While many ECG classification models achieve high accuracy, few address key constraints like patient-independent data partitioning and hardware limitations for embedded deployment.
Systematic Review (n=122)
Highlights the need for standardized performance reporting and addressing hardware constraints in ECG classification models for real-world applicability.
Abstract Background The classification of electrocardiogram (ECG) signals is essential for early arrhythmia detection. However, many studies fail to follow standardization protocols, leading to inconsistencies in performance evaluation and real-world applicability. Additionally, hardware constraints crucial for deployment in pacemakers, Holter monitors, and wearable ECG patches are often overlooked, limiting practical use in resource-constrained environments. Objective This review systematically analyzes ECG classification research published between 2017 and 2024, focusing on studies that meet the E3C (Embedded, Clinical, and Comparative Criteria). Methods These criteria include adherence to the inter-patient paradigm, compliance with AAMI recommendations, and evaluation of model feasibility for embedded deployment. We conduct a comparative analysis of accuracy, inference time, energy consumption, and memory usage, identifying state-of-the-art methods that align with the E3C. Results Our findings show that while many models achieve high accuracy, few address key constraints like patient-independent data partitioning and hardware limitations. Conclusions Finally, we propose standardized performance reporting to enable fair comparisons and enhance real-world applicability. By addressing these gaps, this study aims to guide research toward more robust and clinically viable ECG classification models.
Silva et al. (Fri,) conducted a systematic review in Arrhythmia (n=122). ECG classification models was evaluated on Adherence to inter-patient paradigm, AAMI recommendations, and embedded deployment feasibility. While many ECG classification models achieve high accuracy, few address key constraints like patient-independent data partitioning and hardware limitations for embedded deployment.