Tabea Thies, Felix Dörr, Louisa Schwed, Johannes Tröger, & Michael T. Barbe
*Poster presented at ADPD 2025
Abstract
Objectives: Apathy is a common non-motor symptom in people living with Parkinson’s disease (PwPD) that can significantly impact the quality of life. Detecting and addressing apathy is essential for improving the health management in PwPD. This study investigates if automatic speech analysis is capable of detecting apathy in PwPD.
Methods: Data from 99 PwPD (32 female, mean age 62 ± 8, mean UPDRS III 22 ± 9) were analyzed. All were native German speakers and completed three speech tasks: maximum vowel /a/ phonation, a reading text, and a picture description. Speech was recorded via a condenser microphone headset, and prosodic features were extracted using SIGMA, ki:elements’ speech processing library. Participants also completed the Apathy Evaluation Scale (AES-S).
Spearman Rank-Sum correlations between speech features and AES scores were calculated, with p-values adjusted via the Benjamini-Hochberg method. An ROC analysis identified optimal cut-offs for apathetic and non-apathetic groups (cut-off > 37).
Results: The average AES score was 30 ± 9, ranging from 18 to 59. Of the 99 PwPD, 21 were classified as apathetic and 78 as non-apathetic. The AES score significantly correlated with phonation task features, including spectral_slope_500_1500 (r = .384, p < .001), spectral_slope_0_500 (r = .314, p = .024), h1_h2_harmonic_difference (r = .376, p < .001), and harmonics-to-noise ratio (r = -.301, p = .024). The harmonics-to-noise ratio also differentiated between the two groups (AUC: 0.698, specificity: 0.538, sensitivity: 0.857).
Conclusions: In this study, 21% of PwPD exhibited apathy, which correlated with specific speech motor function parameters. Apathy was associated with changes in speech features related to phonatory control, potentially causing vocal fold tension fluctuations and vocal quality instability. These acoustic features may serve as indicators for detecting apathy in PwPD.
