EEG Signature of Idiopathic Hypersomnia: Insights from Sleep Microarchitecture and Hypnodensity Metrics
Background and Objectives: Patients with idiopathic hypersomnia with long sleep time (IH) report daytime hypersomnolence despite prolonged sleep time and normal sleep macrostructure. As they often have non-restorative sleep, we investigated whether the structure of their sleep is abnormal. Methods: In polysomnography recordings from 80 IH participants and 48 controls, we quantified hypnodensity metrics across the night (macro level), periodic and aperiodic spectral properties, infraslow fluctuations of sigma power within the night (meso level), slow waves, sleep spindles and their clustering (microstructure). Multivariate machine-learning models were used to classify IH vs. control sleep. Results: Hypnodensity metrics were comparable between IH and controls, apart from more mixed wake/N1 sleep epochs in IH, and greater divergence between consecutive epochs of the same stage during NREM sleep in IH. Sigma power was increased in N2 sleep in IH and sleep spindles were more frequent and clustered. Slow wave density was higher in IH. Higher mean spindle cluster size correlated with higher Epworth Sleepiness Scale scores. Multivariate machine learning models incorporating these features achieved a balanced accuracy of 74% in distinguishing IH from controls. Discussion: While spindles and slow waves are typically associated with good sleep quality, they are increased in IH patients. This could reflect greater need for sleep and increased difficulty waking up in IH, which is also characterized by more mixed wake/N1 stages.