ArticlePDF Available

A Novel EEG Based Spectral Analysis of Persistent Brain Function Alteration in Athletes with Concussion History

Authors:

Abstract and Figures

The neurocognitive sequelae of a sport-related concussion and its management are poorly defined. Detecting deficits are vital in making a decision about the treatment plan as it can persist one year or more following a brain injury. The reliability of traditional cognitive assessment tools is debatable, and thus attention has turned to assessments based on electroencephalogram (EEG) to evaluate subtle post-concussive alterations. In this study, we calculated neurocognitive deficits combining EEG analysis with three standard post-concussive assessment tools. Data were collected for all testing modalities from 21 adolescent athletes (seven concussive and fourteen healthy) in three different trials. For EEG assessment, along with linear frequency-based features, we introduced a set of time-frequency (Hjorth Parameters) and nonlinear features (approximate entropy and Hurst exponent) for the first time to explore post-concussive deficits. Besides traditional frequency-band analysis, we also presented a new individual frequency-based approach for EEG assessment. While EEG analysis exhibited significant discrepancies between the groups, none of the cognitive assessment resulted in significant deficits. Therefore, the evidence from the study highlights that our proposed EEG analysis and markers are more efficient at deciphering post-concussion residual neurocognitive deficits and thus has a potential clinical utility of proper concussion assessment and management.
This content is subject to copyright. Terms and conditions apply.
1
Scientific REPORTS | 7: 17221 | DOI:10.1038/s41598-017-17414-x
www.nature.com/scientificreports
A Novel EEG Based Spectral
Analysis of Persistent Brain
Function Alteration in Athletes with
Concussion History
Tamanna T. K. Munia1, Ali Haider
1, Charles Schneider1, Mark Romanick2 & Reza Fazel-Rezai1
The neurocognitive sequelae of a sport-related concussion and its management are poorly dened.
Detecting decits are vital in making a decision about the treatment plan as it can persist one year or
more following a brain injury. The reliability of traditional cognitive assessment tools is debatable, and
thus attention has turned to assessments based on electroencephalogram (EEG) to evaluate subtle
post-concussive alterations. In this study, we calculated neurocognitive decits combining EEG analysis
with three standard post-concussive assessment tools. Data were collected for all testing modalities
from 21 adolescent athletes (seven concussive and fourteen healthy) in three dierent trials. For EEG
assessment, along with linear frequency-based features, we introduced a set of time-frequency (Hjorth
Parameters) and nonlinear features (approximate entropy and Hurst exponent) for the rst time to
explore post-concussive decits. Besides traditional frequency-band analysis, we also presented a new
individual frequency-based approach for EEG assessment. While EEG analysis exhibited signicant
discrepancies between the groups, none of the cognitive assessment resulted in signicant decits.
Therefore, the evidence from the study highlights that our proposed EEG analysis and markers are more
ecient at deciphering post-concussion residual neurocognitive decits and thus has a potential clinical
utility of proper concussion assessment and management.
A concussion is a complex pathophysiological procedure which is induced by sudden impulsive biomechan-
ical forces aecting the brain1. In the US alone, sport and physical activity cause nearly 4 million concussions
each year2,3. It is critical to assess concussion and mild traumatic brain injury (mTBI) with high accuracy to
avoid anxiety, sensitivity and cognitive biases which appear as post-concussion syndrome. Moreover, insucient
follow-up and treatment can put the post-concussive person at the risk of neurobiological depression with anxiety
resulting in a longer concussion recovery time. erefore, proper understanding and measuring of concussions
are essential to treat the psychological factors as a means of eective prevention which, in turn, can lead to a
rapid post-concussion recovery period. When examining performance metrics related to motor control, it is
well established that individuals diagnosed with the post-concussion syndrome can show marked impairments
in reaction times4, visual motor processing5, gait stability6, postural balance7 and dynamic gait analysis8,9. More
importantly, it is a primary concern for both amateur and professional athletes. Because the symptoms of con-
cussions sometimes go unnoticed or are self-reported and tend to subside within 1–2 weeks10, many athletes fail
to seek immediate and proper medical care. Furthermore, high school athletes tend to purposely avoid reporting
their concussions in order to prevent being “benched” during subsequent games11. ough almost all recreational
participants express their concern about post-concussion syndrome, most competitive athletes keep quiet about
their minor physical discomforts or even deny considerable pain for the sake of pursuing their career goals.
