Introduction:
Parkinson’s disease is a progressive neurodegenerative disorder often diagnosed at advanced stages, limiting treatment effectiveness. Early detection is critical for improving patient outcomes, as it enables timely therapeutic strategies and slows disease progression. This research addresses a significant gap in healthcare diagnostics by employing machine learning to develop accurate, early-stage detection tools, integrating speech and motor features as predictive indicators.

Objective:
The study aims to develop and validate ML models capable of detecting Parkinson’s disease in its early stages. The central research question is: Can machine learning effectively combine multimodal data, including speech and motor assessments, to improve early detection of Parkinson’s disease?

Methodology:
The research utilized a dataset containing speech metrics (e.g., pause intervals, respiratory patterns) and motor assessments (e.g., tremor severity). Exploratory Data Analysis identified significant predictors, such as pause intervals and loudness, with dimensionality reduction techniques refining features to enhance model performance. Various ML classification algorithms were evaluated using metrics like accuracy, sensitivity and ROC-AUC.

Results:
Preliminary results indicate ML models achieved sensitivity and specificity rates exceeding 90%, highlighting their reliability for early detection. Speech features such as pause intervals and respiratory patterns were the strongest predictors, while motor assessments showed moderate correlations. Combining multimodal data significantly improved accuracy over single-modality approaches. Dimensionality reduction reduced noise, enhancing robustness and focus on key features. Additionally, dimensionality reduction through PCA reduced noise and improved robustness, highlighting the importance of key features. These findings suggest that ML-driven multimodal diagnostics can capture subtle early symptoms, enabling earlier and more precise detection.

Conclusion:
The study demonstrates the transformative potential of ML-driven multimodal diagnostics for early detection of Parkinson’s disease. These findings provide actionable insights for clinical applications and establish a foundation for integrating personalized monitoring systems. This research contributes to predictive healthcare, offering opportunities for advancements in neurodegenerative disease diagnostics and treatment strategies.