Objectives: High-frequency breath-sound components are attenuated at the chest wall, producing a low-pass effect. We evaluated this pattern across Healthy controls, Stable chronic obstructive pulmonary disease (COPD), and COPD Exacerbation, and assessed whether a mel-frequency cepstral coefficient (MFCC) axis reflected these states. Methods: We retrospectively analyzed 1,440 posterior recordings from 240 adults (80/group; six 15-s recordings per participant). Spectral bandwidth and 85% spectral roll-off were derived using a short-time Fourier transform pipeline. MFCC1–MFCC13 were summarized by principal component analysis (PCA). ModeT was defined as the PCA component most negatively aligned with both spectral anchors; this rule selected PC6. Results: Bandwidth differed across groups (210.19, 204.96, 206.89; p=0.002), as did roll-off (303.35, 299.47, 297.89; p=0.013) and ModeT (?0.08, ?0.05, 0.14; p=0.001). Ordered trends were significant for all three measures. ModeT explained 3.9% of MFCC variance, correlated with both anchors, and remained higher in Exacerbation than Healthy in clustered record-level models (estimate 0.22, 95% CI 0.09–0.34; p<0.001). Conclusion: COPD chest-wall recordings show a low-pass spectral shift. Spectral anchors and ModeT define an interpretable acoustic phenotype requiring prospective validation. Keywords: Chronic obstructive pulmonary disease (COPD); respiratory sound analysis; electronic stethoscope; low-pass acoustic phenotype; mel-frequency cepstral coefficients (MFCC)
Corresponding Author: İhsan Topaloğlu