Automatic segmentation for neonatal phonocardiogram.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Pediatric Cardiology

Background:

• Neonatal phonocardiogram (PCG) analysis is crucial for diagnosing infant heart conditions.
• Accurate segmentation of neonatal PCG is challenging but essential for reliable AI-assisted diagnosis.
• Existing adult PCG segmentation algorithms may not be optimal for the unique characteristics of neonatal heart sounds.

Purpose of the Study:

• To develop and evaluate a novel, automated algorithm for segmenting neonatal phonocardiograms (PCG).
• To assess the algorithm's performance against a baseline adult PCG segmentation method.
• To identify key features for neonatal PCG segmentation and explore its potential to improve heart abnormality detection.

Main Methods:

• Development of a novel PCG segmentation algorithm with a single parameter (maximum heart rate).
• Comparison of the novel algorithm with a baseline algorithm designed for adult PCG.
• Evaluation on a large clinical dataset of neonatal PCG recordings (over 7 hours total duration).

Main Results:

• The novel algorithm achieved a high F1 score of 0.94 on the neonatal PCG dataset.
• The algorithm demonstrated a five-fold increase in cardiac cycle detection compared to manual segmentation.
• Key features relevant for neonatal PCG segmentation were identified and discussed.

Conclusions:

• The proposed algorithm offers accurate and efficient automatic segmentation of neonatal PCG.
• This advancement has the potential to significantly improve the performance of AI-driven diagnostic tools for infant heart abnormalities.
• The increased detection of cardiac cycles can lead to more sensitive detection of heart conditions in newborns.