Johannes Tröger, Ebru Baykara, Felix Dörr, Jan Rusz & Juan Rafael Orozco-Arroyave
* Poster presented at the International Congress of Parkinson’s Disease and Movement Disorders, Philadelphia (USA)
Abstract
Background: This study aims to identify reliable speech biomarkers for assessing dysarthria in individuals with Parkinson’s disease (PD), ensuring clinical validity across languages (Spanish and Czech) and evaluating their performance across the two common speech tasks – sustained phonation /a/ (SPA) and rapid syllable repetition /pa/-/ta/-/ka/ (DDK). Chosen for clinical practicality and reduced patient-burden, these tasks are also considered more language-independent compared to others. Clinically relevant speech features are extracted with the ki:elements speech processing pipeline SIGMA. Age and gender influences are controlled through adjacent pre-processing per language. Composite scores are created and assessed for clinical validity using UPDRS-III, the UPDRS- Speech, as well as H&Y scale and evaluated across both languages.
Methods: The final sample included 96 Spanish and 98 Czech speakers. Out of 93 DDK and 103 SPA features provided by SIGMA, significant features were identified through Spearman Rank Correlations (p < .05) across both languages. Subscores for phonation and articulation —were created by weighting relevant features and building composites. A composite score, derived from the sum of the two subscores, was used to assess clinical significance by comparing diagnostic groups (HC vs PD) with Kruskal-Wallis tests.
Results: Seven and five features were selected for the articulation- and phonation-subscore respectively. Table 1 shows the results of the Kruskal-Wallis tests with regards to the group differences in the full sample, combining Czech and Spanish speakers. The articulation and total scores are significantly different between the groups across both samples while the difference in the phonation score did not reach significance. Density plots of the scores separated by groups can be seen in Figure 1.
Conclusions: Significant differences in articulation and total scores between PD and HC groups were observed in both languages, with no significant findings for phonation. These findings are consistent with prior research that emphasize speech prosody is the most affected dimension in PD, followed by articulation (Orozco-Arroyave et al., 2016), and sustained phonation might not be an appropriate task to measure vowel articulation (Rusz et al., 2013). Considering the tasks used, DDK and articulation emerges as a more appropriate measure for assessing PD-related dysarthria.
References:
Rusz, J., Cmejla, R., Tykalova, T., Ruzickova, H., Klempir, J., Majerova, V., … & Ruzicka, E. (2013). Imprecise vowel articulation as a potential early marker of Parkinson’s disease: Effect of speaking task. The Journal of the Acoustical Society of America, 134(3), 2171-2181.
Orozco-Arroyave, J. R., Hönig, F., Arias-Londoño, J. D., Vargas-Bonilla, J. F., Daqrouq, K., Skodda, S., … & Nöth, E. (2016). Automatic detection of Parkinson’s disease in running speech spoken in three different languages.The Journal of the Acoustical Society of America, 139(1), 481-500.
