Differentiating between light and deep sleep stages using an ambulatory device based on peripheral arterial tonometry

Area of Science:

• Biomedical Engineering
• Sleep Medicine
• Signal Processing

Background:

• The Peripheral Arterial Tone (PAT) signal reflects sympathetic activity and is measured by ambulatory devices like Watch-PAT100.
• Existing algorithms can detect wake, NREM, and REM sleep, but differentiating light and deep sleep remains a challenge for unattended monitoring.

Purpose of the Study:

• To develop and validate an automatic algorithm for differentiating light and deep sleep stages using PAT signals.
• To enhance the capability of ambulatory sleep monitoring devices for comprehensive sleep stage assessment.

Main Methods:

• An algorithm was developed using 14 features extracted from PAT amplitude and inter-pulse period (IPP) time series.
• The algorithm was trained on 49 patients and validated on 44 patients, with simultaneous polysomnography recordings.
• A prediction function was created to determine the likelihood of deep sleep epochs during NREM sleep.

Main Results:

• The algorithm achieved 66% sensitivity, 89% specificity, and 82% agreement for light/deep sleep detection in the training set.
• In the validation set, the algorithm yielded 65% sensitivity, 87% specificity, and 80% agreement.
• Combined with existing algorithms, this offers a near-complete sleep stage detection method using PAT and actigraphy.

Conclusions:

• The developed automatic algorithm effectively differentiates light and deep sleep stages based on PAT signals.
• This algorithm significantly advances unattended ambulatory sleep monitoring by enabling detailed sleep stage analysis without EEG.
• The findings support the use of PAT-based algorithms for comprehensive sleep assessment in clinical and research settings.