Med Biol Eng Comput (2012) 50:1147–1154
DOI 10.1007/s11517-011-0847-7
SPECIAL ISSUE - ORIGINAL ARTICLE
Development of a smart garment to reduce kyphosis during daily
living
Edmond Lou • Gabrielle C. Lam • Douglas L. Hill •
M. S. Wong
Received: 27 March 2011 / Accepted: 4 December 2011 / Published online: 18 December 2011
Ó International Federation for Medical and Biological Engineering 2011
Abstract Many of the aches and pains of adults are the
result of the long-term effects of bad posture or body
misalignment. Postural kyphosis in adolescence, which is
an excessive rounding of the upper spine, may be one of
the effects of poor standing and sitting habits. A smart
garment, consisting of a harness and two data-sensor log-
gers, was developed to monitor and provide vibration
feedback to wearers to improve their posture during daily
activities. Laboratory tests verified that the garment could
provide an accuracy of 2 ± 2° during static measurement
and 3 ± 2° during stable or slowly changing posture
activities and 4 ± 4° during rapidly changing posture
activities. Four volunteers wore the system for 3 h per day
and for 4 consecutive days. The feedback was provided on
the last 2 days and the kyphotic angle reduced by 8 ± 1°
and 8 ± 2° on the last 2 days, respectively. Although the
long-term effects of reminding the subjects’ posture is still
not clear, a short-term result shows promise that the smart
garment may be able to improve the kyphosis.
Keywords Kyphosis  Monitoring  Posture control 
Data loggers  Biofeedback
E. Lou (&)  D. L. Hill
Department of Research and Technology Development, Alberta
Health Services, 10230–112 Ave, Edmonton, AB T5G 0B7,
Canada
e-mail: elou@ualberta.ca
E. Lou  G. C. Lam
Department of Electrical and Computer Engineering, University
of Alberta, Edmonton, AB T6G 2V4, Canada
M. S. Wong
Department of Health Technology and Informatics, The Hong
Kong Polytechnic University, Hung Hom, Hong Kong
1 Introduction
‘‘Straighten up!’’ is not an uncommon phrase in the
everyday North American household. A parent’s concern
for the child’s posture is not without basis. Apart from the
cosmetic aspect of a ‘‘hunchback’’, improper posture has
substantial implications to one’s physical health, and is a
particularly significant problem to the growing, developing
child. In fact, the aches and pains associated with old age
are oftentimes rooted in postural habits affecting spinal
development during crucial growth stages of adolescence
[24].
Postural kyphosis, the rounding of the back in the sag-
ittal plane (as seen from the side), is a flexible spinal
deformity that develops due to poor posture [19]. Unlike
kyphosis of pathogenic origin such as misshaped vertebrae,
postural kyphosis presents as a smooth rounding of the
back upon forward bending [15, 19, 25] and can be cor-
rected entirely at the will of the individual with extension
of back muscles [19]. Although its cosmetic implications
are often parents’ most immediate consideration, recent
studies suggest that there are adverse effects of postural
kyphosis upon physical [14, 20] and respiratory function
[20], relation to neurologic problems [15] and, very
importantly, back pain [5, 13, 16, 20, 24, 26].
Maintaining improper posture for prolonged time exerts
significant stress on the spine [16, 24]. Increased spinal
loading can cause protrusion or rupture of the vertebral
discs, which, when compressed against the spinal nerve,
gives rise to back pain [13]. Excessive biomechanical stress
on vertebral endplates can over time render vertebrae to
become wedge shaped, thereby worsening the kyphotic
curve [14]. Lax ligaments and weak back muscles are other
detrimental results leading to poor spinal support [13, 26]
and further exacerbating the condition. Maintaining proper
123
1148
posture during a child’s developmental stage is thus a key
to reducing the likelihood of aches and pains in later years
of life [24].
Unfortunately, with the advancement and increased
availability of computer-related activities, children nowa-
days may spend a large portion of their leisure time
sitting—more often than not—with improper posture.
Psychosocial factors, such as depression, anxiety or inse-
curity, may also be underlying reasons for poor posture [9].
Irrespective of the cause, improper posture can become a
habit difficult to reverse. Treatment methods that have
received much attention in research include exercise ther-
apy and orthosis (bracing). The primary aim of exercise
