What is the clinical evidence from this study?

Study design: Systematic Review. Population: Phonocardiogram classification (n=151). Intervention: Machine learning approaches.

June 24, 2026Computer Methods in Biomechanics & Biomedical Engineering

A systematic review of machine learning approaches for phonocardiogram classification

View Full Paper

Ask AI

Bookmark

Key result

Deep learning dominates PCG classification but dataset heterogeneity and generalization issues hinder clinical utility.

n=151

Population

151 studies on machine learning approaches for phonocardiogram classification published predominantly…

Design

Systematic_review

Authors

AAAbdulrahman Al-ShanoonUniversity of Wollongong ESEdward R. SykesUniversity of Guelph HNHedyeh NazariUniversity of Guelph

Discussion

Member takes

Implication

Provides clinicians an updated roadmap for ML-assisted PCG interpretation; extends prior surveys with PRISMA rigor, DL trends, and full-pipeline coverage.

Key Points

This review aims to synthesize recent machine learning approaches for phonocardiogram classification and address methodological gaps.
Conducted systematic search across multiple databases including IEEE Xplore and PubMed/MEDLINE.
Included 151 studies published predominantly between 2021 and 2025 with emphasis on deep learning and hybrid models.
Analyzed classification processes including acquisition, preprocessing, segmentation, and evaluation practices.
Deep learning approaches, particularly CNN-based, dominate current research in PCG classification.
Challenges include dataset heterogeneity, generalization under noise, and pediatric versus adult variations.
Key recommendations for practice include enhancing robustness, interpretability, and promoting standardized evaluation protocols.

Study Design

Type

Systematic Review (n=151)

Structured PICO

Population

151 studies on machine learning approaches for phonocardiogram (PCG) classification published predominantly between 2021 and 2025

Intervention

Machine learning (ML) and deep learning (DL) approaches for PCG classification

Outcome

Methodological mapping (acquisition, preprocessing, segmentation, feature extraction, classification) and performance evaluation of PCG classification models

Deep learning models show high benchmark performance for phonocardiogram classification, but clinical utility requires standardized evaluation, robustness to noise, and interpretable modeling.

Limitations

Comparability is hindered by heterogeneous datasets, non-uniform splits and metric choices
Generalization under noise, device variability and pediatric vs. adult domain shift remains a key challenge
Heterogeneous datasets
Non-uniform splits and metric choices
Generalization under noise
Device variability
Pediatric vs. adult domain shift

Cite This Study

Al-Shanoon et al. (2026) conducted a systematic review in Phonocardiogram classification (n=151). Machine learning approaches was evaluated. Deep learning approaches, especially CNNs, dominate recent phonocardiogram classification across 151 reviewed studies, though heterogeneous datasets and generalization issues hinder clinical utility.

synapsesocial.com/papers/6a3c232cd15afadd906f9d8b https://doi.org/10.1080/10255842.2026.2690183

View Full Paper

Ask AI

Bookmark

Implication

Provides clinicians an updated roadmap for ML-assisted PCG interpretation; extends prior surveys with PRISMA rigor, DL trends, and full-pipeline coverage.

A systematic review of machine learning approaches for phonocardiogram classification

Discussion

Member takes

Implication

Key Points

Study Design

Structured PICO

Limitations

Cite This Study

A systematic review of machine learning approaches for phonocardiogram classification

Discussion

Member takes

Implication

Key Points

Study Design

Structured PICO

Limitations

Cite This Study

Also Consider

Also Consider