Receiver operating characteristic analysis of acoustic and electroglottographic parameters with different sustained vowels

Abstract

Objective To examine the power of the parameters obtained from different sustained vowels used in acoustic and electroglottographic (EGG) voice evaluation protocols to discriminate between dysphonic and non-dysphonic voice quality. Methods Sixty non-dysphonic participants and 30 dysphonic participants were included in the study. In addition to the time domain amplitude and frequency perturbation parameters obtained from the sustained phonation of /ʌ/-/ɛ/-/i/-/u/ vowels, several frequency-domain spectral/cepstral parameters and EGG parameters were evaluated. The classification performance of the acoustic and electroglottographic measures was quantified using analysis and receiver operating characteristic (ROC) curve analysis. Results As a result of ROC analysis, the discriminative diagnostic performance (area under the curve, AUC) of the test for low-vowel (/ʌ/-/ɛ/) phonation was higher than values obtained from high-vowel (/i/-/u/) phonation. For /ʌ/ and /ɛ/ sustained vowels, the parameters exhibiting the highest discrimination were fundamental frequency standard deviation (fo/STD), cepstral peak prominence (CPP), relative average perturbation (RAP), pitch perturbation quotient (PPQ), and jitter percent (JITT). In the EGG parameters, on the other hand, average jitter and periodicity parameters obtained from front vowels (/ɛ/-/i/) were found to have higher AUC values compared to back vowels (/ʌ/-/u/). Conclusions In acoustic analyses, /ʌ/ and /ɛ/ sustained vowels give the highest diagnostic performance. In the electroglottographic evaluation, on the other hand, /ɛ/ and /i/ vowels, when the position of the tongue is forward, have better classification performance compared to /ʌ/ and /u/ vowels, when the position of the tongue is back.