regimes is to strengthen back muscles and increase spinal
flexibility of the patient [27]. It is sometimes prescribed to
children to assist in maintaining proper posture while the
kyphotic deformity is still mild [10, 24]. Orthosis, the
standard treatment for moderate kyphosis, supports trunk
alignment via a passive force [27]. Its active component,
however, is believed to provide the greatest therapeutic
effect [1, 8]. By pulling away from uncomfortable pressure
points in brace, the patient is encouraged to actively use
his/her muscles to maintain proper posture and develop
greater awareness of trunk alignment [1, 8].
Despite what appears to be a general consensus
regarding the importance of proper posture in children [5,
13–16, 26], formal medical treatment of postural kyphosis
is generally deemed unnecessary [14, 24], on the basis that
the deformity usually lessens with time [4, 15]. In North
America, minimal treatment is provided to patients dis-
playing mild curvatures, defined by a kyphotic angle
between 20 and 45°, determined radiographically using
Cobb’s method [15, 25]. The extent of treatment involves
patient education, periodic examination and exercise regi-
mens [15, 24]. Active treatment, namely bracing, is pre-
scribed only when the kyphotic curve progresses beyond
50° [15]. Even though an adolescent patient is prescribed
exercise or orthotic therapy, patient compliance remains an
enormous impediment to effective treatment [1, 4, 10].
Physical complications associated with prolonged brace
usage have been commonly reported, including reduced
spinal flexibility, permanent deformation of the rib cage or
soft tissue at pressure points [1], skin breakdown or aller-
gies and altered gastrointestinal motility [10]. More sig-
nificant, however, is its bulkiness, being uncomfortable
especially in hot weather, limiting activities and attracting
unwanted attention, act as a great hurdle for young indi-
viduals undergoing sensitive stages of life [1, 10].
In light of existing challenges, recent studies have
directed attention toward development of less conspicuous
posture training systems, many of which rely on biofeed-
back mechanisms. Biofeedback involves electronic usage
of bodily signals to inform an individual of his/her own
123
Med Biol Eng Comput (2012) 50:1147–1154
functional changes, of which one would not normally be
conscious [7]. By acting as an artificial sixth sense, bio-
feedback encourages reliance on oneself to correct and
control particular problems [7]. For instance, biofeedback
is now commonly used in tele-rehabilitation of patients
with neuromuscular difficulties displaying postural insta-
bility [11, 17, 21, 22], as well as for fall detection in the
elderly population [6]. More recent studies investigate its
potential for improving spinal deformities.
In this study, an inconspicuous smart garment was
developed to continuously monitor an individual’s posture
by direct measurement of the kyphotic angle. Feedback is
provided to its user when his/her kyphotic angle exceeds
personalized thresholds during daily activities. This paper
reports the smart garment system, the accuracy and reli-
ability of the system and the preliminary trial results.
2 Methods
2.1 Smart garment
The smart garment system consists of a harness and two
data-sensor loggers. The harness-like device, worn on the
upper body (Fig. 1a), was put over the arms like a back-
pack and the shoulder straps could be adjusted until snug.
A waist strap was wrapped around the waist to secure the
harness. A flexible plastic strap extended down the back,
approximately from vertebrae T1 to T12. The data-sensor
loggers attached to the harness with Velcro, along the back
on the plastic strap. Each logger consists of a plastic box
(55 mm 9 35 mm 9 15 mm, weight 30 g).
y1
z1
α1
g1
φ
y2
α2
z2 g2
Kyphotic angle = 180 - φ
(a) (b)
Fig. 1 a The smart garment system and b the determination of the
kyphotic angle from the position of two logger modules along the
spine
1148
posture during a child’s developmental stage is thus a key
to reducing the likelihood of aches and pains in later years
of life [24].
Unfortunately, with the advancement and increased
availability of computer-related activities, children nowa-
days may spend a large portion of their leisure time
sitting—more often than not—with improper posture.
Psychosocial factors, such as depression, anxiety or inse-
curity, may also be underlying reasons for poor posture [9].
Irrespective of the cause, improper posture can become a
