Performance=C2=A0of=C2=A0a=C2=A0Wireless=C2=A0Dry=C2=A0Sensor=C2=A0S= ystem=C2=A0in=C2=A0Automatically Monitoring=C2=A0Sleep=C2=A0and=C2=A0Wakefu= lness

Page 1=

Performance=C2= =A0of=C2=A0a=C2=A0Wireless=C2=A0Dry=C2=A0Sensor=C2=A0System=C2=A0in=C2=A0Au= tomatically

Monitoring=C2=A0= Sleep=C2=A0and=C2=A0Wakefulness

Stephan E F=C3=A1br= egas1, Jack Johnstone2, John R Shambroom1

1Zeo, Inc, Newton, MA, USA 2Valley Sleep Center, Burbank, CA, USA

Introduction:=

The ability to wirel= essly monitor and track sleep and wakefulness

in a way that is non= -invasive and inexpensive could be of utility in

many areas including= research medicine and education One

Results:

Mean sleep/wake mea= surements between the three systems appeared to be similar. The WS is desig= ned to

accurately detect= wake periods that last for at least 2 minutes. Therefore, number of awaken= ings is reported as

awakenings lasting= at least 4 continuous epochs

many areas, includin= g research, medicine, and education. One

system with the pote= ntial to effectively accomplish this has been

developed, which is= composed of a single channel no-prep dry

fabric sensor that= rests on the forehead through the use of a

headband. Data are= then wirelessly transmitted to an

accompanying base st= ation and processed through a neural

network to determine= periods of sleep and wakefulness

awakenings lasting= at least 4 continuous epochs.

Sleep=C2=A0and= =C2=A0Wake Summary=C2=A0Measures

ACT<= /div>

Latency=C2=A0to= =C2=A0Persistent=C2=A0Sleep

10 continuous minut= es

LPS=C2=A0(min=C2= =B1SEM)

31.8=C2=B116.1

33.0=C2=B116.5

36.4=C2=B116.3<= /b>

34.7=C2=B115.8<= /b>

Number=C2=A0of= =C2=A0Awakenings

Occurring after ons= et of persistent sleep

NA=C2=A0(#=C2=B1= SEM)

2.6=C2=B10.7=

2.2=C2=B10.4=

2.2=C2=B10.4

3.3=C2=B10.6

network to determine= periods of sleep and wakefulness.

This study was condu= cted to assess the accuracy of derived

measures of this wir= eless system (WS) and Actigraphy (ACT) in

comparison to polyso= mnography (PSG).

Methods:

Participants:=

g

p

p

and lasting at leas= t 2 continuous minutes

(
)
Total=C2=A0Sleep= =C2=A0Time

TST=C2=A0(min=C2= =B1SEM)

312.8=C2=B117.8= 303.3=C2=B117.7 296.8=C2=B116.7 313.2=C2=B113.9

Wake=C2=A0After= =C2=A0Sleep=C2=A0Onset

WASO=C2=A0(min= =C2=B1SEM) 41.7=C2=B113.4

27.8=C2=B13.7

47.1=C2=B18.5

48.5=C2=B19.7

Sleep=C2=A0Effic= iency

SE=C2=A0(%=C2=B1= SEM)

86.7=C2=B13.4

84.5=C2=B14.0

82.6=C2=B13.6

86.4=C2=B12.4

Sensitivity and spe= cificity to sleep were

Participants:=

=E2=80=A210 adults= (4 female)

=E2=80=A233.7 years= old (=C2=B110.7,SD)

=E2=80=A2No sleep co= mplaints

Study Protocol:

=E2=80=A2Sleep in th= e laboratory at the participant=E2=80=99s habitual bedtime

=E2=80=A2Concurrent= measurement of PSG, WS, and ACT

Sensitivity and spe= cificity to sleep were

higher in the WS th= an in ACT when

compared to PSG. In= ter-rater agreement

of sleep/wake betwe= en the WS and PSG

was moderate-to-hig= h, while agreement

between ACT and PSG= was moderate-to-

low

Sleep=C2=A0Sensi= tivity Sleep=C2=A0Specificity Accuracy Cohen's=C2=A0kappa

WS=C2=A0v=C2=A0T= 1

96.2%

72.9%

90.7%

0.63

WS=C2=A0v=C2=A0T= 2

95.9%

69.1%

88.4%

0.57

ACT=C2=A0v=C2=A0= T1

92.8%

65.1%

84.8%

0.41

ACT=C2=A0v=C2=A0= T2

93.5%

65.0%

84.6%

0.43

As an example of a= typical night from this data set, below is a representation of one subject= =E2=80=99s

(35F) night of slee= p.

Concurrent measureme= nt of PSG, WS, and ACT

=E2=80=A2PSG data co= llected with Cadwell Easy II PSG, sampled at 200

samples per second

=E2=80=A2WS data wer= e sampled at 128 samples per second

=E2=80=A2ACT data we= re collected on a Mini-Mitter Actiwatch 64, epoch

length set at 30 sec= onds

=E2=80=A2Sleep recor= ds were scored by 2 trained technicians (T1 and T2)

low.

LPS<= /div>
(min)

NA
(#)<= /div>
TST<= /div>
(min)

WASO=

(min)

SE
(%)<= /div>
WS
42

4

249

44

75

Stages 3 and 4

Wake

Stages 1 and 2

REM

Sleep

p

y

(

)

according to Rechtsc= haffen & Kales

=E2=80=A2Sleep recor= ds were scored automatically by the WS via its neural

network

=E2=80=A2ACT records= were scored automatically by Actiware 5.0 software

at medium wake thres= hold sensitivity (wake threshold value of

40 activity counts)<= /nobr>

T1
40

3

263

31

79

T2
41

2

265

28

80

ACT<= /div>
37

5

290

38

87

=E2=80=A26 subjects= contributed two nights of recordings and 4 subjects

one night, resulting= in 16 total records

=E2=80=A2Inter-rate= r agreement was performed on summary statistics of

sleep parameters on= a 30-second epoch-by-epoch basis for

agreement/disagreem= ent of sleep/wake decisions.

Support: Th= is research was funded by Zeo, Inc, Newton, MA, USA.

Conclusion:<= /nobr>

The wireless system= shows promise as an effective way to distinguish wake and sleep during the= night and has

potential as a cos= t-effective ambulatory system that can distinguish sleep and wake, while al= so providing sleep stage

information over= the course of nighttime recordings.