https://doi.org/10.1007/s00392-025-02625-4
1Justus-Liebig-Universität Giessen Medizinische Klinik I, Kardiologie Bad Nauheim, Deutschland; 2Universitäts-Herzzentrum Freiburg - Bad Krozingen Klinik für Kardiologie und Angiologie Bad Krozingen, Deutschland; 3Karlsruher Institut für Technologie (KIT) Institut für Biomedizinische Technik Karlsruhe, Deutschland
Background and Aims: Physiological circadian changes are known to effect the cardiovascular system. The electrocardiogram (ECG) is a cornerstone of diagnostic evaluation in cardiac diseases. Aim of the study is to potentially investigate daytime dependent changes of ECG signals to discover possible effects of super-imposed circadian and pathological variations in signals and features.
Methods and Results: The PTB-XL database was used as a dataset including raw ECG signals and information on the time of ECG acquisition. It contains 21,799 12-lead ECGs of 18,869 individuals. Only sinus rhythm ECGs without pathologic changes of individuals aged 18 to 89 years were selected resulting in n=3896 ECGs (female: 1784, male: 2112, median age: 52 years). A total of 46,752 heartbeats have been processed to calculate a mean heart beat for 24x 1h time bins. The figure below shows the time dependent mean ECG beats. Especially, the amplitude varies over day, nicely seen eg for the r- or t-peak. Further, variations in timing visualized via the x-axis are observed with eg earlier or later t-peaks according to daytime.
Conclusions: We observed circadian changes in raw surface ECG signals that might relate to known circadian phases. This information may help to better understand circadian effects on the cardiovascular system as well as to disentangle effects of pathologic cardiovascular diseases and circadian changes on ECG signals as diagnostic tool. Even if this might not be relevant for the human ECG interpretation it might facilitate machine learning approaches by providing relevant additional model input data.