. - [18] S. B. e. al., “Hybrid Machine Learning Models for ECG Classification,” PLOS ONE, vol. 20, no. 5, p. e0334607, 2025. - [19] G. a. R. Mark, “The Impact of the MIT-BIH Arrhythmia Database,” IEEE Engineering in Medicine and Biology Magazine, vol. 20, no. 3, pp. 45-50, 2001. [20] M. O. a. R. B. R. P. de Chazal, “Automatic Classification of Heartbeats Using ECG Morphology and Heartbeat Interval Features,” IEEE Transactions on Biomedical Engineering, vol. 51, no. 7, pp. 1196-1206, 2004. [21] R. A. e. al., “Automated Diagnosis of Arrhythmia Using Combination of CNN and ECG Signals,” Information Sciences, no. 145-416, pp. 190-198, 2017. [22] P. R. e. al., “Cardiologist-Level Arrhythmia Detection with Deep Neural Networks,” Nature Medicine, vol. 25, no. 1, pp. 65-69, 2017. - [23] T. I. a. M. G. S. Kiranyaz, “Real-Time Patient-Specific ECG Classification by 1-D CNN,” IEEE Transactions on Biomedical Engineering, vol. 63, no. 3, pp. 664-675, 2016. with the ECG Data from Recent advances in wearable technologies allow tracking of heart rate via ECG sensors embedded into smartwatches. The proposed arrhythmia model could be integrated with those sensors, and this would benefit millions of users. Deep Learning Techniques Application Another area where studies could be carried out is the application of deep learning techniques such as the Long Short Term Memory technique, Transformers, and deep reinforcement learning. Big Data Analysis for Hospitals There are many medical databases available that can be used to train the ECG model.

  Arrhythmia Classification Using Machine Learning Techniques on ECG Data · 2026 · DOI

• While the proposed framework offers a comprehensive and interpretable approach to preliminary cardiac risk assessment, several limitations must be acknowledged to ensure a balanced and realistic perspective. First, the present study is conceptual and does not include empirical validation using real patient cohorts. Although publicly available datasets (e.g., PhysioNet) are discussed for feasibility, the absence of experimental benchmarking limits the immediate clinical applicability of the proposed system. Future work should prioritize quantitative evaluation using standard metrics such as accuracy, sensitivity, specificity, and area under the ROC curve. Second, the framework relies on self-reported symptoms and user-provided inputs, which may be subjective and prone to bias or inaccuracy. In real-world deployments, variability in user understanding, reporting behavior, and health literacy could affect model reliability. Integrating passive sensing (e.g., wearable devices) and validated clinical measurements would help mitigate this limitation. Third, the proposed model assumes availability of multimodal data, including physiological signals such as ECG. However, such data may not always be accessible in low-resource or home settings. Future systems should explore adaptive models that can operate under partial data availability while maintaining acceptable performance. Fourth, although explainability methods such as SHAP and LIME enhance transparency, they do not guarantee clinical interpretability in all cases. Explanations may still be misinterpreted by non-expert users. Therefore, future research should focus on user-centric explanation design, including visualization strategies and clinician-in-the-loop validation. Fifth, the framework currently emphasizes risk estimation rather than definitive diagnosis. There is an inherent risk of false positives and false negatives, both of which carry significant implications. False negatives may delay critical medical intervention, while false positives may increase anxiety and unnecessary healthcare utilization. Future work should incorporate risk calibration and uncertainty quantification to improve decision reliability. Finally, ethical and regulatory considerations remain a critical challenge. Any real-world deployment would require compliance with healthcare regulations, data privacy standards, and clinical validation protocols. Prospective clinical trials and collaborations with medical professionals will be essential to translate this framework into a deployable solution. In terms of future directions, several promising avenues exist. The integration of longitudinal health data could enable personalized risk modeling over time. The use of federated learning may allow privacy-preserving model training across distributed healthcare systems. Additionally, incorporating natural language processing for symptom description and conversational interfaces could improve accessibility and user engagement. JZU - Engineering Science || ISSN:1008-973X Volume-26 | Issue-4https://jzuengineering.org/Page No:289 Overall, while the proposed framework demonstrates strong potential, its transition from conceptual design to clinical tool will require rigorous validation, interdisciplinary collaboration, and careful consideration of ethical and practical constraints.

