Automatic segmentation of the second cardiac sound by using wavelets and hidden Markov models

Area of Science:

• Biomedical Engineering
• Cardiology
• Signal Processing

Background:

• The second heart sound (S2) comprises aortic (A2) and pulmonary (P2) components, crucial for cardiac diagnosis.
• The timing and order of A2 and P2, known as S2 splitting, provide vital diagnostic information.
• Accurate segmentation of S2 components is essential for reliable cardiac assessment.

Purpose of the Study:

• To develop an automated technique for segmenting the second heart sound (S2).
• To estimate the order of occurrence of aortic (A2) and pulmonary (P2) components.
• To quantify the delay (split) between A2 and P2 for diagnostic purposes.

Main Methods:

• Utilized discrete wavelet transform (DWT) for signal processing.
• Employed hidden Markov models (HMMs), including discrete density HMMs (DDHMMs) for segmentation and embedded continuous density HMMs for acoustic modeling.
• Collected and analyzed phonocardiogram (PCG) and electrocardiogram (ECG) data from five subjects.

Main Results:

• Successfully segmented S2 into its A2 and P2 components.
• Accurately estimated the order of A2 and P2 occurrence.
• Quantified the S2 splitting interval, a key diagnostic parameter.

Conclusions:

• The proposed DWT and HMM-based method offers an effective approach for automated S2 segmentation and analysis.
• This technique aids in diagnosing conditions related to S2 splitting, reverse splitting, or reverse component occurrence.
• The findings support the use of advanced signal processing for enhanced cardiac auscultation analysis.