AI, Wearables Transform Sleep Diagnostics: ATS 2026 Preview

Traditional sleep disorder diagnosis relies heavily on polysomnography (PSG), a multi-channel recording conducted in a sleep laboratory.¹ This method provides comprehensive data on sleep stages, respiratory effort, cardiac activity, and limb movements.¹ PSG involves the placement of numerous electrodes and sensors on the patient's scalp, face, chest, and limbs to record electroencephalography (EEG), electrooculography (EOG), electromyography (EMG), electrocardiography (ECG), respiratory airflow, respiratory effort, and oxygen saturation. This detailed physiological monitoring allows for the precise identification of sleep architecture, including rapid eye movement (REM) and non-REM sleep stages, and the detection of various sleep-related events such as apneas, hypopneas, and periodic limb movements. However, PSG is resource-intensive, requiring specialised equipment, trained technicians, and an overnight stay, which can be inconvenient and costly for patients.² The logistical barriers associated with PSG often lead to diagnostic delays and underdiagnosis of prevalent conditions such as obstructive sleep apnoea (OSA) and insomnia.³

Next-Generation Technologies in Sleep Diagnostics

The ATS 2026 conference will highlight advancements in next-generation sleep diagnostics, focusing on technologies that aim to overcome the limitations of conventional PSG. These innovations include the integration of artificial intelligence (AI), wearable devices, and advanced sensor technologies.⁴

Wearable Devices and Sensors: Consumer-grade wearables, such as smartwatches and rings, are increasingly incorporating sensors capable of collecting physiological data relevant to sleep. These include photoplethysmography (PPG) for heart rate and oxygen saturation, accelerometers for actigraphy, and sometimes electrodermal activity sensors.⁵ These devices typically measure peripheral blood flow changes to infer heart rate and, in some cases, estimate oxygen saturation. Accelerometers detect movement patterns, which can be used to estimate sleep-wake cycles and sleep efficiency. Medical-grade wearable sensors are also under development, offering enhanced accuracy and validation for clinical use. These devices enable continuous, unobtrusive monitoring in the patient's natural sleep environment, potentially capturing sleep patterns over multiple nights, which can provide a more representative picture than a single night in a lab.⁶ This extended monitoring period is particularly beneficial for conditions with night-to-night variability, such as insomnia or restless legs syndrome, where a single night's PSG may not capture typical sleep disturbances. Furthermore, the comfort and familiarity of the home environment can reduce the "first-night effect" often observed in laboratory PSG, where sleep patterns are altered due to the unfamiliar setting.

Artificial Intelligence and Machine Learning: AI algorithms are being developed to process the vast amounts of data generated by wearables and other sensors. These algorithms can identify sleep stages, detect respiratory events, and characterise sleep architecture with increasing accuracy.⁷ For example, AI models trained on large PSG datasets can analyse single-channel electroencephalography (EEG) or even non-EEG signals (like heart rate variability and actigraphy) to infer sleep stages, potentially simplifying the diagnostic process.⁸ The application of machine learning techniques, such as deep neural networks, allows these algorithms to learn complex patterns within physiological signals that may not be evident through traditional rule-based analysis. AI also holds promise for predicting treatment response and personalising interventions based on individual sleep phenotypes.⁹ This includes identifying specific subgroups of patients with OSA who may respond better to continuous positive airway pressure (CPAP) therapy versus oral appliances, or tailoring cognitive behavioral therapy for insomnia (CBT-I) based on individual sleep patterns and psychological profiles.

Transformative Technologies Across the Care Continuum: The integration of these technologies is expected to transform the entire sleep care continuum. In primary care, simplified screening tools incorporating wearable data and AI analysis could facilitate earlier identification of individuals at risk for sleep disorders.¹⁰ This is crucial given the high prevalence of conditions like OSA, affecting a significant portion of the adult population, and insomnia, which impacts millions globally. For specialists, these tools could provide longitudinal data, allowing for more nuanced diagnosis and monitoring of treatment efficacy.¹¹ Remote patient monitoring platforms, leveraging these technologies, could also support virtual consultations and reduce the need for in-person visits, improving access to care, particularly in underserved areas.¹² This shift could alleviate the burden on specialized sleep clinics and improve patient adherence to long-term management plans.

While the potential benefits are substantial, several challenges remain. Standardisation of data collection and interpretation across different devices is critical.¹³ The variability in sensor types, algorithms, and data output among different manufacturers necessitates robust validation to ensure clinical utility and comparability. Regulatory pathways for medical-grade wearables and AI algorithms are still evolving.¹⁴ The Food and Drug Administration (FDA) and other regulatory bodies are developing frameworks to ensure the safety, effectiveness, and analytical validity of these novel diagnostic tools. Furthermore, ensuring data privacy and security, as well as addressing potential biases in AI algorithms, are essential considerations for widespread clinical adoption.¹⁵ Bias can arise from training data that do not adequately represent diverse patient populations, potentially leading to disparities in diagnosis and treatment. Validation studies comparing the diagnostic accuracy and clinical utility of these next-generation tools against established PSG are ongoing and will be crucial for their integration into clinical guidelines.¹⁶ These studies must demonstrate not only high sensitivity and specificity but also clinical impact on patient outcomes and cost-effectiveness.