TOWARD VALIDATION OF AN ACOUSTIC INDEX OF DYSPHONIA SEVERITY by

This study investigated the relationship between a cepstral/spectral index of dysphonia severity (i.e., the CSID) and listener severity ratings of disordered voices. To assess the value of the CSID as a potential objective treatment outcomes tool, preand posttreatment samples of continuous speech and sustained vowel /a/ productions were elicited from 112 patients (with varying degrees of dysphonia) from six diagnostic categories: (1) unilateral vocal fold paralysis (UVFP), (2) adductor spasmodic dysphonia (ADSD), (3) primary muscle tension dysphonia (PMTD), (4) benign vocal fold lesions (BVFL), (5) presbylaryngis, and (6) mutational falsetto. Perceptual ratings of dysphonia severity in continuous speech were compared to acoustically-derived severity estimates using a three factor CSID model consisting of the cepstral peak prominence (CPP), the ratio of low-to-high spectral energy, and its standard deviation. A five factor CSID model incorporating all acoustic variables as well as gender and the CPP standard deviation was used to estimate severity in sustained vowel samples. Results showed strong relationships between perceptual and acoustic estimates in dysphonia severity in connected speech (r = 0.72, p < 0.0001) and sustained vowels (r = 0.836, p < 0.0001). A strong relationship between the perceived and predicted change in dysphonia severity from preto posttreatment was also observed for connected speech (r = 0.77, p < 0.001) and sustained vowels (r = 0.81, p < 0.0001). Spectrum effects were also examined, and overall severity (mild, moderate, or severe) did not influence the relationship between perceived and estimated severity ratings in connected speech (F[1, 2] = 0.58, p = 0.56); however, dysphonia severity did influence the relationship in sustained vowels (F[1, 2] = 6.22, p = 0.002). In general, the results confirm a robust relationship between listener perceived and acoustically-derived estimates of severity within the contexts of connected speech and sustained vowels across diverse diagnostic categories and varying degrees of dysphonia severity. As such, the CSID shows considerable promise an objective treatment outcomes measure.