Age-Based Heart Sound Segmentation for Accessible Pediatric Cardiovascular Disease Screening

Area of Science:

• Biomedical Engineering
• Cardiology
• Signal Processing

Background:

• Digital stethoscopes offer accessible cardiovascular disease screening in low-resource settings.
• Accurate heart sound segmentation is vital for computer-aided screening systems.
• Existing algorithms are based on adult physiology, limiting pediatric application.

Purpose of the Study:

• To evaluate the performance of adult-based heart sound segmentation algorithms on pediatric phonocardiograms.
• To develop and validate an age-based heart sound segmentation method for improved pediatric accuracy.
• To enhance automated cardiovascular disease screening tools for children.

Main Methods:

• Analysis of the CirCor DigiScope Dataset (3163 recordings, 942 pediatric patients).
• Examination of adult-based assumptions (heart rate, S1/S2 duration, systolic interval) against pediatric data.
• Implementation of a pediatric-based denoising algorithm and modification of segmentation parameters.
• Validation of the proposed age-based segmentation algorithm.

Main Results:

• Significant deviations from adult-based assumptions were observed in pediatric heart rates (5.4%-80%) and S1/S2 durations (53.5%-96.3% overlap).
• The linear relationship between systolic interval parameters was weaker in younger age groups.
• The pediatric-based segmentation improved S1 and S2 F1-Scores across all pediatric age groups from 69.0%-93.7% to 91.4%-99.8%.

Conclusions:

• Adult-based assumptions in heart sound segmentation algorithms are inadequate for pediatric populations.
• An age-based segmentation approach significantly enhances the accuracy of pediatric phonocardiogram analysis.
• Improved pediatric heart sound segmentation is essential for developing reliable automated screening tools for childhood cardiovascular diseases.