Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Alberts Et Al. - 2015 - Using Accelerometer and Gyroscopic Measures To Qua

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Journal of Athletic Training 2015;50(6):578–588

doi: 10.4085/1062-6050-50.2.01
Ó by the National Athletic Trainers’ Association, Inc original research
www.natajournals.org

Using Accelerometer and Gyroscopic Measures to


Quantify Postural Stability
Jay L. Alberts, PhD*†‡§; Joshua R. Hirsch*‡; Mandy Miller Koop, PhD*; David
D. Schindler*‡; Daniel E. Kana*; Susan M. Linder, DPT*‡; Scott Campbell*‡;
Anil K. Thota, MS*†

Downloaded from http://meridian.allenpress.com/jat/article-pdf/50/6/578/1617372/1062-6050-50_2_01.pdf by guest on 23 August 2020


*Department of Biomedical Engineering; †Center for Neurological Restoration; and ‡Cleveland Clinic Concussion
Center, Cleveland Clinic, OH; §Cleveland Functional Electrical Stimulation Center, L. Stokes Cleveland VA Medical
Center, OH

Context: Force platforms and 3-dimensional motion-capture to compare equilibrium scores produced by the NeuroCom and
systems provide an accurate method of quantifying postural iPad2 devices. Limits of agreement was defined as the mean
stability. Substantial cost, space, time to administer, and need bias (NeuroCom  iPad) 6 2 standard deviations. Mean
for trained personnel limit widespread use of biomechanical absolute percentage error and median difference between the
techniques in the assessment of postural stability in clinical or NeuroCom and iPad2 measurements were used to evaluate
field environments. how closely the real-time COG sway measured by the 2 systems
Objective: To determine whether accelerometer and gyro- tracked each other.
scope data sampled from a consumer electronics device (iPad2)
Results: The limits between the 2 devices ranged from
provide sufficient resolution of center-of-gravity (COG) move-
ments to accurately quantify postural stability in healthy young 0.58 to 0.58 in SOT condition 1 to 2.98 to 1.38 in SOT condition
people. 5. The largest absolute value of the measurement error within
Design: Controlled laboratory study. the 95% confidence intervals for all conditions was 2.98. The
Setting: Research laboratory in an academic medical mean absolute percentage error analysis indicated that the
center. iPad2 tracked NeuroCom COG with an average error ranging
Patients or Other Participants: A total of 49 healthy from 5.87% to 10.42% of the NeuroCom measurement across
individuals (age ¼ 19.5 6 3.1 years, height ¼ 167.7 6 13.2 SOT conditions.
cm, mass ¼ 68.5 6 17.5 kg). Conclusions: The iPad2 hardware provided data of suffi-
Intervention(s): Participants completed the NeuroCom cient precision and accuracy to quantify postural stability.
Sensory Organization Test (SOT) with an iPad2 affixed at the Accuracy, portability, and affordability make using the iPad2 a
sacral level. reasonable approach for assessing postural stability in clinical
Main Outcome Measure(s): Primary outcomes were equi- and field environments.
librium scores from both systems and the time series of the
angular displacement of the anteroposterior COG sway during Key Words: concussions, motor function, motor control,
each trial. A Bland-Altman assessment for agreement was used biomechanics

Key Points
 The accelerometer and gyroscope within the iPad2 provided data of sufficient quantity and quality to enable
accurate evaluation of postural stability.
 The accuracy, portability, availability, and affordability of mobile devices can enable health care providers in various
clinical and field settings to evaluate postural stability in athletes.
 To improve clinical outcomes, mobile devices can be a mechanism by which sophisticated biomechanical
algorithms are translated to the broader field of athletic trainers and clinical teams treating patients with concussions.
 The accuracy and reliability of mobile devices must be validated before these systems are used to assess cognitive
or motor function.

M
aintenance of stable posture depends on the or mild traumatic brain injury (mTBI), has been well-
efficient processing and integration of informa- documented to adversely affect postural stability; however,
tion from the visual, somatosensory, and vestib- debate exists about the time course for resolution of balance
ular systems and the modulation of efferent responses by declines postconcussion.11–21 Often after concussion, static
the musculoskeletal system.1 A decline in postural stability postural-stability declines are most evident when visual and
is often a hallmark of advancing age2–4 and neurologic support-surface conditions are altered.20 Based on the
diseases, such as Parkinson disease5–7 and multiple frequency of postural-stability deficits postconcussion, the
sclerosis.8–10 In addition to neurologic disease, concussion, recent consensus statement on concussion in sport22 and the

578 Volume 50  Number 6  June 2015


National Athletic Trainers’ Association23 (NATA) recom- Table 1. Demographic Information for Participants (Mean 6 SD)
mended that balance assessment be considered part of Participants No. Age, y Height, cm Mass, kg
baseline testing for athletes and that assessment postcon- Total 49 19.5 6 3.1 167.7 6 13.2 68.5 6 17.5
cussion is a ‘‘reliable and valid addition’’ to a multifaceted Male 22 18.5 6 3.1 173.8 6 13.5 82.4 6 16.0
approach to concussion management. Female 27 18.4 6 3.1 161.6 6 10.7 57.6 6 8.5
Current methods for examining postural stability range
from sophisticated biomechanical techniques to subjective
clinical assessments.14,24–27 Biomechanics-focused meth- questioned the interrater and intrarater reliability of its
ods, which include force plate and 3-dimensional motion- scoring method39,40 and have noted floor and ceiling scoring
capture systems, provide the greatest reliability and effects that may limit clinical utility.26,27,41 These reliability
accuracy in assessing balance.28 The Sensory Organization concerns may be exacerbated in environments where
Test (SOT; NeuroCom Smart Balance Master; NeuroCom multiple providers (eg, certified athletic trainers, physi-
International Inc, Clackamas, OR), which is not a cians, and physical therapists) work together to diagnose
traditional biomechanical assessment, uses aspects of and treat concussed athletes and make return-to-play
biomechanical techniques through a force-plate–based decisions. Recent technological advances and the inclusion
posturography system that measures center-of-pressure of inertial-measurement units (ie, accelerometer, gyro-

