A team of researchers from NIMHANS and the International Institute of Information Technology (IIIT), Hyderabad, has developed the first open-access Asian database of overnight sleep studies in patients with ischemic stroke.
The database, which has been made publicly available, is expected to support research on sleep disorders in stroke survivors and help improve post-stroke care.
The dataset, named iSLEEPS (Polysomnography Dataset for Sleep Analysis in Ischemic Stroke Patients), has been published in the Nature Portfolio journal Scientific Data. It comprises 100 overnight polysomnography (PSG) recordings collected at NIMHANS between September 2018 and December 2021.
Each recording includes manually scored sleep stages, respiratory events, oxygen desaturation episodes, periodic limb movements and clinical information, all annotated, according to the American Academy of Sleep Medicine’s 2017 guidelines.
Srijithesh P.R., professor of neurology at NIMHANS and one of the corresponding authors of the study, told The Hindu that the project was conceived to address a major gap in sleep medicine research.
“Sleep-disordered breathing is extremely common after ischemic stroke, but researchers have had very limited access to large, well-annotated datasets from stroke patients. Most publicly available sleep databases are from healthy individuals or non-stroke populations. We wanted to create a high-quality resource that researchers across the world can use to answer clinically relevant questions,” he said.
Dr. Srijithesh said sleep disturbances are often overlooked despite their impact on neurological recovery. “Sleep is not merely a period of rest. It plays an important role in brain repair, memory consolidation and functional recovery after stroke. If sleep disorders such as obstructive sleep apnea are not recognised and treated, they can adversely affect rehabilitation and even increase the risk of recurrent stroke,” he said.
The study pointed out that between 70% and 80% of stroke patients experience some form of sleep-disordered breathing, while Indian studies have reported similar prevalence among patients with ischemic stroke. Yet, comprehensive sleep datasets from this population have remained scarce, particularly from Asia.
The study pointed out that existing international databases either include only a small number of patients, rely on single physiological signals such as electrocardiograms, or do not provide detailed annotations of sleep stages and respiratory events. The iSLEEPS database attempts to bridge this gap by providing multimodal sleep recordings from a stroke-specific cohort.
The cohort comprised 100 patients aged above 50, including 23 women. Overnight sleep studies recorded brain activity, eye movements, muscle activity, heart rhythm, airflow, oxygen saturation, respiratory effort and body position.
Analysis of the dataset showed that 38% of participants had severe sleep apnea, 23% had moderate sleep apnea and 24% had mild sleep apnea, while only 15% had normal breathing during sleep.
To demonstrate the value of the database, the researchers also evaluated artificial intelligence models for automated sleep-stage classification. Among the three models tested, the Long Short-Term Memory (LSTM) model achieved the highest accuracy of 74.7%, followed by the Transformer model (67.4%) and the Convolutional Neural Network (61.7%).
S. Bapi Raju of IIIT Hyderabad, another corresponding author, said the database was intended to advance stroke and sleep research beyond AI applications.
“It can be used to study sleep physiology after stroke, understand disease mechanisms, validate new diagnostic methods and develop better screening tools. Because it is openly available, researchers can build on it instead of starting from scratch,” he said.
The anonymised dataset has been made available through open repositories for research and education. It has already been downloaded by researchers from 20 organisations in nine countries, including China, Belgium, Hong Kong and the United States.
Published - July 02, 2026 07:30 pm IST
Source: The Hindu - India News