Although athletes’ willingness of accepting risks greatly varies with the competition stages, game completion
levels and types of sports, it’s more likely that many individuals will choose to continue to play with a concussion
rather than remove themselves from competition12. However, such a decision can pose a risk to their health with
the potential for repeated head trauma13. Athletes have been shown to suer from cognitive decits up to three
1Department of Electrical Engineering, University of North Dakota, Grand Forks, 58202, USA. 2Department of
Physical Therapy, University of North Dakota, Grand Forks, 58202, USA. Correspondence and requests for materials
should be addressed to R.F.-R. (email: reza@und.edu)
Received: 28 July 2017
Accepted: 21 November 2017
Published: xx xx xxxx
OPEN
Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/
2
Scientific REPORTS | 7: 17221 | DOI:10.1038/s41598-017-17414-x
years aer their brain injury incidents, exhibiting lower performance on select neuropsychological tasks when
compared to an age-matched non-concussed group14.
Evidently, the challenges in concussion assessment have led to the studies exploiting the sensitivity of EEG
spectral features to mild, moderate, and severe traumatic brain injury over the time span as short as 15 days to
four years post-concussion.
Researchers have accomplished the quantitative analysis of the EEG signals collected from the concussed sub-
ject to evaluate the post-concussion physical and clinical recovery. Additional studies suggest that the EEG spec-
tral prole varies with acute mTBI due to the change in the cognitive state during the resting stage15,16. In essence,
the spectral prole of EEG is also altered in acute mTBI and during any anomaly of consciousness. However,
researchers argue whether mTBI can evoke long-term variations in spectral information. Also, identication of
any long-term change is sometimes controversially attributed to psychiatric comorbidity such as posttraumatic
stress disorder (PTSD). So far, long-term neurological changes have remained indistinct. Nevertheless, many
ndings support that brain volume, and white matter can be aected by mTBI17. Likewise, the resting state activa-
tion stage can be sensitive to mTBI. Another study found that EEG measurement was able to predict the return to
play better than other measurement types18. Notably, one study examined EEG and showed that frequency infor-
mation changes for as long as six months aer the mTBI occurrence19. All these ndings underscore the fact that
the power of each frequency component of EEG can reveal signicant physiological and clinical ndings. ough
there is a necessity to examine the details of spectral patterns aer a mTBI incident, only a relatively small number
of studies compared the spectral proles just with a group of frequencies bounded by specic bands.
e goal of the current research is to look into the spectral proles as a potential measurement tool which
can expose the long-term cognitive impairment aer an analytical study of EEG signals. To test our hypothesis,
we utilized visual (King-Devick (K-D) Test), postural (Balance Error Scoring System (BESS)) and neurological
(Immediate Post-Concussion Assessment and Cognitive Testing battery (ImPACT)) tests, along with a novel
EEG spectral analysis that computes the distinguishing features from each individual component of EEG, as
well as from the set of conventional frequency bands. We also utilized novel time and nonlinear feature-based
analysis to evaluate the EEG of injured and healthy athletes that provide unique and complementary measures
of post-concussion deciencies. Herein, we report that though postural, visual and neurological tests were una-
ble to detect the decits associated with a long-term concussion history, the EEG linear and nonlinear feature
based spectral analysis, both in terms of frequency bands and individual frequencies, were sensitive to highlight
post-concussion sequelae.
Methods
Participants. e inclusion criteria for the participants were adolescents high school athletes aged 14–18
years who were actively participating in football games. Adolescent athletes were emphasized in this study since
according to CDC report, youths are at increased risk of concussion, and 65% of these concussions occur in chil-
dren between 5 to 18 years of age20. ese persons are at a larger risk for traumatic brain injury as their brains are
still young and developing, and the brain tissues are not as able to recover as rapidly as an adult brain21. e data
collection was limited to football to align with the highly broadcasted wave of concerns about the sport-related
brain trauma in National Football League (NFL) stars. Exclusion criteria for the participants included any history
of intellectual or learning disabilities, neurological or psychotic disorders, or alcohol/substance abuse.
Following the inclusion and exclusion criteria, we were able to collect data from a total of 21 male participants
who are football athletes from two high schools available in Grand Forks area. e study was performed following
the experimental protocol approved by the Institutional Review Board (IRB) of the University of North Dakota.
e data were collected in accordance with the guidelines and regulation established by the protocol. e par-
ticipation was voluntary, and the participants had the right to withdraw any time from the study. Informed and
written consent for participation was collected from the athletes and also from their parents or legal guardians.
Each participant had to complete a demographic information form with previous concussion history before data
collection.