Downloaded from http://meridian.allenpress.com/jat/article-pdf/50/6/578/1617372/1062-6050-50_2_01.pdf by guest on 23 August 2020


(COP) movements while systematically manipulating scope) in mobile devices may provide a readily available
visual, somatosensory, and vestibular information. Clini- and affordable solution to augment subjective clinical
cally, the SOT has been shown to be sensitive to functional assessments of postural stability with objective and
deficiencies in the visual, vestibular, and somatosensory quantitative measures.
systems often seen after concussion or mTBI29,30 and, in Therefore, the purpose of our study was to determine
turn, has been used to track recovery from concussion and whether postural stability could be quantified accurately
evaluate the effectiveness of rehabilitation.31,32 The primary with data gathered by the embedded accelerometer and
outcome of the SOT is the equilibrium score. Assuming a gyroscope of the iPad2 (Apple Inc, Cupertino, CA). We
maximum of 12.58 of anteroposterior (AP) sway, the compared AP center-of-gravity (COG) sway derived from
equilibrium score is calculated by subtracting the observed iPad2 sensor data with output from the NeuroCom SOT for
peak-to-peak sway range from this value and dividing the amplitude (equilibrium scores) and real-time displacement
difference by 12.5. Scores range from 0 to 100, with 100 goodness of fit (mean absolute percentage error [MAPE])
representing 08 of AP sway range and 0 representing 12.58 during performance of the SOT. The identification of an
or more of AP sway. Despite its sensitivity and precision, accurate method of assessing postural stability with
the SOT is limited as a clinical or field evaluation tool by its affordable and portable consumer electronics devices
expense, size, need for trained operators, and lack of would effectively fill the fundamental gap between
portability.26,33 inexpensive, subjective clinical tests and more expensive
A cost-effective and space-effective alternative to biomechanical measurement techniques and would provide
systems such as the NeuroCom is attaching inertial sensors a mechanism to improve continuity of assessment and care
(eg, accelerometer, gyroscope) to the body to measure across multiple providers.
linear and angular kinematics. Whitney et al26 validated
accelerometry methods using measures of planar acceler- METHODS
ation of the pelvis and reported a correlation with the sway
metric of the SOT. However, their methods were weakened Participants
by the postprocessing synchronization of the data, which
A total of 49 healthy participants met the inclusion
aligned data from the 2 devices according to optimized
criteria: (1) age from 14 to 25 years, (2) no history of
correlation values rather than via real-time synchronization.
concussion in the 6 months before the study, (3) no known
Other researchers26,33–37 have explored the efficacy of
musculoskeletal or neurologic condition resulting in
accelerometry-based balance measures; however, no ap-
impaired balance or postural stability, and (4) complete
proach has combined the use of an accelerometer and a
SOT and iPad2 data sets. Written informed consent was
gyroscope in a commercially available, nondedicated obtained from the parent or legal guardian of all minors
device package and then evaluated its effectiveness in before participation, and the Cleveland Clinic Institutional
assessing postural stability relative to an accepted clinical Review Board approved this project. Participant character-
system, such as the NeuroCom. The recent inclusion of istics are provided in Table 1.
relatively sophisticated inertial-measurement technologies
in consumer electronics devices, such as smartphones and
tablet-based computing devices, provides an opportunity to Data Collection
use these devices to objectively assess postural stability in Participants completed the 6-condition SOT14 on the
athletes during healthy baseline testing, at diagnosis of NeuroCom while wearing a custom-built belt securely
concussion, during the return-to-play process, and when holding the iPad2 at approximately sacral height with the
determining resolution of concussion symptoms. screen of the device facing away from the body (Figure
The most common clinical test to assess postural stability 1A). This placement of the iPad2 positioned its sensors as
in athletes is the Balance Error Scoring System (BESS).14 near as possible to the approximate center of mass (COM)
The complete BESS consists of 6 conditions comprising 3 during upright stance.26 The iPad2 contains motherboard-
stances performed on firm and foam surfaces with eyes level embedded inertial sensors. Linear acceleration of the
closed.14 Whereas the BESS is considered a reliable and device was captured with the embedded 3-axis linear
valid assessment of postural stability,38 researchers have accelerometer (model LIS331DLH; STMicroelectronics,

Journal of Athletic Training 579


Downloaded from http://meridian.allenpress.com/jat/article-pdf/50/6/578/1617372/1062-6050-50_2_01.pdf by guest on 23 August 2020
Figure 1. A and B, Experimental setup. Abbreviations: t1, time 1; t2, time 2.

Geneva, Switzerland), which has a range of 62.0g, computer; only 1 of 882 trials was discarded due to loss
resolution of 0.9 to 1.1 mg, and maximum sampling rate of more than 3 consecutive data points. The median number
of 100 Hz.42 Device-rotation rates were measured by the of lost individual data points was 1 per trial (0.05% of total
embedded 3-axis gyroscope (model L3G4200D; STMi- data points). The NeuroCom force plate measures COP in
croelectronics) with a range of 62508/s, resolution of 8.75 the 2-dimensional plane associated with AP and ML sway.
mdeg/s, and maximum sampling rate of 100 Hz.43 The The ‘‘On’’ signal from the NeuroCom indicating the
Sensor Data iOS application (Sensor Data app by Wave- initiation of each trial was collected and processed through
front Labs; Apple Inc) was used to collect and transmit an analog-to-digital converter connected to the laptop
sensor data from the embedded accelerometer and gyro- computer and was used to mark the start of the 20-second
scope in the x-, y-, and z-directions to a laptop computer trial (Figure 1B). The COP data from the NeuroCom were
(Apple Inc). The coordinate system relative to the device is synchronized with iPad2 data in real time using the same
shown in Figure 1A. We were particularly interested in the LabView data-collection program. Individual trials were
x- and z-axes because they are in the direction of the eliminated if any of the following criteria were met: (1) the
mediolateral (ML) and AP movements, respectively, participant committed a testing error that normally would
associated with postural maintenance. The iPad2 data were invalidate the SOT, such as moving the feet midtrial or
sampled at 100 Hz, the same frequency as that of the falling (8 trials); (2) poor data integrity (1 trial); or (3) data-
NeuroCom system. transcription error, such that files with different data sets
Data were transmitted from the iPad2 with a user- were mistakenly labeled with the same name (18 trials).
datagram protocol over a WiFi connection to another The remaining 855 trials, representing 97% of collected
device. Before data collection, the WiFi connection was trials, were used in the analysis.
established between the iPad2 and a laptop computer, and
data were logged and saved continuously on the laptop
Data Analysis
computer during the trial using a customized LabView
data-collection program (National Instruments Corp, Aus- The AP COG angle was used for all outcome metrics.
tin, TX). This experimental setup minimized the latency of The NeuroCom system measures AP COP for position on
data transmission between the iPad2 and the laptop the force plate at each time point in a trial, converts it to AP

