A Comprehensive Examination of Single Channel ECG, Public Datasets, and Experimental Protocol for the Diagnosis and Detection of Arrhyhtmias
This follow-up Scientific report examines the use of single-channel ECGs, public datasets, and an experimental protocol for arrhythmia detection and diagnosis. It evaluates data quality, methodology, and real-world applicability to support future development in digital cardiac monitoring.
Jul 25, 2025
Abstract:
This comprehensive report encompasses a three-phase investigation into the effectiveness of single-channel electrocardiogram (ECG) for arrhythmia detection, with a focus on leveraging publicly available datasets. Initially, the research carefully examines several databases that have become widely used in scientific literature. The study's second phase uses sophisticated signal processing methods to look deeper into the categorization of various arrhythmias. Furthermore, the effectiveness of deep learning techniques is analyzed in detail, including how well they work in real-time for single-channel ECG classification. The report continues its investigation of experimental protocols designed especially for smartwatch-based ECG techniques targeted at arrhythmia identification in the third stage. The objective of this stage is to evaluate the viability and efficiency of incorporating single-channel ECG monitoring into wearable technology, such as smartwatches. By traversing these three stages, the report seeks to provide a nuanced understanding of the capabilities and limitations of single-channel ECG in the realm of arrhythmia detection. The findings of this review hold implications for both research methodologies and the practical implementation of ECG based healthcare applications, particularly in the context of emerging wearable technologies.
Part 1
Identification and Analysis of Publicly Available Datasets for Single Channel ECG with Specified Arrhythmias
The 2017 PhysioNet/CinC Challenge:
The 2017 PhysioNet/CinC Challenge dataset employed an AliveCor hand-held device with a single lead attached to the patient's chest. The study included 8,528 subjects and recorded the data at a sampling rate of 250 Hz. The study concentrated on brief Electrocardiogram (ECG) recordings, usually lasting between 10 and 60 seconds, to categorize arrhythmias into groups like Normal, Atrial Fibrillation (AF), Non-AF, and Noisy. This dataset has been extensively used in the study of classifying AF using various signal processing techniques along with Machine Learning and Deep Learning [1, 2].
MIT-BIH Arrhythmia database:
48 half-hour snippets of two-channel ambulatory ECG recordings from 47 people investigated by the BIH Arrhythmia Laboratory between 1975 and 1979 are available in the MIT-BIH Arrhythmia Database. A total of 4000 24-hour ambulatory ECG recordings were obtained from a mixed population of patients at Boston's Beth Israel Hospital, consisting of approximately 60% outpatients and 40% inpatients. Of these recordings, 23 were randomly selected, and the remaining 25 recordings were chosen to include less common but clinically significant arrhythmias that would not be well-represented in a small random sample. Over a 10-mV range, the recordings were digitized at 360 samples per second per channel with 11-bit resolution. Each record was independently annotated by two or more cardiologists; conflicts were settled to produce computer-readable reference annotations for each beat (approximately 110,000 annotations in all) included in the database [1,3].
Although this database has two leads and is not a single-channel database, it has been extensively used in the single channel arrhythmia classification as evident in Appendix 1. Table III-V.
Read more: Klick
