Artificial Intelligence Diagnosis of Parkinson's Disease From MRI Scans.

Abstract

Parkinson's disease (PD) is a prevalent neurodegenerative disorder characterized by motor symptoms such as tremors, rigidity, and bradykinesia, affecting approximately 6.1 million people worldwide, according to estimates from the Parkinson's Foundation. Early and accurate diagnosis of PD is crucial for effective management and treatment. In this study, we aimed to develop an artificial intelligence (AI) model capable of distinguishing between magnetic resonance imaging (MRI) scans of individuals with PD and those without PD. A total of 442 MRI scans were utilized for training the AI model, comprising 221 scans of individuals diagnosed with PD and 221 scans of healthy controls. The dataset, obtained from a publicly available image dataset on Kaggle.com, was randomly split into three sets: training, validation, and testing, with 80%, 10%, and 10% of the data allocated to each set, respectively. Leveraging Google's Collaboration platform for model training, the AI model achieved exceptional performance, with accuracy, precision, recall (sensitivity), specificity, and F1-score all measuring at high levels. Additionally, the area under the receiver operating characteristic curve (AUC) for the model was found to be 1, indicating strong discrimination between PD and non-PD cases. This study presents a novel AI model capable of accurately identifying PD from MRI scans with high precision and reliability, offering promise for enhancing early diagnosis and personalized treatment strategies for individuals affected by PD.