Key result
Time-frequency phonocardiographic analysis with a linear SVM classifier distinguished cardiac pathology from healthy cases with 89.29% accuracy and an AUC of 0.9204.
Why the study?
This study aimed to develop a method for diagnosing selected cardiac pathologies using time-frequency analysis of phonocardiographic signals for application by nursing and medical personnel.
Does time-frequency spectral analysis of S1 phonocardiographic signals using genetic algorithm feature selection and a support vector machine classifier accurately distinguish between healthy individuals and patients with cardiac pathologies?
Cross-Sectional (n=231)
Yes
Does time-frequency spectral analysis of S1 phonocardiographic signals using genetic algorithm feature selection and a support vector machine classifier accurately distinguish between healthy individuals and patients with cardiac pathologies?
Time-frequency analysis of S1 heart tones recorded via an electronic stethoscope combined with genetic algorithm feature selection and SVM classification achieved an AUC of 0.92 in distinguishing healthy individuals from those with cardiac pathologies, demonstrating feasibility for non-invasive outpatient screening.
The article presents a method for diagnosing selected pathological heart conditions based on the time-frequency analysis of phonocardiographic signals, enabling its application by nursing and medical personnel. The study describes the process of collecting diagnostic data, methods for assessing data quality, and the use of a classifier to distinguish between healthy and diseased cases, where the diseased group includes multiple cardiac pathologies. Phonocardiographic signals were acquired by medical staff using a Littmann 3200 electronic stethoscope and the Littmann StethAssist application, and each recording was validated by one of the authors who is a physician. A total of 103 signals acquired in the sitting position at the mitral valve auscultation point were used to determine the performance parameters of the proposed method. Key challenges related to phonocardiographic signal acquisition are discussed, including noise elimination techniques that impact analysis accuracy. The selection of diagnostic features and the optimization of the classifier are presented, with features proposed by the authors and a dedicated feature selection strategy to ensure minimal classification errors while maintaining high signal separation efficiency. The conducted studies confirm the feasibility of developing an effective diagnostic method that allows for the classification of patients as healthy or diseased with sufficient accuracy for medical applications, achieving an AUC of 0.92. The results indicate the potential use of this technology in outpatient settings, opening new perspectives for non-invasive cardiac disease diagnostics.
No takes yet. Share an insight, caveat, or question.
Lejkowski et al. (2026) conducted a cross-sectional in Cardiac pathologies (coronary heart disease, heart failure, previous myocardial infarction, atrial fibrillation, hypertension) (n=231). Time-frequency analysis of phonocardiographic signals with Linear Support Vector Machine (LSVM) classifier vs. Clinical diagnosis validated by a physician (healthy vs. pathological) was evaluated on Diagnostic classification accuracy and area under the ROC curve (AUC) for distinguishing cardiac pathology from healthy subjects (5-fold cross-validation) (AUC 0.9204 (±0.0481); Accuracy 89.29% (±2.40%)). Time-frequency phonocardiographic analysis with a linear SVM classifier distinguished cardiac pathology from healthy cases with 89.29% accuracy and an AUC of 0.9204.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: