Sleep, Sleep Apnea
Conditions
Keywords
Automated Sleep Staging, Automated Apnea Detection
Brief summary
This study aims to evaluate the accuracy of apnea detection and automated sleep analysis by the Dreem dry-EEG headband and deep learning algorithm in comparison to the consensus of 5 sleep technologists' manual scoring of a gold-standard clinical polysomnogram (PSG) record in adults during a physician-referred overnight sleep study due to suspicion of sleep-disordered breathing.
Detailed description
The study will enroll up to 70 adults who are referred to the Stanford Sleep Medicine Center by their physician for an overnight polysomnographic sleep study due to suspicion of sleep-disordered breathing, with the aim of collecting 60 usable data sets (i.e., eligible subjects with high-quality PSG and Dreem recordings). Upon arrival to the clinic, patients provide informed consent, are interviewed to determine eligibility, and complete a detailed demographic, medical, health, sleep, and lifestyle questionnaire (Alliance Sleep Questionnaire; ASQ). After the ASQ, participants are fitted with the PSG and the Dreem headband by the sleep technologist. During the PSG sleep study, the Dreem headband records EEG, pulse, oxygen saturation (SO2), movement, and respiratory rate. Many participants may undergo a split-night study with a continuous positive airway pressure (CPAP) device during their participation, as deemed necessary by the clinical staff pursuant to the sleep study. The PSG data from the first 30 eligible participants will be manually scored by 5 sleep technologists. These manually-scored PSG data files (referred to as the training dataset) will be synchronized with Dreem data files from the same night and the synchronized files will be used to train Dreem's deep learning algorithms. Following training, the algorithms will be deployed to automatically score the final 30 participants' Dreem datasets (testing dataset). Finally, PSG records for the second 30 participants will be provided to the sponsor and manually scored by 5 sleep technologists. The manual scoring results will be compared to the Dreem automatic analysis to determine the accuracy of Dreem's apnea-hypopnea index (AHI) severity detection and sleep staging algorithms.
Interventions
DIAGNOSTIC_TESTDreem
Dreem Band to be worn by each participant while undergoing in-lab sleep study with PSG. Pursuant to the physician-ordered diagnostic study, clinical staff may determine a split-night study with CPAP to be appropriate for any participant, simultaneous with the PSG and Dreem.
Sponsors
Stanford University
Dreem
Study design
Allocation
NA
Intervention model
SINGLE_GROUP
Primary purpose
DIAGNOSTIC
Masking
NONE
Intervention model description
Dreem
Eligibility
Sex/Gender
ALL
Age
18 Years to 70 Years
Healthy volunteers
No
Inclusion criteria
* 18-70 years of age * Capable of providing informed consent * Suspicion of sleep breathing disorder (both diagnostic and split-night studies)
Exclusion criteria
* Concomitant diagnosis of a sleep disorder other than sleep apnea syndrome or insomnia * Morbid obesity (BMI \> 39) * Use of benzodiazepines, nonbenzodiazepine (Z-drugs), or Gammahydroxybutyrate (GHB) the day/night of the study * Concomitant diagnosis of cardiopulmonary or neurological comorbidities (such as heart failure, COPD, neurodegenerative conditions)
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Apnea-Hypopnea Index (AHI) severity agreement | Day 1 | AHI severity (normal \[\<5\], mild \[5-14\], moderate \[15-29\], severe \[\>29\]) as automatically determined by the Dreem headband compared to the AHI severity determined by the consensus of 5 sleep technologists' scoring of the subject's PSG record from the same night. |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| EEG Virtual Channel signal quality agreement | Day 1 | Comparison of a proprietary signal quality metric between Dreem and PSG virtual channel, calculated as a ratio of supra-threshold quality signal time over total time in bed trying to sleep (lights off to lights on). |
| Wake After Sleep Onset (WASO) time agreement | Day 1 | Total time (in minutes) the subject spends awake from sleep onset to sleep end as automatically determined by the Dreem headband compared to the WASO determined by the consensus of 5 sleep technologists' scoring of the subject's PSG record from the same night. |
| Time in N1 sleep stage agreement | Day 1 | Total time (in minutes) the subject spends in AASM N1 sleep stage as automatically determined by the Dreem headband compared to the N1 time determined by the consensus of 5 sleep technologists' scoring of the subject's PSG record from the same night. |
| Total Sleep Time (TST) agreement | Day 1 | Total time (in minutes) the subject spends asleep as automatically determined by the Dreem headband compared to the TST determined by the consensus of 5 sleep technologists' scoring of the subject's PSG record from the same night. |
| Time in N3 sleep stage agreement | Day 1 | Total time (in minutes) the subject spends in AASM N3 sleep stage as automatically determined by the Dreem headband compared to the N3 time determined by the consensus of 5 sleep technologists' scoring of the subject's PSG record from the same night. |
| Time in REM sleep stage agreement | Day 1 | Total time (in minutes) the subject spends in AASM REM sleep stage as automatically determined by the Dreem headband compared to the REM time determined by the consensus of 5 sleep technologists' scoring of the subject's PSG record from the same night. |
| Time in N2 sleep stage agreement | Day 1 | Total time (in minutes) the subject spends in AASM N2 sleep stage as automatically determined by the Dreem headband compared to the N2 time determined by the consensus of 5 sleep technologists' scoring of the subject's PSG record from the same night. |
Countries
United States
Outcome results
None listed