Background: The disorders of voice and speech in Parkinson\'s disease (PD) result from the
involvement of several subsystems including respiration, phonation, articulation, and prosody. We have designed a quick vocal test consisting of sustained phonation, diadochokinetic task, and running speech, and assessed its performance in separating PD patients from healthy controls (HC).
Methods: 24 untreated patients with recently diagnosed PD and 22 age-matched HC were tested. In total, 116 vocal recordings were collected and the voice parameters were obtained using 11 measurements designed with the possibility of automatic extraction in a common acoustic environment. Subsequently, a predictive model was built using kernel support vector machine to find the best combination of measurements to differentiate PD from HC subjects.
Results: Significant differences between both groups were found in 10 out of 11 measurements. The best classification performance of 85.02% has reached in a combination of four measures that represent all PD-related speech subsystems, including the ability to maintain sound pressure level, noise-to-harmonics ratio, accuracy of articulation, and melody variations. Reduced melody in running speech appeared essential in characterizing the vocal impairment in PD. In addition, correlations were found between the measures of articulation and phonation, and subscores of bradykinesia and rigidity.
Conclusions: Our designed configuration of acoustic vocal tests can detect abnormalities of
speech since the early untreated stages of PD. Thus, these tests can ease the clinical assessment of voice and speech disorders, and serve as measures of clinical progression as well as in the monitoring of treatment effects.
|
||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||
