Accuracy and Cut-Off Values of Different Acoustic Measures in the SustainedVowel Task of Brazilian Portuguese Speakers
Objective: To define cut-off values for acoustic measures extracted from the
sustained vowel [a] task to discriminate the presence or absence of vocal disorders
in Brazilian Portuguese (BP) speakers. Methods: Cross-sectional diagnostic
accuracy study. The sample consisted of 376 participants (288 female and 88 male),
with a mean age of 41.20 ± 14.04 years. All individuals underwent a laryngological
evaluation and a voice recording session, during which sustained vowel [a]
productions were recorded. Subsequently, both auditory-perceptual judgment and
acoustic analyses of the voice were performed. Among the participants, 277 were
classified as having a vocal disorder, and 99 were classified as vocally healthy. A
total of 44 acoustic measures were extracted and analyzed using receiver operating
characteristic (ROC curves) to determine cut-off values and performance metrics
(sensitivity, specificity, and area under the curve – AUC). Results: Cut-off values
were established for 23 of the 44 extracted acoustic measures. The acoustic
measures with AUC values > 0.70 and balanced sensitivity and specificity above
70% that demonstrated robust and clinically relevant performance in discriminating
between voices with and without vocal disorders in BP speakers were: Amplitude
Variability Index (AVI) (≤ 0.91), Glottal-to-noise excitation (GNE) 2000Hz (> 0.92),
Spectral Noise Level (SNL) 100-2600Hz (≤ 7.71), SNL 100-3000Hz (≤ 3.38), SNL 100-8000Hz (≤ 10.95) and SNL 5100-8000Hz (≤ –1.84). Conclusion: The acoustic measures AVI, GNE 2000Hz, SNL 100-2600Hz, SNL 100-3000Hz, SNL 100-8000Hz, and SNL5100-8000Hz demonstrated satisfactory diagnostic accuracy in the Sustained Vowel Task
for discriminating between voices with and without vocal disorders in BP speakers. These
measures stand out as potential markers for both clinical evaluation and voice screening, supporting their applicability in diagnostic confirmation and preventive
protocols.