Time-Frequency Representations for second heart sound analysis.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Cardiology

Background:

• Accurate analysis of the second heart sound (S2) is crucial for understanding cardiac function and diagnosing conditions.
• Traditional methods for S2 analysis have limitations in capturing its complex genesis.

Purpose of the Study:

• To evaluate and compare various Time-Frequency Representations (TFRs) for analyzing simulated and real second heart sounds (S2).
• To identify the most effective TFR for S2 analysis and diagnostic applications.

Main Methods:

• Simulated S2 signals were analyzed using classical and modern TFRs: Spectrogram, Wigner-Ville Distribution, Time Varying Autoregressive (TVAR) model, Scalogram, and Hilbert-Huang Spectrum (HHS).
• Performance was assessed using local 2D correlations (rho) and normalized root-mean-square error (NRMSE) based on time moments.
• The best-performing TFR was then applied to real S2 signals from aortic and pulmonary locations.

Main Results:

• Under no noise conditions, HHS and TVAR (Burg algorithm) showed high performance with average rho values of 0.788 and 0.812, respectively.
• HHS achieved the lowest normalized root-mean-square error (NRMSE) of 0.172, indicating superior accuracy.
• The study confirmed HHS as the TFR with the best performance for S2 analysis.

Conclusions:

• The Hilbert-Huang Spectrum (HHS) is a highly effective method for analyzing the second heart sound (S2).
• HHS shows significant promise for improving the diagnosis of cardiac conditions through detailed S2 signal analysis.