habit difficult to reverse. Treatment methods that have
received much attention in research include exercise ther-
apy and orthosis (bracing). The primary aim of exercise
regimes is to strengthen back muscles and increase spinal
flexibility of the patient [27]. It is sometimes prescribed to
children to assist in maintaining proper posture while the
kyphotic deformity is still mild [10, 24]. Orthosis, the
standard treatment for moderate kyphosis, supports trunk
alignment via a passive force [27]. Its active component,
however, is believed to provide the greatest therapeutic
effect [1, 8]. By pulling away from uncomfortable pressure
points in brace, the patient is encouraged to actively use
his/her muscles to maintain proper posture and develop
greater awareness of trunk alignment [1, 8].
Despite what appears to be a general consensus
regarding the importance of proper posture in children [5,
13–16, 26], formal medical treatment of postural kyphosis
is generally deemed unnecessary [14, 24], on the basis that
the deformity usually lessens with time [4, 15]. In North
America, minimal treatment is provided to patients dis-
playing mild curvatures, defined by a kyphotic angle
between 20 and 45°, determined radiographically using
Cobb’s method [15, 25]. The extent of treatment involves
patient education, periodic examination and exercise regi-
mens [15, 24]. Active treatment, namely bracing, is pre-
scribed only when the kyphotic curve progresses beyond
50° [15]. Even though an adolescent patient is prescribed
exercise or orthotic therapy, patient compliance remains an
enormous impediment to effective treatment [1, 4, 10].
Physical complications associated with prolonged brace
usage have been commonly reported, including reduced
spinal flexibility, permanent deformation of the rib cage or
soft tissue at pressure points [1], skin breakdown or aller-
gies and altered gastrointestinal motility [10]. More sig-
nificant, however, is its bulkiness, being uncomfortable
especially in hot weather, limiting activities and attracting
unwanted attention, act as a great hurdle for young indi-
viduals undergoing sensitive stages of life [1, 10].
In light of existing challenges, recent studies have
directed attention toward development of less conspicuous
posture training systems, many of which rely on biofeed-
back mechanisms. Biofeedback involves electronic usage
of bodily signals to inform an individual of his/her own
123
Med Biol Eng Comput (2012) 50:1147–1154
functional changes, of which one would not normally be
conscious [7]. By acting as an artificial sixth sense, bio-
feedback encourages reliance on oneself to correct and
control particular problems [7]. For instance, biofeedback
is now commonly used in tele-rehabilitation of patients
with neuromuscular difficulties displaying postural insta-
bility [11, 17, 21, 22], as well as for fall detection in the
elderly population [6]. More recent studies investigate its
potential for improving spinal deformities.
In this study, an inconspicuous smart garment was
developed to continuously monitor an individual’s posture
by direct measurement of the kyphotic angle. Feedback is
provided to its user when his/her kyphotic angle exceeds
personalized thresholds during daily activities. This paper
reports the smart garment system, the accuracy and reli-
ability of the system and the preliminary trial results.
2 Methods
2.1 Smart garment
The smart garment system consists of a harness and two
data-sensor loggers. The harness-like device, worn on the
upper body (Fig. 1a), was put over the arms like a back-
pack and the shoulder straps could be adjusted until snug.
A waist strap was wrapped around the waist to secure the
harness. A flexible plastic strap extended down the back,
approximately from vertebrae T1 to T12. The data-sensor
loggers attached to the harness with Velcro, along the back
on the plastic strap. Each logger consists of a plastic box
(55 mm 9 35 mm 9 15 mm, weight 30 g).
y1
z1
α1
g1
φ
y2
α2
z2 g2
Kyphotic angle = 180 - φ
(a) (b)
Fig. 1 a The smart garment system and b the determination of the
kyphotic angle from the position of two logger modules along the
spine
Med Biol Eng Comput (2012) 50:1147–1154
2.2 Kyphotic angle measurement
The kyphotic angle was measured using two units placed
on the vertebral column of the subject. The / angle was the
difference between the tilt angle of the upper unit a1 and
the tilt angle of the lower unit a2, as shown in Fig. 1b.
/ ¼ 180o  ða2  a1 Þ
