Enhancing PCG Signal Quality Through Cascaded Adaptive Noise Cancelling with Metaheuristic Optimization

Phonocardiogram (PCG) signals, vital for accurate cardiac monitoring and diagnostics, are often compromised by noise from various sources, including lung sounds, environmental sounds, and stethoscope movement. This contamination severely impacts the precision of cardiac assessments. This paper introduces a robust optimization approach that combines a cascaded adaptive noise canceller (ANC) with the greater cane rat algorithm (GCRA) to significantly enhance PCG signal quality. The proposed method first subjects real PCG signals to diverse noise types, such as uniform noise, Gaussian noise, and pink noise. The corrupted signals are then processed through the cascaded ANC, which dynamically adjusts its coefficients to effectively minimize noise while preserving the integrity of the clean PCG signal. The GCRA fine-tunes the filter parameters, optimizing noise suppression and ensuring the preservation of essential cardiac acoustic details. The performance of the GCRA-optimized infinite impulse response (IIR) ANC is thoroughly evaluated using metrics like signal-to-noise ratio (SNR), mean square error (MSE), maximum error (ME), normalized root mean square error (NRMSE), and correlation coefficient (CC). Moreover, the approach is benchmarked against two well-established optimization algorithms such as the gazelle optimization algorithm (GOA) and the dwarf mongoose optimization algorithm (DMOA). The results clearly demonstrate that the GCRA-optimized IIR ANC not only surpasses GOA and DMOA-based ANCs but also outperforms all previously reported PCG signal enhancement techniques, delivering superior noise reduction and preserving critical cardiac information. Moreover, the effectiveness of the proposed GCRA-based noise removal process is confirmed by using a deep learning model to classify normal (NOR) and abnormal (ABNOR) PCG, demonstrating its practical use.