Validity of the Cepstral Spectral Index of Dysphonia in the Persian Language in Sustained Vowel, Sentence, and All-Voiced Sentence Contexts.

PURPOSE: Cepstral Spectral Index of Dysphonia (CSID) characterizes and objectively quantifies characteristics of voice quality. The findings of CSID's impact in determining dysphonia characteristics in different languages, however, are unclear but important if this tool is to be used cross-linguistically. Therefore, the purpose of the study was to determine the validity of CSID in Persian speakers.

METHOD: Persian-speaking participants were assigned to two groups: patients with voice disorders (n = 102) and vocally healthy speakers (n = 38). They produced a sustained vowel /ɑ/ and a Persian connected speech sample representing a sentence with a majority of voiced sounds. Three speech-language pathologists conducted auditory-perceptual ratings on a 4-point ordinal scale (vocally healthy, mild, moderate, or severe). Acoustic analysis was conducted using the Analysis of Dysphonia in Speech and Voice software. CSID values in addition to cepstral peak prominence (CPP), CPP standard deviation, low-to-high (L/H) spectral ratio, and L/H spectral ratio standard deviation were extracted from (a) a 1-s segment of the middle part of the sustained vowel, (b) the onset to offset of the connected speech sample, and (c) the same sentence further trimmed to contain only voiced parts by omitting voiceless sections.

RESULTS

The area under the curve (AUC) values for the three tasks were as follows: sustained vowel, AUC = 0.978; connected speech sentence, AUC = 0.961; and all-voiced connected speech sentence, AUC = 0.957. The cutoff points of 4.88 (a sensitivity of 92.2% and a specificity of 94.7%), -4.14 (a sensitivity of 91.2% and a specificity of 92.1%), and 6.76 (a sensitivity of 83.3% and a specificity of 97.4%) yielded the highest level of diagnostic findings for CSID in sustained vowel, connected speech sentence, and all-voiced connected speech sentence, respectively. In addition, significant and strong correlations were found between the CSID values and dysphonia category in the sustained vowel (r = .815, p < .001), connected speech sentence (r = .841, p < .001), and all-voiced connected speech sentence (r = .833, p < .001).

CONCLUSION

The findings of this study showed that CSID is a valid acoustics-based index with the potential to assist clinicians in achieving a more complete characterization and quantification of dysphonia in Persian-speaking clients during voice assessments.