o modes: laboratory sampling and long-term sampling. Lab-
Kyphotic angle ¼ 180  / ¼ a2  a1
2.3 Data-sensor logger
The block diagram of the data-sensor logger is shown in
Fig. 2. Both loggers, called a master and slave pair, con-
sisted of the same hardware except that a buzzer and a
vibration motor were placed on the master unit. The major
integrated circuit (IC) of the unit was the CC2430 (Texas
Instrument, USA), which is a wireless transceiver
combined with a microcontroller (MCU). The maximum
current consumption of this IC was 27 mA at 3 V with
radio-on and 7 mA for processing with radio-off. A real-time
clock (M41T93) (RTC) was used to wake up the CC2430
for data collection to provide the time stamp and to cal-
culate the kyphotic angle. A backup battery was connected
to the RTC to ensure that data were retained even if power
failure occurred. A 4-MB flash memory was used to store
measured data, which was able to store 262,144 samples.
Power management circuitry was used to convert a single
3.7 V, 500 mAh lithium-ion battery to a 3 V regulated
supply voltage. Both master and slave units used the same
three-axis accelerometer (LIS302SG, ST Microelectronics
Inc.) and two-axis gyroscope (LPR450AL, ST Microelec-
tronics Inc.) to measure the orientation. Low power, high
sensitivity and small packages (0.65 mA at 3 V, 0.435 V/g
and 3 9 5 9 0.9 mm for accelerometer and 6.8 mA at
3 V, 2 mV/dps (degree-per-second) and 4 9 5 9 1 mm
for gyroscope) were the selection criteria in these two
sensors.
3-axis accelerometer and 2–axis gyroscope
12-bit A/D
Vibration
Memory MCU General
Motor
& I/Os
RTC
Buzzer
Transceiver
CC2430
PC Optional – only on the Master unit
Fig. 2 The block diagram of the data-sensor logger
1149
2.4 Wireless communication and firmware
The communication between the two units was implemented
using the ZigBee protocol. It operated in the industry-sci-
entific medical (ISM) frequency band (2.4 GHz) and had a
range of 30 m for normal operation. The two units (master
and slave) could communicate with a PC via a ZigBee USB
ð1Þ dongle. The system could be set to two different operation
ð2Þ
oratory sampling allowed researchers to set up and validate
the functionality of the system. Commands from the PC were
sent through a ZigBee dongle and waited for its reply. If
invalid commands were received on both units, no response
would be given and a message would display on the com-
puter screen after 1 s. When valid commands such as ‘‘Set
Sample Rate’’ or ‘‘Set Clock’’ or ‘‘Measured Angle’’, were
received, both units would respond, but the slave was
delayed 10 ms to avoid data collision (Fig. 3a). Special
commands such as ‘‘feedback command’’ or ‘‘threshold
value’’ were only sent to the master. After the set up, the units
would be placed in the long-term sampling mode, which
recorded the posture orientation at 30-s intervals. The clock
was synchronized at the beginning to ensure that both units
would wake up at the same time. Both units were put into
processing mode to save power. The communication
between the dongle and both units was off. The wireless
transceivers woke up when a sample was acquired. The
typical sample rate was one sample per 30 s, and it could be
adjustable by the user. The master then sent a command to
the slave to request the angle measurement. If no answer was
received from the slave, the master re-transmitted the com-
mand up to 16 times. After receiving the angle from the slave,
the master calculated the kyphotic angle. If it was larger than
the user-defined threshold (undesirable posture detected),
the master would request 20 more samples at 10 Hz sample
rate to validate the bad posture (Fig. 3b). An alarm was
generated if all 20 measurements were above the threshold.
This method was chosen to avoid providing unnecessary
feedback during transient activities such as bending down to
pick up a dropped object. The angle measurement on each
unit was saved to its own on board memory with a flag
indicating whether an alarm was given. The clock then
synchronized again before both units switched back to pro-
cessing mode.
2.5 Angle calculation algorithm
Two sensors, an accelerometer and a gyroscope, were used
to calculate the inclination angle continuously. A custom
algorithm was developed to minimize computational power
and reduce the size of the source code. This algorithm
compromised between the speed of calculation and the
acceptable error. The following section describes the
123
1150 Med Biol Eng Comput (2012) 50:1147–1154

