Hyderabad: Artificial Intelligence (AI) is transforming the way sleep disorders are diagnosed, with researchers at the International Institute of Information Technology, Hyderabad (IIITH) spearheading innovations that promise faster, more accurate, and non-intrusive solutions. Prof S Bapi Raju highlighted the institute’s advancements in automatic classification of sleep stages, a critical step in diagnosing conditions such as insomnia, sleep apnea, and narcolepsy.

Sleep plays a vital role in physical, cognitive, and emotional health. Poor sleep has been linked to cardiovascular diseases, diabetes, obesity, cognitive decline, and mental health issues like anxiety and depression. Traditional diagnosis relies on polysomnography (PSG), which involves labor-intensive manual scoring of sleep stages from EEG signals. This process is time-consuming and prone to errors. AI, particularly deep learning (DL), is now offering a paradigm shift.

AI in sleep stage classification

Sleep is divided into NREM (N1, N2, N3) and REM stages, each with distinct brain wave patterns. DL models such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) have demonstrated remarkable accuracy in classifying these stages. Supervised learning uses annotated datasets to train models. On the other hand, unsupervised learning, including clustering and self-supervised techniques, identifies patterns without labels. IIITH’s mulEEG, a multi-view self-supervised learning method, has outperformed existing baselines, delivering an 8x increase in training efficiency.

Wearable innovations

Recognising the limitations of current consumer sleep-monitoring devices, IIITH has developed an EOG-based wearable mask. Electrooculography (EOG) tracks eye movements and, when combined with electrode-based sensors, provides multi-channel data for real-time sleep stage classification.

Unlike traditional PSG setups, this solution is non-intrusive, portable, and suitable for home-based monitoring. It addresses challenges of inconsistent data quality across diverse populations, offering a reliable alternative for clinical and community healthcare.

iSLEEPS dataset collaboration

In collaboration with NIMHANS Neurology Department, IIITH has launched the Indian SLEEP Stroke dataset (iSLEEPS), hosted at IHub-Data, IIIT-H. This dataset includes PSG recordings and clinical annotations of 100 ischemic stroke patients, most of who suffer from sleep disorders. By adhering to strict ethical guidelines and anonymisation protocols, iSLEEPS provides a robust foundation for AI models tailored to the Indian population.

It enables researchers worldwide to identify unique risk factors and patterns, enhancing global understanding of sleep health.

Future scope

IIITH’s research demonstrates that EOG signals processed through AI models can effectively classify sleep stages, paving the way for non-intrusive, home-based sleep monitoring devices. Future developments may integrate additional physiological signals such as heart rate and respiratory rate, further improving diagnostic accuracy. Such innovations could revolutionize wearable sleep technology, offering personalized healthcare solutions and advancing community-level interventions.

Prof. Raju emphasised that these breakthroughs not only strengthen diagnostic capabilities but also open avenues for global collaboration in sleep medicine.

By combining AI with innovative wearable technology and curated datasets, IIITH is positioning India at the forefront of sleep research and healthcare innovation.