Non-linear analytical methods surpass jitter and shimmer in quantifying dysphonia in vocal cord paralysis

9 Clinical acoustic voice recording analysis is usually performed using classical perturbation measures 10 including jitter, shimmer and noise-to-harmonic ratios. However, restrictive mathematical 11 limitations of these measures prevent analysis for severely dysphonic voices. Previous studies of 12 alternative nonlinear random measures addressed wide varieties of vocal pathologies. Here, we 13 analyze a single vocal pathology cohort, testing the performance of these alternative measures 14 alongside classical measures. 15 We present voice analysis preand post-operatively in 17 unilateral vocal fold paralysis (UVFP) 16 patients and 11 healthy controls, patients undergoing standard medialisation thyroplasty surgery, 17 using jitter, shimmer and noise-to-harmonic ratio (NHR), and nonlinear recurrence period density 18 entropy (RPDE), detrended fluctuation analysis (DFA) and correlation dimension. Systematizing 19 the preparative editing of the recordings, we found that the novel measures were more stable and 20 hence reliable, than the classical measures, on healthy controls. 21 RPDE and jitter are sensitive to improvements preto post-operation. Shimmer, NHR and DFA 22 showed no significant change (p > 0.05). All measures detect statistically significant and clinically 23 important differences between controls and patients, both treated and untreated (p < 24 0.001, AUC > 0.7). Preto post-operation, GRBAS ratings show statistically significant and clinically 25 important improvement in overall dysphonia grade (G) (AUC = 0.946, p < 0.001). 26 Re-calculating AUCs from other study data, we compare these results in terms of clinical 27 importance. We conclude that, when preparative editing is systematized, nonlinear random 28 measures may be useful UVFP treatment effectiveness monitoring tools, and there may be 29 applications for other forms of dysphonia. 30