Fig. 3 a Lab sample (a) PC Master Slave

communication protocol. b The
long-term communication Broadcast:
protocol Request angle
request angle

Calculate Calculate
angle angle
Current angle

Store master 10ms guard

angle interval

Store slave angle

Calculate
kypothic angle
Update display

(b) Master Slave

Wake up Wake up
Request angle
wireless wireless
Calculate
Calculate angle
Current angle angle

Check
kyphotic angle Resend request if no answer was
received
Alarm:
Request Additional Samples
Take 20
additional Take 20 additional
samples samples
Additional Samples
Check new
samples / give Shut down
alarm wireless

Shut down
wireless

 
sources and magnitudes of errors obtained from the indi- 0:732
vidual sensor and the developed algorithm. Dh0 ¼ arcsin ¼ 0:096 ð4Þ
435
During static or slow movement, the acceleration from  
435  0:732
each axis relative to gravity was used to calculate the change Dh90 ¼ 90  arcsin ¼ 3:33 ð5Þ
of the angle. The sensitivity of the selected accelerometer 435
was 435 mV/g (gravity) and the system used a 12-bit ADC. To reduce the error due to orientation near 90°, the z-axis
At a supply voltage of 3 V, 12-bit resolution was equivalent value was used when the angle h was over 45°. Therefore, the
to 4,096 steps, which meant that each step was 0.732 mV maximum error occurred at 45° and it was improved from
(3 V/4,096 steps). At 0 g, the output voltage was 1.5 V (mid- 3.33° to 0.14°. Use of the z-axis value increased the range
range). If the sensor only tilted in the yz-plane, where the error due to orientation was small.
 
Vout;Y  1500  pffiffiffi 
h ¼ arcsin ð3Þ  1= 2  435  0:732
435 Dh45 ¼ 45  arcsin ¼ 0:14
435
Since this was a nonlinear function, the sensitivity ð6Þ
varied depending on the orientation of the sensor. The
minimum error occurred at 0° relative to the horizontal and To calculate the angle from the gyroscope, the angular
maximum error at 90°. velocity xGyro was integrated numerically:

123
Med Biol Eng Comput (2012) 50:1147–1154 1151

hkþ1 ¼ hk þ Dt  xGyro ð7Þ

This calculation was updated at 140 Hz (Dt & 7 ms) to
minimize the integration error. The gyroscope has a
sensitivity of 3.33 mv
=s, which corresponds to a minimal
detectable change of
VADCStep 0:732
xmin ¼ ¼ ¼ 0:22 =s ð8Þ
Sensitivity 3:33
At the beginning of the data collection, the smart
garment user was required to stand still at a proper posture
for 1 s to get a baseline reading. The accelerometer was
updated at 30 Hz and the last 16 angle measurements were
stored into a buffer. During the static or slowly moving
cases, total acceleration from x, y and z axis should be
within 1 ± 0.1 g. The angle can be accurately calculated
from the accelerometer only. This approach reduced the
calculation time. If the total acceleration was out of the
1 ± 0.1 g boundary (\0.9 g or [1.1 g) or the angle
measurement was greater than 2° difference than the
buffer data (last 16 data), the gyroscope value would be
used as the angle measurement. This represents rapidly
changing orientation. The gyroscope values were used to
calculate the orientation while the total acceleration was
beyond the 1 ± 0.1 g boundary. The angle calculation Fig. 4 The rotating wheel apparatus
switched to the accelerometer method while acceleration
was within the boundary. Using this approach also reduced the logger module to ensure that it was flat. After proper
the long-term drifting in the gyroscope. setup of the equipment, real-time samples from the logger
module were taken as the experimenter turned the rotating
2.6 System validations arm counterclockwise at 10° intervals through 360°.
Between each step, the experimenter held the logger
Validation of the smart garment system was conducted in a module at its new position for approximately 5 s. The
static stage and a dynamic stage using a rotating wheel procedure was repeated another two times for the master
apparatus (Fig. 4) and an optical motion tracking system, logger for a total of three data sets. The same process was
respectively. Details of the associated procedures and replicated for validation of the slave module.
apparatus are described below. After the system was vali- The second portion of the experiment sought to assess
dated, four young men (age 28 ± 5 years) used the system measurement accuracy of the posture monitoring system
in a laboratory environment while they were sitting and under dynamic conditions. A series of regimented activities
working at their computers. were performed by a volunteer subject while wearing the
The initial stage of validation involved the use of a smart garment in a gait and motion laboratory with eight
rotating wheel apparatus, with which a logger module was motion analysis camera systems (Motion Analysis Corpo-
positioned and rotated to known angles. By comparing the ration, USA). Kyphotic angles logged by the smart garment
angle measured manually from the rotating wheel to that were compared with those determined by the optical
logged by the posture monitoring modules, measurement motion tracking system. The motion analysis system has a
accuracy under static conditions was assessed. The appa- field tested accuracy of 0.5 mm to locate the 3D coordi-
ratus consisted of a level plastic arm, on which a logger nates of the reflective markers. At 45° orientation, with
module was stationed. The rotating wheel had angle marks markers 40 mm apart, a 0.5 mm error will provide accu-
on the wheel, which allowed for manual measurement to a racy to within 1°. The smart garment was donned by a
precision of ±1°. The rotating wheel was secured to a level volunteer subject, with shoulder and waist straps adjusted
surface using a metal clamp, fixing the position of the to achieve a proper fit. Two reflective markers were taped
apparatus and minimizing undesired movement. The mas- onto the outer face of each logger module. The master
ter logger was initially laid horizontally, denoting the 0° module was adhered to the upper thoracic region of the
position. A leveler was placed atop the rotating wheel and spine (around the T3 vertebra position) and the slave