580 Volume 50  Number 6  June 2015


COG angular sway via a trigonometric relation incorporat- Table 2. Equilibrium Scores From Our Study Compared With
ing the height of the participant, and calculates the Scores Reported by Wrisley et al46
equilibrium score metric using Equation 1: Mean 6 SD Equilibrium Score
Sensory Wrisley et al46 Our Study
Equilibrium Score
h    i Organization Test (N ¼ 13)a (N ¼ 49)b P Value
12:58  maximum SwayAP ðhÞ minimum SwayAP ðhÞ Condition
¼ 1 95.3 6 1.6 95.3 6 1.9 ..99
12:58 2 93.6 6 2.2 93.0 6 2.4 .40
*100%; ð1Þ 3 91.6 6 4.0 91.4 6 4.1 .87
4 87.3 6 6.0 89.8 6 6.5 .22
where SwayAP is AP sway. 5 74.6 6 3.6 71.0 6 10.9 .25
To generate an AP COG sway metric from the iPad2, we 6 72.9 6 7.1 78.3 6 10.3 .08
used a mathematical model to combine the accelerometer
Composite score 83.4 6 3.0 84.0 6 3.1 .54
and gyroscope data to best predict the COG sway metric
a
from the NeuroCom system. Specifically, accelerometer The mean 6 SD age of participants was 24 6 4 y.
b
and gyroscope data (recorded in ms1s1 and rad/s, The mean 6 SD age of participants was 19.5 6 3.1 y.

Downloaded from http://meridian.allenpress.com/jat/article-pdf/50/6/578/1617372/1062-6050-50_2_01.pdf by guest on 23 August 2020


respectively) were initially filtered using a fourth-order,
low-pass Butterworth filter with a cutoff frequency of 1.25 NeuroCom systems tracked each other on a sample-by-
Hz. To account for initial orientation of the device on the sample basis. The filtered AP COG sway angle of the
patient’s lower back, accelerometer fields were offset by NeuroCom and the modeled AP COG sway of the iPad2
the mean of the first 10 samples (0.1 seconds) of the trial. were used in the analysis. The MAPE values range from 0
Rotation-rate data were integrated once to provide to infinity, with larger values representing greater error.44
rotational displacement in degrees. Using a nonlinear For each COG sway data point in a trial, the absolute
mixed-effects model, iPad2 sensor data were fit to the difference or error between the NeuroCom and the iPad2
COG sway-angle output from the NeuroCom. Before COG sway metric was divided by the measured value
fitting, we filtered the COG sway-angle data from the (NeuroCom) and multiplied by 100. The MAPE value then
NeuroCom system with the same low-pass filter used on the was calculated as the mean value of this metric across all
iPad2 data. The resulting function, a 5-knot restricted cubic samples within a trial. We collapsed the MAPE values
spline and a sine function, was the mathematical model for across trials within a condition and participant. In addition
predicting COG movement in the AP plane using the iPad2 to the MAPE, the true error between the NeuroCom and
inertial sensor data. iPad2 COG sway angle was calculated across all samples in
The accuracy of the modeled COG sway angle from the each trial. The median error across all samples is reported.
iPad2 data was evaluated in 2 ways. First, the maximum All offline analyses were performed using custom scripts in
and minimum predicted values were compared between the MATLAB (The MathWorks, Inc, Natick, MA).
2 devices using the SOT equilibrium scores. The Neuro- Sex and age differences were evaluated on the averaged
Com equation used for the equilibrium score calculation is SOT equilibrium score in each condition using linear
shown in Equation 2: random models for the iPad2 and NeuroCom and R
software (R Project for Statistical Computing; Institute
COG AP swayðdegÞ
  for Statistics and Mathematics of Wirtschafts Universitat
COG AP position ðcmÞ 1808 Wien Vienna University of Economics and Business,
¼ arctan * ð2Þ Vienna, Austria). The a level was set at .05.
:55*participant’s height ðcmÞ p
For comparison, the equilibrium scores from the Neuro- RESULTS
Com system were exported directly, and the modeled COG
sway data from the iPad2 sensors were used in Equation 2 Sex and age were not predictors of the equilibrium score
to calculate the iPad2 equilibrium scores. Equilibrium in any SOT condition (SOT-1 through SOT-6; P . .05 for
scores from the NeuroCom and the iPad2 sensors were both predictors in all conditions) for either the iPad2 or
evaluated for goodness of fit via Bland-Altman plots, NeuroCom systems. Therefore, we collapsed data across
treating each trial as a separate observation. In this age and sex for all other analyses. The equilibrium scores
approach, the differences between the equilibrium scores calculated by the NeuroCom for our sample were not
calculated by the NeuroCom and iPad2 were subtracted different from data reported previously (Table 2).45,46
from each other (ie, NeuroCom  iPad2) to quantify the Grouped equilibrium scores (mean 6 SD) from the
measurement error and were plotted against the average of NeuroCom and iPad2 are provided for all trials in Figure
the equilibrium scores from the 2 devices, which repre- 2A and B. Bland-Altman plots revealed that the mean
sented the best approximation of the ‘‘true’’ equilibrium difference (bias) in equilibrium scores for each condition
score. The mean and standard deviation (SD) of the was close to 0, with SOT-1 showing the smallest (0.01%)
differences then were determined. The limits of agreement and SOT-5 showing the largest (6.2%) mean difference
were defined as mean 6 2 SD and represent the limits (Figure 2C through H). For all other conditions except
within which one can be 95% confident the measurement SOT-5, the mean difference was equal to or less than 2.6%
error resides. The mean of the difference and the respective (Table 3). Using Equation 1, even the largest mean
limits of agreement are reported for each SOT condition. difference in equilibrium scores between the 2 devices
We used the MAPE to assess how well the time-series (6.2% in SOT-5) represented a small discrepancy in actual
data of the AP COG sway-angle data from the iPad2 and sway angle (ie, less than 18).

