The validity of using phonatory jitter to detect laryngeal pathology

Abstract

A multiple regression model was developed to account for the variance in phonatory jitter among normal speakers and for predicting an expected jitter value and confidence interval for individual subjects. Jitter was measured with a system resolution capable of measuring a minimum perturbation of 2.15 μs. Measures were made from maximum phonation lengths of 95 adults without laryngeal pathology. Seven factors were examined for contributions to the prediction of jitter: sex, age, smoking history, drinking habits, F0, vocal intensity, and length of phonation. A multiple correlation coefficient of determination of 97.6% was obtained for the normal subject pool with a three-factor model including: vocal intensity, F0 and phonation length. For 20 patients with laryngeal pathology, individual predicted jitter values and 90% confidence intervals were computed using the normal regression model, for determining when patients' actual jitter values were outside of their expected confidence intervals. Only 30% of the patients with confirmed laryngeal pathology had jitter values outside of their predicted intervals, indicating that this measure would not be valid for use in screening for laryngeal pathology.