Machine Learning-Based Classification of Parkinson's Disease Patients Using Speech Biomarkers

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computational Biology

Background:

• Parkinson's disease (PD) is a prevalent neurodegenerative disorder, particularly affecting aging populations.
• Early diagnosis of PD is crucial but challenging for healthcare systems.
• The prevalence of PD increases significantly with age, highlighting the need for accessible diagnostic tools.

Purpose of the Study:

• To classify Parkinson's disease (PD) patients and healthy controls using voice recordings.
• To evaluate the efficacy of machine learning (ML) algorithms in analyzing speech patterns for PD detection.
• To explore the potential of ML-driven voice analysis for early diagnosis of PD.

Main Methods:

• Collected voice recordings from 252 individuals (aged 33-87) with and without PD.
• Applied supervised machine learning (ML) algorithms and pipelines to analyze voice data with 754 attributes.
• Utilized 10-fold cross-validation to validate the performance of ML models in classifying PD.

Main Results:

• ML models achieved high accuracy (84.21%) and precision (93%) in classifying PD patients.
• Pipeline methods further improved classification performance, reaching 85.09% accuracy and 91% sensitivity.
• The analysis demonstrated ML's capability to distinguish PD from healthy controls based on voice biomarkers.

Conclusions:

• Machine learning classifiers and pipelines can accurately classify Parkinson's disease (PD) using speech biomarkers.
• Pipelines are effective in feature selection and enhancing classification accuracy for PD detection.
• This voice analysis approach offers a viable method for the early diagnosis of initial Parkinson's disease forms.