Journal of Athletic Training 581


Downloaded from http://meridian.allenpress.com/jat/article-pdf/50/6/578/1617372/1062-6050-50_2_01.pdf by guest on 23 August 2020
Figure 2. A, Equilibrium scores (mean and standard deviation values) from NeuroCom Sensory Organization Test (SOT) and calculated
from iPad2 (Apple Inc, Cupertino, CA) sensor data are shown for the 6 SOT conditions and composite score. B, The 6 SOT conditions are
illustrated (adapted with permission from NeuroCom International Inc, Clackamas, OR). C–H, Bland-Altman plots for each SOT condition,
where the difference in equilibrium scores (NeuroCom  iPad2) is plotted against the average 2 equilibrium scores on a trial-by-trial basis.
The solid lines represent the mean difference in equilibrium scores, where a score closer to 0 indicates more similarity in the values from
the 2 devices. The dashed lines represent the upper and lower limits of agreement (mean difference 6 1.96 SD of the difference in
equilibrium scores), where a smaller gap between the dashed lines indicates less variability and more consistency across most of the
values from the 2 devices. Continued on next page.

Additional inspection of the limits of agreement in the rapid movements (Figure 3A and B) and for large
Bland-Altman plots also revealed that 95% of the measure- movements (Figure 3C and D) in the time domain while
ment error between the 2 devices was from 3.8% to 4.0% or also capturing the extreme values within each trial. The
from 0.58 to 0.58 in the condition of best agreement (SOT-1) MAPE metric, which was used to quantify the goodness of
and from 23.1% to 10.7% or from 2.98 to 1.38 in the fit between the iPad2 and the NeuroCom COG time-series
condition with the poorest agreement (SOT-5; Table 3). data, was 7.93% for the entire cohort, signifying that the
Extrapolating from this result, the largest absolute value of COG value predicted with the iPad2 sensor data was similar
the measurement error within the 95% confidence intervals to the NeuroCom COG measurement across all trials.
for all conditions was 23% or 2.98. Further inspection of Furthermore, the model had a median difference of 0.018
Table 3 shows that for most conditions (SOT-1 through SOT- (first and third quartiles ¼ 0.278 and 0.248, respectively)
4) the largest absolute value of the measurement error in the between iPad2 and NeuroCom measures of COG sway.
95% confidence intervals was 11% or at most 1.48. Table 4 shows MAPE per condition for all analyzed trials.
Four representative trials with MAPE values ranging from The SOT-5 had the smallest MAPE value (5.87%),
0.68% to 4.53% are depicted in Figure 3. Visual inspection representing the lowest error and best fit between the iPad2
of Figure 3 indicates that the iPad2 measurements tracked and NeuroCom COG sway metrics despite having the largest
the NeuroCom COG measurement very closely for small, peak-to-peak movements (Table 3). The SOT-2 had the

582 Volume 50  Number 6  June 2015


Downloaded from http://meridian.allenpress.com/jat/article-pdf/50/6/578/1617372/1062-6050-50_2_01.pdf by guest on 23 August 2020

80

Figure 2. Continued from previous page.

largest MAPE value (10.42%) and, based on the equilibrium be quantified accurately in healthy adolescents and young
score metric (Table 3), the smallest sway range. adults using data from the accelerometer and gyroscope
embedded within the iPad2. The Bland-Altman analysis
DISCUSSION assessed the ability of the iPad2 to accurately capture the
peak-to-peak sway magnitude within a trial as quantified by
Comparison between the NeuroCom- and iPad2-generat- the equilibrium score. The SOT-1 had the smallest spread
ed equilibrium scores indicated that postural stability can between the 95% confidence intervals (mean bias ¼ 0.01%;

Journal of Athletic Training 583


Table 3. Results From Bland-Altman Analysis predicted iPad2 and the actual NeuroCom COG sway
Mean Difference in Equilibrium Scores angles was very close to 0 (median difference ¼ 0.018).
Between the Devices These results are timely because evidence suggests that
Limits of Agreement, % whereas the accurate characterization of AP peak-to-peak
Sensory Organization range of sway has clinical utility, the tracking and
Test Condition Mean Bias, % Lower Upper
quantification of balance reactions throughout an entire
1 0.01 3.8 4.0 trial may be a source of added sensitivity.47 In future
2 1.5 5.3 2.3 studies, including cross-correlation analyses of the real-
3 2.0 8.3 4.4 time sway data may provide even better agreement between
4 2.4 11.0 6.2
the 2 devices than what we reported.
5 6.2 23.1 10.7
6 2.6 15.6 10.4
Whereas posturography methods estimate COG sway via
ground reaction forces48 to quantify postural stability,
inertial measurements aim at characterizing sway relative
limits of agreement ¼ 3.8%, 4.0%), and the trial with the to COM. From a methodologic standpoint, the placement of
smallest mean measurement error was in this condition. the iPad2 at the sacrum reasonably approximated COM
Overall, the iPad2 consistently provided a measure similar position, which was important for an accurate character-