123
1152 Med Biol Eng Comput (2012) 50:1147–1154

module to the lower thoracic region (around the T12 ver-
tebra position). Three regimented activities were performed
by the volunteer, namely: (1) typical walking, (2) walking
with slouched posture and (3) mixed activity. The mixed
activity was designed to represent an action typical of that
performed in daily living; the subject walked to the middle
of the room, sat down on a chair, bent down quickly to pick
up an object, resumed an upright sitting position, got up
from the chair and continued to walk to the other end of the
room at a moderate pace. Each of the three actions was
repeated for two trials.
The smart garment system was tested by four healthy
control subjects (male, age 28 ± 5 years) over a 4-day
period to evaluate its effect on improving posture during
daily activities. The smart garment was worn by each
subject for 3 h each day. The data collection rate was one
sample per 30 s. On the first 2 days, no feedback was
provided. The normal sitting posture on day 1 was used as
the reference. A vibration feedback was provided on days 3
and 4 when the kyphotic angle was greater than 5° of the
reference angle and lasted beyond a 2-s interval (20 sam-
ples at 10 Hz).

3 Results

Results from the static measurements are shown in Fig. 5.

The average error was 2 ± 1° and 2 ± 2° for the master
and slave loggers, respectively. By comparing the angle
determined manually from the rotating wheel apparatus
and that logged by the modules, Pearson’s correlation
coefficient was [0.999 for both units. Results from the
dynamic measurements during normal and slouch walking
Fig. 5 Correlation between the actual angle as determined from the are shown in Table 1. The maximum kyphotic angle dif-
rotating wheel apparatus and that measured by both slave and master ference between the loggers and the camera measurements
was 3 ± 2°. During the mixed activities, the kyphotic
Table 1 The kyphotic angle measurements from the smart garment angle difference was 4 ± 4°. Figure 6 shows the master
and camera systems during walking upright and slouching and slave angles compared to the camera measurements
Kyphotic angle (°) during the mixed activities. The maximum kyphotic angle
Logger Camera Differences
difference was 13° when the subject was bending down to
measurement measurement pick up an object and then resumed an upright position.
The rapid change of posture created a significant error.
Walking 30.1 ± 3.6 26.7 ± 2.0 3.4 ± 1.8
However, 90% of the measured angles during the mixed
upright
activities were within 5° difference. During the daily
Slouching 35.7 ± 3.6 36.4 ± 2.1 1.0 ± 1.5
activities tests, all subjects had a slight improvement on

Fig. 6 The tilt angle from a the master and b slave units separately (The kyphotic angle = master angle ? 180 - slave angle)

