Aims: This study aims to develop a novel multimodal Signal Quality Indicator (SQI) for assessing the fidelity of synchronous electrocardiogram (ECG) and phonocardiogram (PCG) signals acquired under ambulatory, non‐standardized conditions. By evaluating quality in both modalities without a fixed reference, the approach ensures mutual consistency, leverages the physiological coupling between electrical and mechanical events, and yields a SQI that is transparent, interpretable, and strongly correlated with expert annotations.
Methods: Fiducial events were extracted from ECG and PCG signals using U-Net–based segmentation, with energy envelopes as input features. The algorithm evaluates temporal alignment between key ECG landmarks (QRS complexes and T waves) and PCG landmarks (S1 and S2), under the assumption that QRS complexes precede S1, and that T waves precede S2 within defined tolerances. Signal quality was scored using a penalized recall metric that rewards correct matches while penalizing mismatches.
Results: The SQI was evaluated on 564 manually annotated ECG–PCG pairs. Among the variants, the minimum linear score (min_lin) showed the strongest correlation with expert labels (Kruskal Wallis test; p < 1e-6). Multinomial logistic regression applied to a 3-level quality description (bad, uncertain, good) achieved an AUC of 0.79, confirming min_lin as the top predictor by permutation-based importance.
\Conclusion: The proposed multimodal SQI provides an interpretable measure of cardiac signal fidelity by leveraging physiological alignment between ECG and PCG fiducial events. Its bidirectional matching framework avoids reliance on baseline modalities and proves effective under opportunistic acquisition conditions.