Toward the development of an objective index of dysphonia severity: a four-factor acoustic model

During assessment and management of individuals with voice disorders, clinicians routinely attempt to describe or quantify the severity of a patient's dysphonia. This investigation used acoustic measures derived from sustained vowel samples to predict dysphonia severity (as determined by auditory-perceptual ratings), for a diverse set of voice samples obtained from 134 adult females, with and without voice disorders. Stepwise multiple regression analysis on all voice samples, followed by randomized and repeated cross-validation (random selection of 75% of the original 134 voice sample corpus; 100 iterations) indicated that a four-variable model comprised of time and spectral-based acoustic measures was able to strongly predict perceived severity of dysphonia (mean R = .880; mean R(2) = .775). A cepstral-based measure (CPP/EXP ratio) was determined to be the most significant contributor to the prediction of dysphonia severity, though it is clear that the addition of other acoustic measures (pitch sigma; shimmer (dB); and the Discrete Fourier Transformation ratio, a measure of low versus high frequency spectral energy) add substantially to the accurate prediction of severity. The results are interpreted and discussed with respect to the key acoustic characteristics that contributed to the prediction of severity, the value of identifying a subset of time and spectral-based acoustic measures which appear sensitive to a perceptually diverse set of voices, and the possible use of acoustic models in guiding auditory-perceptual ratings.