Assessment of the state of health by the measurement of a set of biophysiological signals

Abstract

The dissertation studies the estimation of the degree of self-similarity and entropy of Shannon of several real electrocardiography (ECG) signals for healthy and non-healthy humans. The goal of the dissertation is to create a starting point algorithm which allows distinguishing between healthy and non-healthy subjects and can be used as a basis for further study of a diagnosis algorithm, necessarily more complex. We used a novel Hurst parameter estimation algorithm based on the Embedded Branching Process, termed modified Embedded Branching Process algorithm. The algorithm for estimation of entropy was based on Shannon‟s entropy. Both algorithms were applied on the spatial distribution of ECG signals in a windowed manner. The studied signals were retrieved from the Physionet website, where they are diagnosed as normal or as having certain pathologies. The results presented for the Hurst parameter estimation allow us to confirm the results already published on the temporal self-similarity of ECG signals, this time for its spatial distribution. We also conclude that the non-self similar signals belong to non-healthy subjects. The results obtained for entropy estimation on the spatial distribution of ECG signals also allowed a comparison between healthy and non-healthy systems. We obtained high entropy estimates both for healthy and non-healthy subjects; nevertheless, non-healthy subjects show higher variability of Shannon‟s entropy than healthy ones.