Optimizing EKG signal analysis using the MaxEnt-Burg algorithm

Through its repeating patterns, the electrocardiogram (EKG) is an essential tool for diagnosing cardiac activity. Applying the Maximum-Entropy (ME) Burg algorithm is planned for EKG analysis. A comparison is made with Fourier Transformation (FT). Detecting anomalous activity in EKG time series is challenging for FT. (F)FT problems include noise transformation & sinc peaks caused by time-window effects. These become evident, if weak signals or deviations from normalcy are present. To overcome these issues, the ME-Burg method is proposed. This technique uses autoregression to maximize information. In the Burg algorithm, each signal S(i) at time 𝑖 is related to earlier S(i - k); a true statement for magnetic resonance & EKG data. As shown by magnetic resonance data analysis, [2] this efficiently reduces noise, prevents sinc peaks and improves the detection of anomalous or weak signals. Two approaches are promising: •1> To further reduce noise, one geometrically averages signals over a period To (repeated several times in an interval Tm) & applying the ME-Burg algorithm to this <To-period>. •2> Applying ME-Burg to each To period and geometrically averaging the resulting ME transforms.

[1] S Odugoudar & C Boekema, SJSU-AFC internal report (2024)

[2] C Boekema & MC Browne, AIP Conf Proc #1073 (2008) 260.