NeuroPark: An Automated Approach for Parkinson's Disease Detection

Abstract

Parkinson's disease stands as a pervasive neurodegenerative condition, casting a substantial impact on global health. Swift and precise identification of Parkinson's, in its early stages, holds pivotal importance for efficacious intervention and proficient disease management. Nevertheless, conventional machine learning techniques encounter formidable challenges when deciphering intricate electroencephalogram signals, often necessitating arduous manual interventions. In this research endeavor, we introduce a pioneering methodology for automated Parkinson's detection employing electroencephalogram signals. Our novel approach harnesses Mel spectrogram images, derived from a preprocessed electroencephalogram dataset, seamlessly integrated with convolutional neural networks. This strategic amalgamation enables the extraction of both frequency nuances and temporal patterns from the visual representations, thereby bestowing our model with a remarkable upswing in the accuracy of Parkinson's detection. The methodologies we propose not only hold substantial promise in advancing Parkinson's diagnosis but also bear the potential to foster tailored approaches in the realm of personalized treatment strategies. Key Words: Parkinson’s disease (PD), Classification, Electroencephalogram (EEG), Deep learning, Convolutional Neural Network (CNN), Mel Spectrogram.