Downloaded from http://meridian.allenpress.com/jat/article-pdf/50/6/578/1617372/1062-6050-50_2_01.pdf by guest on 23 August 2020


to the NeuroCom system during a condition in which very ization of body sway when using accelerometry.31 The
minute balance movements were made. The largest mean iPad2 measures linear acceleration and angular rotation
measurement error was in SOT-5, which also had the independent of gravitational effects in 3-dimensional space
largest span between the 95% confidence intervals (mean via an embedded triaxial accelerometer and gyroscope.
bias ¼6.2%; limits of agreement ¼23.1%, 10.7%). Even Within traditional inertial-measurement methods, sway-
in the most difficult balance condition, the mean bias and angle measurements using linear accelerometers may not
the limits of agreement represent less than 18 and 2.98, provide the best estimate of stability when COM motion
respectively, of sway difference between the 2 devices. becomes less planar and more angular with rotation about
Overall, the Bland-Altman analysis highlighted the low the ankle joint.26 In these cases, gravitational effects
measurement error using the iPad2 when the peak-to-peak become large and affect axial-acceleration measurements.26
sway magnitude of a trial was small (ie, average Incorporation of the triaxial gyroscope allows rotation to be
equilibrium scores .80 in Figure 2C through H, all measured, which is particularly useful when sway is less
conditions). We do not believe this is a limitation of the planar and larger movements are made, as evidenced by
iPad2 sensors themselves but of the limited variability in greater correlation of equilibrium scores in the more
the dataset. In 70% of all the trials from all conditions, the challenging SOT conditions (ie, SOT-3 through SOT-6).
peak-to-peak sway magnitude was equal to or less than 28, Whereas the utility of commercially available sensor
measured as NeuroCom equilibrium scores equal to or hardware, such as that built into the iPad2, has not been
greater than 84. Given that the statistical model that investigated systematically in a clinical environment, the
generated the sway estimate of the iPad2 was based on this specifications of the sensors and methods used were
well-performing dataset, the model could predict good consistent with the sophisticated and single-purpose
performance (ie, most of the dataset) very well and had balance-assessment and movement-assessment devices
more difficulty and measurement error when estimating and techniques.26 As sensor specifications and processing
poor performance. Particularly in SOT-5 and SOT-6, when rates continue to improve with new tablet-device hardware,
test conditions were more challenging and elicited more the resolution of these techniques will become more
sway and variability in the performances, the measurement precise. In addition, consumer electronic devices, such as
errors of the peak-to-peak sway values were increased and tablets and smartphones, offer vast opportunities for
more variable. Our study indicated that data from the developers and users to create custom, nonproprietary
sensors within the iPad2 can predict time-series sway applications that can be shared across investigators to
position without sacrificing accuracy in quantifying peak- facilitate the collection of a common group of data
to-peak sway displacements for most trials in the studied elements for population-based studies. These types of
population. Once we have additional data in other devices offer the possibility of providing a complete,
populations or those after concussion, with the anticipated portable tool for collecting, processing, and analyzing
increased variability in performance, the statistical model clinical assessments of postural stability, unlike stand-alone
will adapt to better capture larger peak-to-peak sway inertial sensors. Whereas the NeuroCom SOT improves our
values. understanding of the balance declines associated with
The iPad2 sensors accurately tracked the NeuroCom real- concussion, it is not feasible or practical for use in most
time COG sway in all SOT conditions as measured by the environments in which athletic trainers practice due to
MAPE values and the median difference between the extensive cost, space, and personnel requirements.
NeuroCom and iPad2 measurements. The accuracy of the Given that we have shown the validity of using the
model was evaluated with the MAPE metric, giving a embedded sensors of the iPad2 to assess postural stability,
relative percentage value that signified the ‘‘error’’ of the we have opened an additional avenue of investigation that
predicted value from the actual value. The MAPE values involves quantification of movements in real time in more
were smallest during SOT-3 through SOT-6 despite the than the AP direction. This is possible with the 3-
much higher amplitude and volatility of balance reactions dimensional inertial sensors of the iPad2. The triaxial
by the individual. The less demanding tasks (SOT-1 and accelerometer and gyroscope allow the expansion of
SOT-2) had larger but still relatively small MAPE values. balance evaluation to a more comprehensive quantification
Furthermore, the absolute median difference between the of COM movement in 3 dimensions (AP, ML, and trunk

584 Volume 50  Number 6  June 2015


Downloaded from http://meridian.allenpress.com/jat/article-pdf/50/6/578/1617372/1062-6050-50_2_01.pdf by guest on 23 August 2020

Figure 3. Example of real-time anteroposterior center-of-gravity sway data from NeuroCom (NeuroCom Smart Balance Master;
NeuroCom International Inc, Clackamas, OR) and iPad2 (Apple Inc, Cupertino, CA), with corresponding mean absolute percentage error
values for 4 randomly sampled trials: A, Sensory Organization Test (SOT) condition 1; B, SOT condition 4; C, SOT condition 5; and D, SOT
condition 6. Positive anteroposterior center-of-gravity sway values correspond to movements in the anterior direction and negative values
correspond to movements in the posterior direction relative to the center of gravity. The mean absolute percentage error values were used
to determine the goodness of fit for the iPad2 data relative to NeuroCom data and are provided for each condition in Table 4.

Journal of Athletic Training 585


Table 4. Overall Mean Absolute Percentage Error Values Per assessment across the multidisciplinary team of providers
Condition for Center-of-Gravity Sway Time-Series Data From iPad2a involved in the care and treatment of concussion by using
and NeuroComb Systems
objective measures will provide clinicians with a more
Sensory Organization Mean Absolute Percentage precise method of evaluating motor function. This ap-
Test Condition Error 6 SD, % proach, which is affordable and scalable, can provide new
1 9.34 6 5.46 and unique information about the effects of concussion on
2 10.42 6 6.87 balance and its recovery. We recently completed a project
3 7.40 6 5.08 in which the balance module of the Cleveland Clinic
4 8.42 6 6.30 Concussion App (Cleveland Clinic, Cleveland, OH) was
5 5.87 6 4.80
used to augment the error scoring of the BESS in a group of
6 6.25 6 4.62
Overall 7.93 6 5.76
healthy and concussed athletes.49
a
Whereas the BESS is currently the preferred clinical
Apple Inc, Cupertino, CA. assessment of postural stability postconcussion, the devel-
b
NeuroCom Smart Balance Master; NeuroCom International Inc,
Clackamas, OR.
opment of a portable and sophisticated method of
quantifying postural stability opens the possibility of using
potentially more sensitive and challenging dynamic testing