123
Med Biol Eng Comput (2012) 50:1147–1154
Fig. 7 The average kyphotic angle of each subject (S1–S4) during the
test period
their posture with feedback. Feedback was provided at
every measurement (one sample per minute) as needed.
Better sitting posture is reflected by a smaller kyphotic
angle. Figure 7 shows the average kyphotic angle during
the test period. On comparing days 1 to 2, days 1 to 3,
and days 1 to 4, the kyphotic angles improved as 3 ± 3,
8 ± 1 and 8 ± 2°, respectively. When no feedback was
given (day 2), the average kyphotic angle was within 5°
of day 1. However, when the feedback were given on
days 3 and 4, the improvement was greater than 5°. The
two-tailed p values (\0.05) on day 1 versus day 4 of the
average kyphotic angle showed a statically significant
difference. All subjects showed an improvement on days
3 and 4. Subject 1 reported that he was very sensitive to
the vibration feedback and tried to maintain his posture
to minimize the vibration. The average number of
vibration feedbacks received by subjects 1–4 on days 3
and 4 were 35, 65, 40 and 38 out of 180 samples,
respectively. Subject 1 showed better posture and less
feedback than the other three. He also had superior
posture prior to the feedback. All subjects reported that
the smart garment was easy to put on and comfortable to
wear. The shoulder straps only needed to be adjusted on
the first day. A waist strap was wrapped around the
waist to secure the harness. No extra attention was
required during the tests.
4 Discussion
Postural kyphosis is a rounding back that can be improved
by intensive rehabilitation. In a study by Dworkin et al. [8]
and Birbaumer et al. [4], a device was developed and
evaluated that monitored spinal curvature indirectly via
measurement of the user’s torso length. Feedback was
provided to the subject whenever improper posture was
assumed for extended lengths of time. Using a mechanism
similar to the active component of an orthotic treatment, it
1153
was postulated that the biofeedback system would develop
an over-learned correction response that, with prolonged
usage, would become a pseudo-reflex [4]. Once a patient
has successfully completed a training period with the
feedback, the learned improved posture will be maintained
even without further feedback. Results of clinical trials
[4, 8] suggest that if worn compliantly, the developed
posture training device could halt curve progression in
patients with postural kyphosis. Currently, improved pos-
ture training devices have been developed, integrating the
use of accelerometer and/or gyroscope technology to
directly monitor the orientation of the spine to treat spinal
deformities [1–3, 18, 23, 27]. A primary objective of these
developments was to increase comfort of wear and reduce
notability of the device, so as to improve patient accep-
tance and facilitate effective treatment. The improvements
of the current design were: smaller size, lower power
consumption, lighter weight and easier to put on as a single
unit when compared with previous designs [2, 3, 28].
This preliminary study validated that an inconspicuous
smart garment was able to measure and provide feedback
on the kyphotic angle during daily activities. Faulty posture
might cause fatigue, muscular strain and, in later stages,
pain. Good posture also contributed to good appearance;
the person with good posture projected poise, confidence
and dignity. This approach was convenient, as the subjects
were able to train their posture anywhere and anytime. The
smart garment could be hidden under clothing and would
not attract undesired attention. There was no significant
difference on the kyphotic angle during the day compared
to when the subjects first put on the garment at the
beginning of the day. Data were automatically saved into
the system.
A smart garment which could monitor posture and
provide feedback to alert a subject to maintain a more
balanced posture was developed. Measurement accuracy
of this garment was within 5° over 90% of the time
during daily activities. The 5° accuracy is within typically
accepted accuracy of the Cobb angle [12]. Although a
direct comparison between the current technique and the
Cobb method could not be made, a moderate relationship
between the internal alignment and external appearance
could be assumed. Also, the threshold value of the
kyphotic measurement was personalized so that each
subject had a target range, which was attainable. The fast
checking algorithm as described in Sect. 2.4 was imple-
mented to avoid providing too many unnecessary feed-
backs during transient activities. Subjects who tested the
system did not notice too many false feedbacks. Although
the long-term effect of this biofeedback could not be
concluded, an immediate improvement was found. Further
clinical trials will be conducted to validate the long-term