  AI-Driven Non-Invasive Cardiac Risk Assessment in Women: A Technical Review and Conceptual Framework · 2026 · DOI

• Despite its advantages, the proposed system has certain limitations that need to be considered. One of the primary challenges is the requirement for a large and diverse dataset. Deep learning models generally perform better when trained on extensive datasets, and limited data can affect the model’s ability to generalize across different patient conditions. Another limitation is the high computational cost associated with training deep learning models. The use of multiple architectures, especially hybrid models, requires significant processing power and memory, which may not be readily available in all environments. The current system is also limited to ECG image-based classification, which means it does not directly process raw ECG signal data or multi- lead inputs. This restricts its applicability in more advanced clinical scenarios. Furthermore, the system is not yet clinically deployed, meaning it still requires validation and approval before being used in real-world healthcare settings. 7. Future Enhancement: To overcome the current limitations and improve the system further, several enhancements can be considered in future work. One important direction is the integration of multi-lead ECG data, which can provide more comprehensive information about cardiac activity and improve classification, accuracy. Another promising enhancement is cloud- based deployment, which would allow the system to be accused remotedly and handle large- scale data efficiently. © 2026 The Author(s). Published by IJCOPE Journal. Website: https://ijcope.org/ 5 International Journal of Creative and Open Research in Engineering and Management ISSN: 3108-1754 (Online) Volume 02 Issue 04 April-2026 | Impact Factor: 3.5 This would also reduce the need for high-end local hardware. Additionally, developing a mobile application can make the system more accessible to healthcare professionals, especially in remote or rural areas. In corporating Explainable AI (XAI) techniques is another crucial improvement. This would allow the system to provide clear explanations for its predictions, increasing trust and transparency among medical practitioners. Finally, integrating the system with hospital management systems can streamline workflows and enable seamless data sharing, making the solution more practical for real-world applications. 8. Conclusion: In conclusion, this research demonstrates the effectiveness of deep learning techniques in the classification of ECG signals for detecting heart diseases. By comparing multiple models, including CNN, MobileNet and DenseNet, the study highlights the superior performance of the hybrid MobileNet + LSTM model. This model successfully combines spatial features extraction with temporal pattern recognition, resulting in improved accuracy and reliability. The proposed system not only enhances diagnostic accuracy but also reduces the burden on healthcare professionals by automating the analysis process. Its ability to deliver fast and consistent results makes it a valuable tool for supporting clinical decision-making. Although there are certain limitations, the system shows strong potential for real- world implementation with further improvements and validation. Overall this work represents a meaningful step toward the development of intelligent healthcare solutions, where artificial intelligence can assist doctors in providing faster, more accurate, and more efficient patient care.

  Automated Heart Disease Detection from ECG Signals using a Hybrid Deep Learning Approach · 2026 · DOI

• and hyperparameter tuning. This proactive optimization avoided excessive computational redundancy and ensured deterministic execution. 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

• Past methods tackle issues like spatial–temporal feature extraction, class imbalance and dataset generalisation, but they are limited by a number of shortcomings: the traditional ML models depend on hand-crafted features and thus lack scalability, and the stand-alone deep models (CNNs or LSTMs) do not capitalize on spatial and sequential information simultaneously.

  HybridCardioNet: A CNN-LSTM-Based Deep Learning Framework for ECG Signal Classification and Cardiac Anomaly Detection · 2025 · DOI

• The optimal screening interval and patient population for AF detection via mobile cardiac telemetry in primary care or telehealth settings remains undefined. Paper [1] examined a large cohort with median 15.4 days monitoring, and [9] states 'additional data are needed' for screening recommendations, but no study establishes cost-effective screening protocols for telecommunications-based AF detection.

  Diagnostic yield is dependent on monitoring duration. Insights from a full-disclosure mobile cardiac telemetry system · 2022 · DOI

• No comparative effectiveness study evaluates AF outcomes (symptom control, hospitalization, mortality) when managed via continuous mobile telemetry versus standard episodic monitoring in integrated care pathways. Paper [1] shows diagnostic yield increases with monitoring duration, and [8] demonstrates early rhythm control reduces cardiovascular events, but no study combines continuous telemetry-enabled early intervention with outcome assessment.

  Diagnostic yield is dependent on monitoring duration. Insights from a full-disclosure mobile cardiac telemetry system · 2022 · DOI

• No study integrates real-time mobile cardiac telemetry data with clinical decision-support systems to optimize AF management pathways in telecommunications settings. Paper [1] demonstrates that diagnostic yield increases with monitoring duration, but [9] notes that 'further high quality evidence is necessary' for mHealth-based integrated care implementation, leaving a gap in how continuous telemetry data should inform treatment decisions across care settings.

  Diagnostic yield is dependent on monitoring duration. Insights from a full-disclosure mobile cardiac telemetry system · 2022 · DOI

• Although awareness and improved detection of AF have improved over the last decade as the incidence and prevalence of AF has increased, current trends in using machine learning approaches to diagnose AF are still lacking in precision.

  A model to enhance the atrial fibrillations’ risk detection using deep learning · 2022 · DOI

• Due to the lack of data in the ST elevation categories, the classifier was only trained to identify different types of ST depressions (horizontal, upsloping and downsloping).

