Paediatric heart sound signal analysis towards classification using multifractal spectra

Area of Science:

• Cardiology
• Biomedical Engineering
• Signal Processing

Background:

• Computer-aided classification of heart sounds is crucial for clinical decision support.
• Existing methods often rely on detecting fundamental heart sounds (S1, S2), which can be challenging.
• A need exists for robust heart sound classification methods that do not require precise localization of S1 and S2.

Purpose of the Study:

• To propose a new statistical method for classifying heart sound recordings.
• To differentiate between healthy heart sounds, click syndrome, and other heart dysfunctions.
• To evaluate the method's performance in the presence of additive noise.

Main Methods:

• Employs singularity spectra analysis and long-term dependency of irregular structures.
• Classifies signals statistically without detecting or localizing fundamental heart sounds (S1, S2).
• Separates healthy signals first, then identifies click syndrome within the unhealthy group.

Main Results:

• Achieved high recall and precision values for all three classified categories: healthy, click syndrome, and other heart dysfunctions.
• Demonstrated high accuracy in classifying heart sound signals, even when subjected to additive noise.
• Successfully classified auscultatory recordings into distinct pathological groups.

Conclusions:

• The proposed statistical method offers an effective approach for heart sound classification.
• The method is robust and performs well even in noisy conditions, supporting clinical decision-making.
• This technique provides a valuable tool for analyzing heart sound recordings without relying on traditional S1/S2 detection.