Sleep medicine is entering a new era — one in which artificial intelligence (AI) and physiological insight are redefining how sleep is measured, interpreted, and delivered.
Transforming Home Sleep Testing Through AI
Home sleep testing expands access to sleep studies but has important limitations. Without EEG, conventional home sleep testing cannot detect sleep stages or arousals, both of which are critical for diagnostic confidence, treatment planning, and accurate AHI scoring. This gap particularly impacts women, younger, and non-obese individuals, who often experience arousal-based hypopneas that home sleep tests systematically under detect. These patients face increased risk of false negatives, underestimated OSA severity, and delayed treatment.
Nox BodySleep™ 2.0 directly addresses this gap. Using deep learning to interpret respiratory inductance plethysmography (RIP) signals, it automatically scores sleep states (NREM, REM, Wake) and arousals — without EEG. By analyzing subtle changes in breathing patterns and identifying hypopneas without desaturations, it enables calculation of an AHI value that closely aligns with PSG results, and helps clinicians achieve conclusive HSAT results on the first try, leading to reduced delays in care, missed diagnoses, and unnecessary retesting.
“BodySleep 2.0 brings physiological depth to home sleep testing,” says Jón Skírnir Ágústsson, PhD, VP of AI at Nox Medical. “Grounded in well-understood respiratory physiology, it delivers AHI values that strongly correlate to PSG results — extending clinical insight beyond what standard HSATs can offer and potentially reducing the need for retesting.”
Scientific Foundation: From Breathing Patterns to Sleep States
The foundation of BodySleep 2.0 lies in the well-established link between respiratory patterns and sleep architecture. As described in Eysteinn Finnsson et al., 2025, Detecting Arousals and Sleep from Respiratory Inductance Plethysmography, breathing becomes steady and metabolically driven in NREM sleep, irregular in REM due to muscle atonia, and variable during wakefulness. Arousals trigger distinct bursts of respiratory effort — patterns that can be recognized and classified by deep learning models trained on large, diverse datasets.
“Breathing reflects sleep state in reliable, measurable ways,” explains Ágústsson. “Teaching AI to interpret those patterns allows us to narrow the gap between polysomnography and home testing, demonstrating accuracy on par with human EEG-based scoring.”
The model behind BodySleep 2.0 was validated across nearly 3,500 sleep studies, including over 1,200 gold-standard PSGs with human scoring. Its performance showed intraclass correlation coefficients (ICC) of 0.91 for total sleep time and 0.74 for arousal index, confirming strong agreement with manual scoring.
Clinical Application: Improving Confidence and Access
From a clinical perspective, Nox BodySleep 2.0 enhances HSAT by allowing physicians to exclude wake periods, include arousal-related hypopneas, and achieve PSG-validated correlation.The result is a more representative measure of disease severity and fewer inconclusive tests.
According to Heidi Riney, MD, Chief Medical Officer at Nox Medical, “BodySleep 2.0 has the capacity to help clinicians make confident diagnostic decisions after a single HSAT. It is particularly valuable for patients who are often overlooked, including women, younger adults, and individuals who are not obese, whose respiratory events may not cause oxygen desaturation but do trigger cortical arousals. These arousal-based hypopneas disrupt sleep and activate the autonomic nervous system, leading to fluctuations in heart rate and blood pressure that carry real cardiovascular implications. By detecting these clinically important events that traditional HSATs frequently miss, BodySleep 2.0 has to potential to address gender related disparities in diagnosis and supports more equitable and accurate identification of sleep disordered breathing.”
By improving diagnostic accuracy and reducing repeat studies, Nox BodySleep 2.0 not only supports greater clinical confidence but also potentially supports reduced patient burden, shorter care timelines, and increased therapy initiation rates.
Join the Discussion
To dive into some of the critical ways sleep medicine is changing as it converges with advanced technology, Nox Medical will host a Sleep Research Society webinar on December 3rd, exploring the science and clinical impact of Nox BodySleep™ 2.0, available within the DeepRESP medical software, which has received FDA clearance.
The webinar session will explore how AI-driven respiratory analysis is transforming home-based sleep diagnostics — bringing the precision of PSG to the accessibility of HSAT.
Speakers:
- Jón Skírnir Ágústsson, PhD, VP of AI at Nox Medical — The scientific foundation and validation of Nox BodySleep 2.0
- Heidi Riney, MD, Chief Medical Officer — Clinical integration and real-world outcomes in sleep care management
Learning Objectives:
- Understand the physiological principles linking respiratory patterns to sleep and arousals.
- Review validation data indicating BodySleep 2.0’s consistency with PSG- based scoring.
- Explore how AI integration enhances HSAT accuracy, reduces inconclusive results, and expands diagnostic reach across diverse populations.
Click here to sign up for the webinar
Topic: Events