Downloaded from http://meridian.allenpress.com/jat/article-pdf/50/6/578/1617372/1062-6050-50_2_01.pdf by guest on 23 August 2020


rotation). Assessment of postural control, taking all 3 protocols.11,50–53 Researchers52 using more dynamic assess-
dimensions into consideration, provides a more precise and ments of postural stability postconcussion have noted that,
objective assessment of motor functioning. Multidirectional in some athletes, declines in balance may last from 10 to 30
characterization of balance may be particularly important days postconcussion; investigators20,54 using static tests
for optimizing the clinical management of athletes with have reported that most balance declines are resolved
lingering balance dysfunction by identifying vestibular, within 3 to 5 days postconcussion. Although dynamic tests
neuromuscular, and musculoskeletal deficiencies and pro- of postural stability may be more sensitive than static tests,
viding guidance for the rehabilitative management of they are susceptible to the same, if not greater, bias as
patients. We fully acknowledge that this model of balance subjective rating scales because the rater must monitor and
assessment cannot replace the ability to systematically evaluate multiple degrees of freedom and potentially
manipulate afferent inputs to distinguish among somato- manage a more complex scoring schema and rubric. The
sensory, vestibular, and visual contributions to postural
development of mobile applications capable of objectively
stability as provided by the NeuroCom. If the ability to
quantifying postural stability during dynamic-balance tasks
distinguish among the various systems that contribute to
will provide a pathway to their use in athletic or clinical
balance is a goal for future uses of this technology, the
development and validation of protocols and additional, environments. Providers will then be better informed about
more inexpensive equipment to create those conditions is the true postural-stability capabilities or impairments of
necessary. athletes suspected to have concussions and better able to
Based on strong agreement between equilibrium scores assess and treat athletes with lingering balance impair-
from both devices across SOT test conditions and low error ments.
in predicting time-series sway position as evaluated by the A potential limitation of this study is that the sample
MAPE metric, this approach successfully provided a viable included only healthy individuals rather than populations
measure of postural stability in a cohort of healthy young with neurologic conditions or concussions. Although we
adults that could be used to assess baseline levels of considered including these populations, our fundamental
postural stability as part of a preseason evaluation, as goal was to determine whether this tablet device provided
recommended in the recent position statement on concus- data that could be used to accurately quantify postural
sion from the NATA.23 In an early publication, Guskiewicz stability. Including a population with neurologic conditions
et al20 correctly noted that most clinicians did not have or concussion would have added more variance to the data,
access to biomechanical laboratories or postural-stability which would have compromised our ability to systemati-
systems for the proper assessment of how concussion may cally address this question. Given that this approach has
be affecting balance. Thus, in the absence of balance been validated, we are conducting large-scale, population-
testing, they recommended a conservative strategy for based studies in athletes with concussions (eg, baseline
returning athletes to participation. A lack of postural- testing and immediate postconcussion through return to
stability measures led to the development of the BESS for play) and populations with neurologic conditions (eg,
the field and clinical evaluation of balance during preseason Parkinson disease, multiple sclerosis). Furthermore, we
baseline testing, postconcussion, and when tracking recov- are evaluating the use of the balance module of the
ery. Unfortunately, although many advances have been Cleveland Clinic Concussion App to estimate postural
made in the assessment of various aspects of cognitive stability during more dynamic-stability tasks (eg, tandem
functioning and better classification of concussion-related gait) and dual-task paradigms (eg, balance plus cognitive
symptoms, the objective quantification of balance remains tasks) in healthy and concussed athletes. These studies
largely subjective via BESS testing for most certified represent the next phase to improve our understanding of
athletic trainers, physicians, and physical therapists. The how concussion affects dynamic postural stability, dual-
accurate and reliable assessment of postural stability in a task performance, and the recovery of these functions
cohort of healthy young adults, similar to the population postconcussion in a large population. In these projects, we
most often experiencing sport-related concussion, poten- are using measures of sway in the AP, ML, and trunk-
tially can dramatically improve the treatment and manage- rotation planes to provide 2- and 3-dimensional character-
ment of concussion. Minimizing the subjectivity of izations of postural stability.

586 Volume 50  Number 6  June 2015


A possible limitation from an analytical perspective is the REFERENCES
use of average peak-to-peak sway within a trial. The model 1. Horak FB. Postural orientation and equilibrium: what do we need to
relating the iPad2 and NeuroCom data sets was based on know about neural control of balance to prevent falls? Age Ageing.
the range of values that was recorded in this cohort; overall 2006;35(suppl 2):ii7–ii11.
minimum and maximum sway angles in this group were 2. Melzer I, Benjuya N, Kaplanski J. Postural stability in the elderly: a
7.88 and 10.28, respectively. It is unclear how well the comparison between fallers and non-fallers. Age Ageing. 2004;33(6):
model would perform on trials with data outside of these 602–607.
values. However, this limited range is not a major concern 3. Menz HB, Lord SR, Fitzpatrick RC. Age-related differences in
considering that the limits of stability cited by the walking stability. Age Ageing. 2003;32(2):137–142.
NeuroCom SOT are 58 and 7.58 in the posterior and 4. Teasdale N, Simoneau M. Attentional demands for postural control:
anterior directions, respectively.55 We anticipate that the effects of aging and sensory reintegration. Gait Posture. 2001;
limitations inherent to many clinical balance tests, such 14(3):203–210.
as reliability of scoring and ceiling and floor effects, can be 5. Horak FB, Dimitrova D, Nutt JG. Direction-specific postural
instability in subjects with Parkinson’s disease. Exp Neurol. 2005;
mitigated or eliminated by testing individuals while they
193(2):504–521.
perform various tasks or stances.
6. Jankovic J, McDermott M, Carter J, et al. Variable expression of

Downloaded from http://meridian.allenpress.com/jat/article-pdf/50/6/578/1617372/1062-6050-50_2_01.pdf by guest on 23 August 2020


