Polysomnography Dataset for Sleep Analysis in Ischemic Stroke Patients.

Area of Science:

• Neurology
• Sleep Medicine
• Data Science

Background:

• Sleep disturbances are common in ischemic stroke survivors, impacting neural recovery and cognitive function.
• Sleep-disordered breathing (SDB) is a frequent comorbidity in this population.
• A lack of stroke-specific sleep data hinders research into these interactions.

Purpose of the Study:

• To introduce the Polysomnography Dataset for Sleep Analysis in Ischemic Stroke Patients (iSLEEPS).
• To provide a comprehensive, annotated sleep database for stroke research.
• To investigate the prevalence of SDB in an Asian stroke cohort.

Main Methods:

• Collected 100 overnight polysomnography (PSG) recordings from ischemic stroke patients in India (Sept 2018 - Dec 2021).
• Expertly annotated sleep stages, respiratory events, limb movements, and oxygen desaturation per AASM (2017) guidelines.
• Applied deep learning models (LSTM, Transformer, CNN) for automated sleep stage classification.

Main Results:

• The iSLEEPS dataset includes detailed PSG data and clinical metrics for 100 stroke patients.
• A high prevalence of SDB was observed in the study cohort.
• The Long Short-Term Memory (LSTM) model achieved the highest automated sleep stage classification accuracy at 74.70%.

Conclusions:

• The iSLEEPS dataset addresses a critical gap in stroke sleep research.
• This resource facilitates the study of stroke-sleep pathophysiology.
• It supports comprehensive analysis of sleep disturbances post-ischemic stroke.