In Silico Decoding of Parkinson's: Speech & Writing Analysis.

Area of Science:

• Neurology
• Biomedical Engineering
• Artificial Intelligence
• Linguistics

Background:

• Parkinson's disease (PD) is the fastest-growing neurological disorder globally, with cases projected to double by 2040.
• Rising life expectancy and population growth contribute to the increasing prevalence of PD.
• The economic burden of PD is substantial, with significant socio-economic implications for patients and caregivers.

Purpose of the Study:

• To investigate the potential of using speech and handwriting analysis for early detection of Parkinson's disease (PD).
• To identify prodromal PD biomarkers through advanced AI analysis of speech and handwriting features.

Main Methods:

• A transdisciplinary approach combining medical, biomedical engineering, AI, and linguistic expertise.
• Analysis of running speech and continuous handwriting from 20 PD patients (Hoehn and Yahr stages 1-4) and 10 healthy controls.
• Utilized advanced AI techniques, including the ParkinsonNet model, to analyze speech and handwriting features.

Main Results:

• Distinct speech and handwriting patterns were identified in PD patients compared to controls.
• The ParkinsonNet model achieved high predictive accuracy: 95.74% for speech and 96.72% for handwriting.
• Speech and handwriting show potential as effective early biomarkers for PD detection.

Conclusions:

• Integrating AI for analyzing speech and handwriting offers a promising approach for early PD detection.
• This methodology provides a novel diagnostic tool and enhances understanding of early PD manifestations.
• Further validation in diverse cohorts and clinical integration are necessary for timely PD pre-diagnosis and management.