Classifying Parkinson’s Disease With or Without Rapid Eye Movement Sleep Behavior Disorder Using Machine Learning-Based Analysis of Single-Lead Electrocardiogram

Abstract

Background and Objective Rapid eye movement sleep behavior disorder is a prodromal stage of alpha-synucleinopathy. Parkinson’s disease with rapid eye movement sleep behavior disorder is associated with more severe symptoms and cerebral pathology compared to Parkinson’s disease without rapid eye movement sleep behavior disorder. This study classified idiopathic rapid eye movement sleep behavior disorder, Parkinson’s disease with rapid eye movement sleep behavior disorder, and Parkinson’s disease without rapid eye movement sleep behavior disorder using single-lead electrocardiogram signals from polysomnography. Methods Subjects who underwent polysomnography and dopamine transporter positron emission tomography between January 2010 and December 2021 were retrospectively analyzed. The study included 4 patients with idiopathic rapid eye movement sleep behavior disorder, 9 with Parkinson’s disease with rapid eye movement sleep behavior disorder, 8 with Parkinson’s disease without rapid eye movement sleep behavior disorder, 9 control subjects, and 15 healthy controls. Heart rate variability features were extracted from electrocardiogram signals, and machine learning models classified the groups. Results No significant differences in demographics or obstructive sleep apnea severity were found between the groups, except for healthy controls. Machine learning classifiers effectively distinguished idiopathic rapid eye movement sleep behavior disorder, Parkinson’s disease with rapid eye movement sleep behavior disorder, and Parkinson’s disease without rapid eye movement sleep behavior disorder based on electrocardiogram features. Conclusions This study demonstrated the potential of single-lead electrocardiogram signals to differentiate idiopathic rapid eye movement sleep behavior disorder, Parkinson’s disease with rapid eye movement sleep behavior disorder, and insight for future prospective studies to predict the conversion.