results.
123
1154
References
1. Bazzarelli M, Durdle N, Lou E, Raso J (2001) A low power
portable electromagnetic posture monitoring system. Proceedings
of the 18th IEEE Instrumentation and Measurement Technology
Conference 1:619–623
2. Bazzarelli M, Durdle N, Lou E, Raso J (2001) A low power
hybrid posture monitoring system. Can Conf Electr Comp Eng
2:1373–1377
3. Bazzarelli M, Durdle NG, Lou E, Raso VJ (2003) A wearable
computer for physiotherapeutic scoliosis treatment. IEEE Trans
Instrum Meas 52(1):126–129
4. Birbaumer N, Flor H, Cevey B, Dworkin B, Miller ME (1994)
Behavioural treatment of scoliosis and kyphosis. J Psychosom
Res 38(6):623–628
5. Borenstein DG, Wiesel SW, Boden SD (2004) Low back and
neck pain: comprehensive diagnosis and management, 3rd edn.
Saunders Elsevier Inc., USA, pp 380–383
6. Bourke AK, Van de Ven PWJ, Chaya AE, O’Laighin GM, Nel-
son J (2008) The design and development of a long-term fall
detection system incorporated into a custom vest for the elderly.
30th Annual International Conference of the IEEE Engineering in
Medicine and Biology Society, pp 2836–2839
7. Davis CM (2009) Complementary therapies in rehabilitation:
evidence for efficacy in therapy, prevention, and wellness. Slack
Inc, USA
8. Dworkin B, Miller NE, Dworkin S, Birbaumer N, Brines ML,
Jonas S, Schwentker EP, Graham JJ (1985) Behavioral method
for the treatment of idiopathic scoliosis. Proc Nat Acad Sci USA
82(8):2493–2497
9. Feldenkais M (1949) Body and mature behaviour: a study of
anxiety, sex, gravitation and learning. New York International
University Press Inc, USA
10. Frontera WR, Silver JK, Rizzo TD (2008) Essentials of physical
medicine and rehabilitation: musculoskeletal disorders, pain and
rehabilitation: 2nd edn. Saunders Elsevier Inc, Philadelphia
11. George CL (1992) The posture monitor: an automated prompting
device for body alignment. Proceedings of the Johns Hopkins
National Search for Computing Applications to Assist Persons
with Disabilities, pp 143–145
12. Goh S, Price RI, Leedman PJ, Singer KP (2000) A comparison of
three methods for measuring thoracic kyphosis: implications for
clinical studies. Rheumatology 39(3):310–315
13. Gotasche A (2007) Postures: the book of the back, 2nd edn.
Anne-Lise Gotzsche, Great Britain
123
Med Biol Eng Comput (2012) 50:1147–1154
14. Hart ES, Merlin G, Harisiades J, Grottkau BE (2010) Scheuer-
mann’s thoracic kyphosis in the adolescent patient. Orthop Nurs
29(6):365–371
15. Heary RF, Albert TJ (2007) Spinal deformities: the essentials.
Thieme Medical Publishers Inc., USA, pp 192–196
16. Lewis JS, Valentine RE (2010) Clinical measurement of thoracic
kyphosis: a study of the intra-rater reliability in subjects with and
without shoulder pain. BMC Musculoskelet Disord 11(39)
17. Lorussi F, Scilingo EP, Tesconi A, Tognetti A, De Rossi D
(2003) Wearable sensing garment for posture detection, rehabil-
itation and tele-medicine. 4th International IEEE EMBS Special
Topic Conference on Information Technology Applications in
Biomedicine, pp 287–290
18. Lou E, Bazzarelli M, Hill D, Durdle N (2001) A low power
accelerometer used to improve posture. Can Conf Electr Comp
Eng 2:1385–1389
19. Lovell WW, Winter RB, Morrissy RT, Weinstein SL (2006)
Lovell and Winter’s Pediatric orthopaedics, 6th edition, vol 2.
Lippincott Williams & Wilkins, USA, p 7998
20. Macagno AE, O’Brien MF (2006) Thoracic and thoracolumbar
kyphosis in adults. Spine 31(19):S161–S170
21. Motoi K, Higashi Y, Kuwae Y, Yuji T, Tanaka S, Yamakoshi K
(2005) Development of a wearable device capable of monitoring
human activity for use in rehabilitation and certification of eli-
gibility for long-term care. Conf Proc IEEE Eng Med Biol Soc
1:1004–1007
22. Motoi K, Ikeda K, Kuawe Y, Yuji T, Higashi Y, Nogawa M, Tanaka
S, Yamakoshi K (2006) Development of an ambulatory device for
monitoring posture change and walking speed for use in rehabili-
tation. Conf Proc IEEE Eng Med Biol Soc 1:5940–5943
23. Nevins RJ, Durdle NG, Raso VJ (2002) A posture monitoring
system using accelerometers. Canadian Conference on Electrical
and Computer Engineering 2:1087–1092
24. Sinett T, Sinett S (2008) The truth about back pain: a revolu-
tionary, individualized approach to diagnosing and healing back
pain, 1st edn. Penguin Group, New York
25. Skaggs DL, Flynn JM (2006) Staying out of trouble in paediatric
orthopaedics. Lippincott Williams & Wilkins, USA, p 286
26. Widhe T (2001) Spine: posture, mobility and pain. A longitudinal
study from childhood to adolescence. Eur Spine J 10(2):118–123
27. Wong WY, Wong MS (2008) Smart garment for trunk posture
monitoring: a preliminary study. Scoliosis 3(7):1748–1761
28. Wong MS, Mak AFT, Luk KDK, Evans JH, Brown B (2001)
Effectiveness of audio-biofeedback in postural training for AIS
patients. Prosthet Orthot Int 25(1):60–70