僧帽弁逸脱症の病理学的重症度評価のための高次スペクトル解析

Mitral valve prolapse (MVP) is a prevalent cardiac disorder affecting approximately 2%-3% of the population. Accurate early diagnosis is essential to prevent progression into more severe conditions. This study introduces a non-invasive methodology for assessing MVP severity using phonocardiogram signals analyzed through bispectral (third-order spectral) techniques. MVP signals were categorized into four types based on murmur intensity and the presence of an ejection click (EC). Following wavelet-based denoising and energy-based segmentation, the energetic ratio (ER%) was computed as a clinical indicator of severity. Bispectral analysis was then applied to extract higher-order spectral (HOS) features including bispectral magnitude, entropies, spectral moments, and the weighted bispectrum center. These features were analyzed to distinguish between severity categories and correlate with murmur energy. An ANOVA test was conducted to assess the statistical significance of each feature and its discriminative power.