Validation of Basis Science
Advanced Sleep Analysis
Estimation of Sleep Stages and Sleep Duration
Sarin Patel
Tareq Ahmed
Jonathan Lee
Biosignals Algorithms Engineer 1
Senior Software Engineer 1
Director of Research 1
Leslie Ruoff
Tejash Unadkat
Sleep Core Director 2
Director of Product 1
Abstract
Basis Science launched Advanced Sleep Analysis in January 2014 and became the first wrist-based tracker to detect
REM, deep (slow wave), and light Non-REM sleep stages. Advanced Sleep Analysis combines sleep stage, toss-andturn, sleep duration metrics, and more to provide users with a comprehensive view of their sleep.
To further develop and validate the sleep staging and sleep duration components of Advanced Sleep Analysis,
Basis is partnering with sleep researchers at the Stress and Health Research Program, a joint venture between the
University of California, San Francisco (UCSF), the San Francisco Veterans Affairs Medical Center (SFVAMC), and the
Northern California Institute of Research and Education (NCIRE). None of the institutions were compensated in any
form for these studies.
These researchers conducted sleep studies known as polysomnography, the gold standard for examining sleep, in
order to evaluate the Basis sleep algorithm’s estimation of sleep duration and sleep staging. Preliminary results from
these ongoing studies are described in this report. Sleep duration and sleep staging detected via the Basis Band were
compared to polysomnography (PSG) data scored by sleep technologists. The Advanced Sleep Analysis algorithm
demonstrated excellent agreement with polysomnography data for sleep duration (4.3% mean difference) and sleep
staging (r = 0.92).
1
Basis Science, San Francisco, CA,
2
University of California, San Francisco /San Francisco Veterans Affairs Medical Center/
Northern California Institute of Research and Education
Validation of Basis Science Advanced Sleep Analysis
Page 1
Introduction
During sleep, the body cycles between different sleep states in order to recover mentally and physically. Rapid eye
movement (REM) sleep is theorized to be essential for strengthening memories, boosting mood, and consolidating
information learned during the day[1],[2]. During deep sleep, your body repairs muscles and tissues, stimulates growth
and development, and boosts your immune system[3],[4],[5]. Light sleep allows the body to transition to either deep or
REM sleep[6].
Polysomnography (PSG), a comprehensive sleep study measuring several physiological signals, is the gold standard
for measuring these sleep states. PSG studies are typically used by clinicians to evaluate sleep quality in those with
suspected sleep disorders or by researchers in both normal subjects, or patients with sleep disorders. During PSG
studies, an array of sensors is attached to the patient, measuring activity in the brain, eye, muscles, and heart. These
signals are then analyzed and interpreted by a sleep technologist, who creates an in-depth report of the patient’s night
sleep, including time spent in each sleep state.
The Basis Band’s sensors measure heart rate, motion, temperature, and perspiration. Through combinations of multiple
sensors, the Basis sleep algorithm is able to measure physiological correlates of sleep stages, and thereby estimate
times each user was in light, REM or deep sleep. These measurements are not possible with sensors that only detect
motion (actigraphs). To ensure the accuracy of the sleep algorithm, Basis partnered with clinical sleep researchers who
performed studies comparing the Basis sleep algorithm with PSG data. Excellent agreement was observed between
Basis and PSG estimation of sleep duration and sleep states. Results from the initial phase of this study are described in
this report, with additional studies ongoing.
Methodology
To further develop and validate the sleep duration estimation and sleep staging of our Advanced Sleep Analysis, Basis
partnered with The Stress and Health Research Program, a joint venture between the University of California, San
Francisco (UCSF), the San Francisco Veterans Affairs Medical Center (SFVAMC), and the Northern California Institute
of Research and Education (NCIRE). Prior to work with SFVAMC, Basis developed, trained and validated the Advanced
Sleep Analysis algorithm on data from over 600 sleep events from studies at Basis. For initial external validation of the
algorithm, sleep studies, known as polysomnography (PSG), were performed at SFVAMC on 12 subjects for one or two
nights per subject for a total of 19 subject-nights. Basis B1 Bands were worn concurrently during these studies. PSG
data were scored by registered sleep technologists at SFVAMC. Analysis of sleep duration and sleep stage percentages
comparing the Basis sleep algorithm to PSG were performed at Basis.
Validation of Basis Science Advanced Sleep Analysis
Page 2
350
Results & Discussion
300
Detection of Sleep Stages
night sleep spent in each state between the
Basis sleep algorithm estimation and the
PSG (see Figures 1 and 2). The results were
Basis Duration
observed for duration and percentage of
(minutes)
250
Excellent correlation (r = 0.92) was
200
150
highly statistically significant (p < 0.01).
