Sex Differences in the ECG Interpretation: a Functional Data Analysis Perspective

Prompt diagnosis and correct therapy are essential for the treatment of cardiovascular diseases. A proper evaluation of electrocardiography (ECG) tracings could help to analyze the different effects of drug therapies in men and women and therefore would allow the prescription of more specific drugs and dosages. For this purpose, it is essential to know the morphology of the ECG and to know how to read it correctly. In recent years there has been an increasing awareness of the differences between a male and female ECG trace and, therefore, of the need to take them into account to obtain a correct diagnosis. This work proposes a method for the recognition of the subject's sex starting from the morphology of the ECG trace alone to highlight sex differences. This technique employs Functional Data Analysis (FDA), a statistical approach specifically developed for the analysis of curves and surfaces. We proved the adequacy of our method by evaluating its ability to classify the traces correctly and is proposed as a smart analysis tool to be employed in the development of ad hoc cardiac drugs. The procedure foresees a preprocessing of the signal through a Butterworth filtering, wavelet-based smoothing, and alignment of the traces. We then classify the signals through a cluster analysis in the form of multivariate functional k-mean procedures. The result is a semi-automatic assignment of the sex of the subject to which the ECG belongs. The method reaches better performances when considering younger subjects because morphological differences between sexes are more evident in this subpopulation, as previously highlighted in literature. More specifically, the accuracy is 77.8% in the younger population and 71.4% in elderly subjects. The technique hereby proposed is a valuable tool for the exploration of ECG tracings to be employed in clinical and pharmacological research.