Abstract

Parkinson's disease is a common neurological condition that predominantly im-pacts people who are older than fifty, leading to speech impairments and movement diffi-culties. Timely diagnosis of PD is essential for efficient care and improved patient outcomes, especially considering the global aging population. Our paper introduces a new method for the early identification of Parkinson's disease utilizing machine learning techniques and the Xception architecture. Our focus lies on analyzing spiral and wave drawings, commonly used in clinical diagnosis. We curated a dataset comprising such drawings from individuals with and without PD and processed the data for model training. Leveraging the Xception architecture, our machine learning models achieved promising results, with a training accu-racy of 95.34% for spiral drawings and 93.34% for wave drawings, along with validation accuracies of 93.00% and 86.00%, respectively. Our highlight the capabilities of machine learning methodologies and the Xception architecture in improving the precision and accu-racy of the diagnosis of Parkinson's disease. Implementing our proposed approach could significantly advance early detection efforts, thereby enhancing patient care and quality of life.