100
The agreement observed between Basis
and the scored PSG data is especially
50
encouraging considering even trained
sleep scorers routinely disagree in their
0
assessments of sleep stages. In a study of
0
50
100
over 2,500 trained sleep scorers recently
150
200
PSG Duration
published in the Journal of Clinical Sleep
Deep
Medicine, inter-scorer agreement was
Figure 1
250
300
350
(minutes)
REM
Light
Comparison of PSG and Basis estimation of sleep stage duration.
observed to be approximately 83%[7].
Given this inherent variability in scoring,
complete agreement between the
70
PSG and Basis is unlikely, and the Basis
sleep algorithm performance should be
60
overlaying sleep stages estimated by the
Basis sleep algorithm and PSG. As shown
in the hypnograms, the algorithm shows
good agreement with PSG for timing of
sleep stages.
Basis Percentage of Sleep Event
See Figure 3 for example hypnograms
(% )
considered very good.
50
40
30
20
10
0
0
10
20
30
40
50
PSG Percentage of Sleep Event
Deep
Figure 2
Validation of Basis Science Advanced Sleep Analysis
REM
60
70
(%)
Light
Comparison of PSG and Basis estimation of sleep stage as a percentage of total sleep time.
Page 3
Example Hypnogram #1
REM
Light
Deep
Midnight
1:12 am
2:24 am
3:36 am
4:48 am
Basis
6:00 am
7:12 am
PSG
Example Hypnogram #2
REM
Light
Deep
10:19 pm
11:31 pm
12:43 am
1:55 am
Basis
Figure 3
Validation of Basis Science Advanced Sleep Analysis
3:07 am
4:19 am
5:31 am
6:43 am
PSG
Example hypnograms overlaying PSG and Basis estimated sleep stage during sleep event.
Page 4
Estimation Of Sleep Duration
The Basis sleep algorithm estimation
10
of sleep duration matched closely with
of duration was 4.3% different from the
PSG*. A difference in duration of 10% or
less was observed in 89% of nights.
* Percentage difference calculated as the absolute
value of the difference between PSG and Basis,
divided by PSG
(hours)
4). On average, the Basis estimation
9
Measured Sleep Duration
PSG-measured sleep duration (see Figure
8
7
6
5
4
3
2
1
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Subject—Night
PSG
Figure 4
Basis
Comparison of PSG and Basis estimation of total sleep duration.
Summary
The Basis sleep algorithm demonstrated excellent agreement with polysomnography (PSG) data for both sleep duration
and sleep state estimation. Preliminary results are promising and demonstrate that the Basis Band provides a level of
sleep analysis previously unavailable outside a sleep laboratory.
Acknowledgments
The authors wish to acknowledge and thank Dr. Thomas Neylan and the staff of The Stress and Health Research
Program at the San Francisco VA Medical Center for their support and advice throughout these studies
Validation of Basis Science Advanced Sleep Analysis
Page 5
References
1. Diekelmann, S. & Born, J. The memory function of sleep. Nature Rev Neurosci. 11, 114-126 (2010).
2. Cartwright, R., Luten, A., Young, M., Mercer, P. & Bears, M. Role of REM sleep and dream affect in overnight mood
regulation: a study of normal volunteers. Psychiatry Res. 81, 1-8 (1998).
3. Kamdar, B., Needham, D. & Collop, N. Sleep Deprivation in Critical Illness: Its Role in Physical and Psychological
Recovery. J Intensive Care Med. 27, 97-111 (2012).
4. Veldhuis, J.D. & Iranmanesh, A. Physiological regulation of the human growth hormone (GH)-insulin-like growth
factor type I (IGF-I) axis: predominant impact of age, obesity, gonadal function, and sleep. Sleep. 19, S221-224 (1996).
5. Cardinali, D. & Esquifino, A. Sleep and the Immune System. Curr Immunol Rev. 8, 50-62 (2012)
6. Kishi, A. et. al. NREM Sleep Stage Transitions Control Ultradian REM Sleep Rhythm. Sleep. 34, 1423-1432 (2011).
7. Rosenberg, R.S. & Van Hout S. The American Academy of Sleep Medicine inter-scorer reliability program: sleep stage
scoring. J Clin Sleep Med. 9, 81-87 (2013)
Validation of Basis Science Advanced Sleep Analysis
Page 6