The Effectiveness of Acoustic Voice Quality Index to Identify People With Parkinson’s Disease

Developing portable computerized diagnostic tools for Parkinson’s disease (PD) based on Parkinsonian dysarthria requires identifying the voice features most sensitive to PD symptoms. Numerous studies have explored the correlations of various acoustic features with PD diagnosis and progression. Despite achieving high accuracies in some cases, these proposed features or methods have not yet been integrated into publicly available PD diagnostic tools due to their limited robustness. The Acoustic Voice Quality Index (AVQI), introduced by Maryn, has the potential to be a voice severity measure capable of distinguishing people with PD from healthy individuals. This study investigates the effectiveness of AVQI in identifying people with PD and the reliable preprocessing techniques to compute AVQI from short-duration sustained vowels. AVQI data extracted from the PC-GITA dataset underwent several statistical analyses to compare the validity of AVQI using truncated and concatenated sustained vowels. Statistical distribution and receiver operating characteristic (ROC) curve analyses were used to assess AVQI’s effectiveness. The results demonstrate that AVQI can distinguish PD among female participants with an accuracy of 75.33% and an area under the curve (AUC) of 0.74. These findings also confirm the suitability of using concatenated sustained vowels for calculating AVQI from short recordings. They highlight AVQI’s potential application in portable PD diagnostic tools, underscoring its relevance in clinical practice.