Early Detection of Parkinson's Disease through Voice Recording of Phone Calls

Parkinson Disease is a widespread neurological disorder and the second-most common cause of death and disability worldwide. The rapid increase in Parkinson’s disease cases in the last 15 years emphasizes the need for early detection to effectively manage PD while minimizing disruptions in patient lives. However, early diagnosis is challenging due to the overlap of PD symptoms with normal aging and other conditions like essential tremor, coupled with a global shortage of trained neurologists. This study addresses the challenges of early detection of Parkinson’s disease by leveraging deep learning models, specifically convolutional neural networks (CNN), to analyze the characteristics of voice signals. The CNN model processes acoustic features such as pitch, amplitude, jitter, shimmer, and rhythmic patterns, effectively distinguishing between healthy individuals and patients with Parkinson’s disease (PD). By utilizing spectrogram-based visualizations as input, the CNN model captures subtle variations in voice characteristics associated with PD. This approach enhances diagnostic accuracy and efficiency, offering an accessible and scalable solution to mitigate the global shortage of neurologists. The proposed method demonstrates the potential for early-stage detection, paving the way for timely interventions that improve patient care and quality of life.