Differential diagnosis of muscle tension dysphonia and adductor spasmodic dysphonia using spectral moments of the long-term average spectrum.

Abstract

Adductor spasmodic dysphonia (ADSD) can mimic the voice characteristics of muscle tension dysphonia (MTD) contributing to diagnostic confusion and inappropriate management. Elevated spectral noise has been reported in MTD, which may aid in differential diagnosis. The long-term average spectrum (LTAS) can be compared to a Gaussian bell curve using spectral moments analysis. Four moments describe features of the LTAS: spectral mean (moment 1), standard deviation (moment 2), skewness (moment 3), and kurtosis (moment 4). This investigation evaluated spectral moments analysis of the LTAS as an objective test to distinguish ADSD from MTD. Case-control comparison. Pretreatment voice samples from 59 subjects with MTD (10 males and 49 females) and 41 subjects with ADSD (19 males and 22 females) were analyzed. Groups were separated by gender, adjusted for age, and results from the analysis were compared across different analyzing bandwidths. Diagnostic precision estimates were calculated including sensitivity, specificity, positive and negative predictive values, and positive and negative likelihood ratios. No significant spectral differences were found between men in both groups. However, significant between-group differences were identified for all spectral moments for women. Logistic stepwise regression identified that spectral standard deviation (moment 2) uniquely distinguished women with MTD and ADSD. No other spectral moments contributed significant discriminatory information. The results suggest that moment 2 of the LTAS provides respectable diagnostic precision by highlighting spectral noise differences between females with MTD and ADSD. Automated spectral moments analysis deserves further attention as a possible test for differential diagnosis.