An electroglottographical analysis-based discriminant function model differentiating multiple sclerosis patients from healthy controls

Abstract

Dysarthrophonia is a predominant symptom in many neurological diseases, affecting the quality of life of the patients. In this study, we produced a discriminant function equation that can differentiate MS patients from healthy controls, using electroglottographic variables not analyzed in a previous study. We applied stepwise linear discriminant function analysis in order to produce a function and score derived from electroglottographic variables extracted from a previous study. The derived discriminant function’s statistical significance was determined via Wilk’s λ test (and the associated p value). Finally, a 2 × 2 confusion matrix was used to determine the function’s predictive accuracy, whereas the cross-validated predictive accuracy is estimated via the “leave-one-out” classification process. Discriminant function analysis (DFA) was used to create a linear function of continuous predictors. DFA produced the following model (Wilk’s λ = 0.043, χ2 = 388.588, p < 0.0001, Tables 3 and 4): D (MS vs controls) = 0.728*DQx1 mean monologue + 0.325*CQx monologue + 0.298*DFx1 90% range monologue + 0.443*DQx1 90% range reading − 1.490*DQx1 90% range monologue. The derived discriminant score (S1) was used subsequently in order to form the coordinates of a ROC curve. Thus, a cutoff score of − 0.788 for S1 corresponded to a perfect classification (100% sensitivity and 100% specificity, p = 1.67e⁻²²). Consistent with previous findings, electroglottographic evaluation represents an easy to implement and potentially important assessment in MS patients, achieving adequate classification accuracy. Further evaluation is needed to determine its use as a biomarker.