Title : Cognitivevoice: Novel machine learning model leveraging acoustic features to predict future cognitive decline in Parkinson’s Disease
Abstract:
Parkinson’s Disease (PD) is the second most common neurodegenerative disorder in the world. While it primarily hinders motor functions, PD also causes cognitive dysfunctions which can progress to severe conditions like dementia. Despite the significant impacts of cognitive dysfunction on PD patients, challenges persist in diagnosing and predicting its onset. Therefore, to prevent cognitive deterioration in PD and allow early therapeutic interventions, there is an urgent need for effective prognostic methods. Vocal impairment is known to be one of the earliest and most prevalent symptoms of PD. However, recent research has shown that vocal changes are not only a hallmark symptom of PD, but also potential indicators of cognitive decline. Thus, this project is the first to assess whether acoustic parameters can predict future neurological dysfunction in PD. In this project, a longitudinal study over three years from the Parkinson’s Progression Markers Initiative Dataset was analyzed. PD patients underwent vocal analysis used to extract mel-frequency cepstral coefficients (MFCCs), which are representations of a sound's acoustic characteristics, and cognitive testing using the Montreal Cognitive Assessment (MoCA). Using machine learning models, the relationship between baseline MFCCs and subsequent changes in MoCA scores was established. By achieving a 75% accuracy using data from actual PD patients, this model showed that MFCCs can be effective predictive biomarkers for cognitive decline in PD. Hence, acoustic features can serve as non-invasive and efficient biomarkers and contribute to the prognosis of future cognitive decline in PD patients, allowing early therapeutic interventions and prevention of cognitive deterioration.