Parkinson’s disease: a base-line analysis of the DATATOP cohort.
CONCLUSIONS The Parkinson Study Group. Neurology. 1990;40(10):1529–1534.
To our knowledge, we are the first to evaluate the 7. Morris M, Iansek R, Smithson F, Huxham F. Postural instability in
accuracy of a mobile device against a clinically accepted Parkinson’s disease: a comparison with and without a concurrent
task. Gait Posture. 2000;12(3):205–216.
method of assessing postural stability under a range of
8. Frzovic D, Morris ME, Vowels L. Clinical tests of standing balance:
conditions. Overall, the results presented here indicate that
performance of persons with multiple sclerosis. Arch Phys Med
the accelerometer and gyroscope within the iPad2 provided Rehabil. 2000;81(2):215–221.
data of sufficient quantity and quality to enable accurate 9. Martin CL, Phillips BA, Kilpatrick TJ, et al. Gait and balance
evaluation of postural stability. Mobile devices, which have impairment in early multiple sclerosis in the absence of clinical
continued to decline in cost and increase in availability and disability. Mult Scler. 2006;12(5):620–628.
hardware capability, are ideally suited to rapidly enable the 10. Morris ME, Cantwell C, Vowels L, Dodd K. Changes in gait and
field of providers associated with concussion management fatigue from morning to afternoon in people with multiple sclerosis. J
(eg, athletic trainers, physicians, physical therapists) to Neurol Neurosurg Psychiatry. 2002;72(3):361–365.
meet the recommendations set out in the NATA’s recent 11. Chou LS, Kaufman KR, Walker-Rabatin AE, Brey RH, Basford JR.
position statement on the management of sport concus- Dynamic instability during obstacle crossing following traumatic
sion.23 As mentioned in this position statement, assessment brain injury. Gait Posture. 2004;20(3):245–254.
of motor control is an integral part of the concussion 12. Geurts AC, Ribbers GM, Knoop JA, van Limbeek J. Identification of
baseline and postinjury examination. To understand motor- static and dynamic postural instability following traumatic brain
control processes, objective and quantitative methods using injury. Arch Phys Med Rehabil. 1996;77(7):639–644.
13. Guskiewicz KM. Assessment of postural stability following sport-
biomechanical principles to characterize movement are
related concussion. Curr Sports Med Rep. 2003;2(1):24–30.
necessary. Mobile devices equipped with inertial-measure-
14. Guskiewicz KM, Ross SE, Marshall SW. Postural stability and
ment hardware provide an opportunity to transition these neuropsychological deficits after concussion in collegiate athletes. J
devices from expensive electronic notebooks into data- Athl Train. 2001;36(3):263–273.
collection devices that can aid in understanding motor- 15. Thompson J, Sebastianelli W, Slobounov S. EEG and postural
control processes. The widespread availability of mobile correlates of mild traumatic brain injury in athletes. Neurosci Lett.
devices provides individuals practicing in rural or under- 2005;377(3):158–163.
served locations with minimal resources an opportunity to 16. Cavanaugh JT, Guskiewicz KM, Giuliani C, Marshall S, Mercer V,
evaluate athletes in an objective and quantitative manner Stergiou N. Detecting altered postural control after cerebral
that previously was available only to individuals in large concussion in athletes with normal postural stability. Br J Sports
academic or medical environments with sophisticated and Med. 2005;39(11):805–811.
expensive biomechanical equipment. We are not advocat- 17. Cavanaugh JT, Guskiewicz KM, Giuliani C, Marshall S, Mercer VS,
ing that mobile devices replace biomechanical analyses, but Stergiou N. Recovery of postural control after cerebral concussion:
they can be the mechanism by which sophisticated new insights using approximate entropy. J Athl Train. 2006;41(3):
biomechanical algorithms could be translated to the broader 305–313.
field of athletic trainers and clinical teams treating 18. Cavanaugh JT, Guskiewicz KM, Stergiou N. A nonlinear dynamic
approach for evaluating postural control: new directions for the
concussion to improve clinical outcomes.
management of sport-related cerebral concussion. Sports Med. 2005;
35(11):935–950.
ACKNOWLEDGMENTS 19. Fox ZG, Mihalik JP, Blackburn JT, Battaglini CL, Guskiewicz KM.
This study was supported by the Lincy Foundation and Edward Return of postural control to baseline after anaerobic and aerobic
F. and Barbara S. Bell Family Endowed Chair (Dr Alberts). We exercise protocols. J Athl Train. 2008;43(5):456–463.
thank Patrick Cummings, PTA, ATC, for participating in initial 20. Guskiewicz KM, Perrin DH, Gansneder BM. Effect of mild head
discussions about the clinical utility of this approach and Sarah injury on postural stability in athletes. J Athl Train. 1996;31(4):300–
Ozinga and Emma Phillips for assisting in data collection. 306.
We have filed an invention disclosure form to protect the 21. Collie A, Darby D, Maruff P. Computerised cognitive assessment of
intellectual property that is associated with the metric used in this athletes with sports related head injury. Br J Sports Med. 2001;35(5):
study. 297–302.

Journal of Athletic Training 587


22. McCrory P, Meeuwisse WH, Aubry M, et al. Consensus statement on 38. McCrea M, Guskiewicz KM, Marshall SW, et al. Acute effects and
concussion in sport: the 4th International Conference on Concussion recovery time following concussion in collegiate football players: the
in Sport, Zurich, November 2012. J Athl Train. 2013;48(4):554–575. NCAA Concussion Study. JAMA. 2003;290(19):2556–2563.
23. Broglio SP, Cantu RC, Gioia GA, et al. National Athletic Trainers’ 39. Finnoff JT, Peterson VJ, Hollman JH, Smith J. Intrarater and
Association position statement: management of sport concussion. J interrater reliability of the Balance Error Scoring System (BESS).
Athl Train. 2014;49(2):245–265. PM R. 2009;1(1):50–54.
24. Riemann BL, Guskiewicz KM, Shields EW. Relationship between 40. Hunt TN, Ferrara MS, Bornstein RA, Baumgartner TA. The
clinical and forceplate measures of postural stability. J Sport Rehabil. reliability of the modified Balance Error Scoring System. Clin J
1999;8(2):71–82. Sport Med. 2009;19(6):471–475.
25. Boulgarides LK, McGinty SM, Willett JA, Barnes CW. Use of 41. Pardasaney PK, Latham NK, Jette AM, et al. Sensitivity to change
clinical and impairment-based tests to predict falls by community- and responsiveness of four balance measures for community-
dwelling older adults. Phys Ther. 2003;83(4):328–339. dwelling older adults. Phys Ther. 2012;92(3):388–397.
26. Whitney SL, Roche JL, Marchetti GF, et al. A comparison of 42. LIS331DLH: MEMS digital output motion sensor ultra low-power
accelerometry and center of pressure measures during computerized high performance 3-axes ‘‘nano’’ accelerometer. STMicroelectronics
dynamic posturography: a measure of balance. Gait Posture. 2011; Web site. http://www.st.com/web/catalog/sense_power/FM89/
33(4):594–599. SC444/PF218132. Accessed June 9, 2014.
27. Blum L, Korner-Bitensky N. Usefulness of the Berg Balance Scale in 43. L3G4200D: MEMS motion sensor: ultra-stable three-axis digital