  Improving interpretability: combined use of LVQ and ARTMAP in decision support · 2005 · DOI

• Whether deep learning applied to ECGs can deliver individualized, time-resolved, and biologically interpretable risk estimates for incident HFrEF across diverse populations remains uncertain.

  Electrocardiogram-Based Deep Learning for Time-Resolved Prediction of Heart Failure With Reduced Ejection Fraction: A Multinational Study · 2026 · DOI

• By effectively resolving inter patient variability and modeling disease correlations, the proposed framework provides a scalable and accurate solution for diagnosing cardiac conditions in complex clinical environments where annotated data is limited.

  Inter-patient multi-label ECG classification via low-rank adaptation fine-tuned large language models with dynamic graph convolutional network · 2026 · DOI

• If validated prospectively, such models could support lower-cost screening, triage, and longitudinal monitoring to help prioritize downstream echocardiography or cardiac magnetic resonance imaging, particularly in congenital heart disease and other settings where labelled imaging data are limited.

  An ECG foundation model for generalizable cardiac function prediction across the lifespan · 2026 · DOI

• Electrocardiogram (ECG) signal compression suffers of lack of standards for analogue-digital conversion.

  Predictive ECG coding using linear time-invariant models · 2004 · DOI

Most-cited papers in ECG Monitoring and Analysis

• Artificial intelligence-enhanced electrocardiography for accurate diagnosis and management of cardiovascular diseases · Journal of Electrocardiology · 2024 · 103 citations
• CAT-Net: Convolution, attention, and transformer based network for single-lead ECG arrhythmia classification · Biomedical Signal Processing and Control · 2024 · 99 citations
• Healthcare diagnostics with an adaptive deep learning model integrated with the Internet of medical Things (IoMT) for predicting heart disease · Biomedical Signal Processing and Control · 2024 · 96 citations
• A novel approach for denoising electrocardiogram signals to detect cardiovascular diseases using an efficient hybrid scheme · Frontiers in Cardiovascular Medicine · 2024 · 77 citations
• Artificial intelligence-enabled electrocardiogram for mortality and cardiovascular risk estimation: a model development and validation study · The Lancet Digital Health · 2024 · 67 citations
• The accuracy of Gemini, GPT-4, and GPT-4o in ECG analysis: A comparison with cardiologists and emergency medicine specialists · The American Journal of Emergency Medicine · 2024 · 64 citations
• Prediction of incident atrial fibrillation using deep learning, clinical models, and polygenic scores · European Heart Journal · 2024 · 55 citations
• Deep Learning Based Healthcare Method for Effective Heart Disease Prediction · EAI Endorsed Transactions on Pervasive Health and Technology · 2023 · 25 citations
• Fetal electrocardiography and artificial intelligence for prenatal detection of congenital heart disease · Acta Obstetricia Et Gynecologica Scandinavica · 2023 · 18 citations
• Studying accelerated cardiovascular ageing in Russian adults through a novel deep-learning ECG biomarker · Wellcome Open Research · 2021 · 18 citations

Most recent work

• Electrocardiographic Signatures of Dysglycaemia: Mechanistic Foundations, Digital Biomarkers, and Artificial Intelligence for Non-Invasive Diabetes Risk Stratification · Applied Sciences · 2026
• CODE-II: a large-scale dataset for artificial intelligence in ECG analysis · npj Digital Medicine · 2026
• Test-Retest Reliability of Artificial Intelligence-Enhanced Electrocardiography: A Multi-Center Study · medRxiv · 2026
• Simplified Electrocardiogram Criteria to Guide Conduction System Pacing: Is it Possible to Democratize the Procedure? · Canadian Journal of Cardiology · 2026
• HuBERT-ECG as a self-supervised foundation model for broad and scalable cardiac applications · medRxiv · 2026
• Detection of Hypokalemia, Hyponatremia, and Hyperkalemia in Heart Failure Patients Using Artificial Intelligence Techniques via Electrocardiography · Turk Kardiyoloji Dernegi Arsivi-Archives of the Turkish Society of Cardiology · 2026
• A Cascaded Quantized Spiking Neural Network for Real-Time ECG Arrhythmia Detection on Edge Hardware · Sensors · 2026
• Modeling and Evaluating an Intelligent Health Monitoring System for Detecting Atrial Fibrillation · International Journal of Network Dynamics and Intelligence · 2026
• 26-A-17850-ACC ARTIFICIAL INTELLIGENCE-ENABLED ELECTROCARDIOGRAM ANALYSIS TO DETECT DIGOXIN EXPOSURE AND TOXICITY · Journal of the American College of Cardiology · 2026
• A Hybrid Deep Learning Framework using CNN and Cascaded Discrete Wavelet Transform for ECG-Based Cardiovascular Disease Detection · International Scientific Journal of Engineering & Management · 2026

Related topics in Medicine