Analyzing long wearable electrocardiogram recordings: R-spike detection and cycle pattern analysis

Introduction: This paper considers very long ECGs (electrocardiograms), looking mainly at leads with salient QRS complexes (deflections in an ECG that represent electrical activity generated by ventricular depolarization). The study focuses on healthy or asymptomatic individuals in order to identify ECG pattern dynamics of potential interest to clinicians, especially for prevention and in the context of personalized medicine. Materials and methods: The raw data comes from portable recording devices running for up to twenty-four hours; hence, it tends to be rather noisy, even along segments that might be fairly long. The raw data is intervened minimally: only a linear convolution filter is used or the Fasano–Villani method to correct for the wandering baseline. An automatic procedure is used to detect zones devoid of meaningful ECG signal. In order to detect the R-spikes in the good zones of the ECG, a new methodology is offered, taking advantage of the fact that the amplitude of the R-spike is within a narrow range with respect to the height of the wandering line. The proposed method compares favorably with the Pan–Tompkins algorithm. The cycles of the good zones are grouped in blocks, which are then classified as coherent and non-coherent. Certain averages of cycles comprising coherent blocks are the detected patterns. All computations are in Matlab™ (version 2025a). Results: Graphics devices are presented that allow for the inspection of large sequences of patterns exhibiting the shape change in the morphology of the cycles as based on their representing patterns. Conclusions: The paper demonstrates an automatic procedure to detect possible anomalies in portable ECG recordings.