Downloaded from http://meridian.allenpress.com/jat/article-pdf/50/6/578/1617372/1062-6050-50_2_01.pdf by guest on 23 August 2020


stroke rehabilitation: a systematic review. Phys Ther. 2008;88(5): output gyroscope. STMicroelectronics Web site. http://www.st.com/
559–566. web/catalog/sense_power/FM89/SC1288/PF250373. Accessed June
28. Tarantola J, Nardone A, Tacchini E, Schieppati M. Human stance 9, 2014.
stability improves with the repetition of the task: effect of foot 44. Hamilton JD. Time Series Analysis. Princeton, NJ: Princeton
position and visual condition. Neurosci Lett. 1997;228(2):75–78. University Press; 1994:435–453.
29. Caeyenberghs K, Leemans A, Geurts M, et al. Brain-behavior 45. Borah D, Wadhwa S, Singh U, Yadav SL, Bhattacharjee M, Sindhu
relationships in young traumatic brain injury patients: DTI metrics V. Age related changes in postural stability. Indian J Physiol
are highly correlated with postural control. Hum Brain Mapp. 2010; Pharmacol. 2007;51(4):395–404.
31(7):992–1002. 46. Wrisley DM, Stephens MJ, Mosley S, Wojnowski A, Duffy J,
30. Sosnoff JJ, Broglio SP, Shin S, Ferrara MS. Previous mild traumatic Burkard R. Learning effects of repetitive administrations of the
brain injury and postural-control dynamics. J Athl Train. 2011;46(1): sensory organization test in healthy young adults. Arch Phys Med
85–91. Rehabil. 2007;88(8):1049–1054.
31. Whitney SL, Marchetti GF, Schade AI. The relationship between 47. Horak FB, Frank J, Nutt J. Effects of dopamine on postural control in
falls history and computerized dynamic posturography in persons parkinsonian subjects: scaling, set, and tone. J Neurophysiol. 1996;
with balance and vestibular disorders. Arch Phys Med Rehabil. 2006; 75(6):2380–2396.
87(3):402–407. 48. Winter DA. Human balance and posture control during standing and
32. Alsalaheen BA, Mucha A, Morris LO, et al. Vestibular rehabilitation walking. Gait Posture. 1995;3(4):193–214.
for dizziness and balance disorders after concussion. J Neurol Phys 49. Alberts JL, Thota A, Hirsch J, et al. Quantification of the balance
Ther. 2010;34(2):87–93. error scoring system with mobile technology. Med Sci Sports Exer. In
33. O’Sullivan M, Blake C, Cunningham C, Boyle G, Finucane C. press. DOI: 10.1249/MSS.0000000000000656.
Correlation of accelerometry with clinical balance tests in older 50. Catena RD, van Donkelaar P, Chou LS. Cognitive task effects on gait
fallers and non-fallers. Age Ageing. 2009;38(3):308–313. stability following concussion. Exp Brain Res. 2007;176(1):23–31.
34. Adlerton AK, Moritz U, Moe-Nilssen R. Forceplate and accelerom- 51. Parker TM, Osternig LR, van Donkelaar P, Chou LS. Gait stability
eter measures for evaluating the effect of muscle fatigue on postural following concussion. Med Sci Sports Exerc. 2006;38(6):1032–1040.
control during one-legged stance. Physiother Res Int. 2003;8(4):187– 52. Slobounov S, Slobounov E, Newell K. Application of virtual reality
199. graphics in assessment of concussion. Cyberpsychol Behav. 2006;
35. Mancini M, Horak FB, Zampieri C, Carlson-Kuhta P, Nutt JG, Chiari 9(2):188–191.
L. Trunk accelerometry reveals postural instability in untreated 53. Buckley TA, Munkasy BA, Tapia-Lovler TG, Wikstrom EA. Altered
Parkinson’s disease. Parkinsonism Relat Disord. 2011;17(7):557– gait termination strategies following a concussion. Gait Posture.
562. 2013;38(3):549–551.
36. Martinez-Ramirez A, Lecumberri P, Gomez M, Rodriguez-Manas L, 54. McCrea M, Guskiewicz K, Randolph C, et al. Incidence, clinical
Garcia FJ, Izquierdo M. Frailty assessment based on wavelet analysis course, and predictors of prolonged recovery time following sport-
during quiet standing balance test. J Biomech. 2011;44(12):2213– related concussion in high school and college athletes. J Int
2220. Neuropsychol Soc. 2013;19(1):22–33.
37. Mathie MJ, Coster AC, Lovell NH, Celler BG, Lord SR, Tiedemann 55. Nashner LM. Computerized dynamic posturography. In: Jacobson
A. A pilot study of long-term monitoring of human movements in the GP, Newman CW, Kartuch JM, eds. Handbook of Balance Function
home using accelerometry. J Telemed Telecare. 2004;10(3):144–151. Testing. St Louis, MO: Mosby Year Book; 1993:280–305.

Address correspondence to Jay L. Alberts, PhD, Cleveland Clinic Concussion Center, Cleveland Clinic, 9500 Euclid Avenue,
Cleveland, OH 44195. Address e-mail to albertj@ccf.org.

588 Volume 50  Number 6  June 2015

You might also like