Integrating Multimodal Data for Parkinson’s Disease Prediction: An Ensemble Approach

Parkinson's disease is a global health concern, impacting millions of people worldwide, with a report of 6 million individuals affected in 2021, including a significant presence in India. This neurodegenerative disorder primarily manifests with challenging motor symptoms, often leading to a late diagnosis, particularly during the dopamine-deficiency stage, resulting in high misdiagnosis rates. The need for early detection is paramount, as Parkinson's disease significantly affects individuals’ daily lives, work, and social activities. This study presents a novel multimodal ensemble approach to address these challenges. By integrating speech and drawing datasets, we aim to substantially improve prediction accuracy for early diagnosis. This paper focuses on two modules: Module 1, incorporating Gray Wolf Optimization and SVM for speech data analysis, and Module 2, utilizing Squeezenet for drawing data examination. Remarkably, our results demonstrate 100% accuracy in binary classification for both modalities, highlighting the potential for enhanced early Parkinson's disease detection and improved patient care. This contribution is crucial in bridging the existing diagnostic gaps, as previous literature has made strides but leaves room for optimization. In summary, this newly proposed model holds significant promise in advancing Parkinson's disease detection, emphasizing its importance in the global medical landscape.