Individuals recruited for this concussion analysis study were assigned to a particular group based on the
history of concussion. e healthy group consists of 14 subjects (Age 15.86 ± 0.67 years, Height: 1.75 ± 0.09 m,
Weight: 72.82 ± 10.03 Kg) with no history of concussion while the concussed group has 7 subjects (Age
15.97 ± 0.74 years, Height: 1.77 ± 0.09 m, Weight: 73.20 ± 12.56 Kg) who suered from one or multiple previous
concussions. e concussion was detected by the concussion management team (including athletic trainer and
team physicians) assigned by the respective schools who was present on the sideline during the athletic contest.
e concussion management team detected the concussion by following the established criteria suggested by
American Academy of Neurology Guideline for Management of Sports Concussion22 and state law of North
Dakota23.
All participants were actively participating in sport and athletes with concussion history made a complete
return to play within four weeks of injury. All athletes with a history of concussion (12 days to 15 months from
injury) reported being symptom-free at the time of testing. Control participants were teammates who had never
suered a sport or non-sport related brain injury. Concussed participants’ post-concussion status is shown in
Table1.
From each subject, the traditional assessment data and EEG signals were collected in three dierent trials
with 30-days’ time dierence between the trials. e total number of data collection trials for the healthy group
was multiplied by fourteen (total 42 trials) and for the concussed subject was three multiplied by seven (total 21
trials).
Postural Data Collection Protocol. Balance Error Scoring System (BESS). e BESS is one of the most
popular tests used to nd balance decit in concussed and fatigued athletes24. Postural stability is measured using
Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/
3
Scientific REPORTS | 7: 17221 | DOI:10.1038/s41598-017-17414-x
three stances, named double leg stance, single leg stance, and tandem stance. Each test is done on two dierent
surfaces, rst on a rm surface and then on a foam surface. During these stances, athletes’ eyes are closed, and
their hands are placed on the iliac crests and feet positions are dierent based on three distinct stances. Each of
these six subtests is performed for 20 seconds. Deviation from proper stance is referred to an error, and the total
number of errors during the subtests are counted.
Visual Data Collection Protocol. King-Devick (K-D) Test. e K-D test is a test of the visual system and is
based on measurement of the speed of rapid number naming25. e K-D test is faster than other standardized tests
like ImPACT, Military Acute Concussion Evaluation (MACE) and the sports concussion assessment tool (SCAT
3) as it takes just two minutes to complete the testing and thus is more practical in case of sideline application25.
e K-D test consists of three test cards and the athlete’s need to name the numbers from the cards rapidly with-
out any error. e score for the test is calculated by combining the amount of the three times, in seconds, required
to read the three cards. e test involves attention, rapid eye movements as well as language operation. ese
three functions may be adversely aected, resulting in a poor K-D test performance. e test purports to measure
any suboptimal brain functional decits aer a concussion incident, as well as sometimes reects decits due to
sleep deprivation, Parkinsons disease, hypoxia and multiple sclerosis. In this experiment, we performed the K-D
test to nd out the ecacy of this test to assess suboptimal brain functional decit due to a concussion aer a time
gap between concussion incident and data collection.
Neuropsychological Data Collection Protocol. Immediate Post-Concussion Assessment and Cognitive
Testing (ImPACT). e ImPACT battery is the most common computerized test that can be used in cognitive
concussion assessment26. e test battery consists of three dierent measures: Demographic data, neuropsycho-
logical tests, and the Post-Concussion Symptom Scale (PCSS). e assessment results from these three sections
are combined to assist in accurate evaluation and management of concussion27. e demographic data section
mainly consists of all the important sport, medical, and concussion history related information. For the neuropsy-
chological test sections, ImPACT (version 3.0) contains six dierent neuropsychological tests, and each of these
tests is intended to target dierent parts of cognitive functioning comprising attention, verbal and visual memory,
control, reaction time and processing speed. Combining the results from these six dierent tests, a set of com-
posite scores are produced containing separate measures named verbal memory, visual memory, motor speed,
reaction time and impulse control. e detailed description of these tests can be found at2628. e last section
named PCSS is also utilized in the ImPACT battery study28. e scale is reported by various sports organizations
to manage and track post-concussion symptoms2629. is section has a 21-symptom checklist which mainly asks
the athlete to specify a rate for each symptom on a scale of one to seven, with zero representing no presence of a
symptom and six representing a severe symptom. An ImPACT test was performed by all participants during all
three trials.
EEG Data Collection Protocol. EEG activities were measured using a 9-lead wireless B-Alert headset30.
Electrode impedance was kept below 50 k. During data collection, the le mastoid was used as a reference, and
the right mastoid was used as a ground. e sampling rate for data collection was 256 Hz, and data were acquired
by placing nine electrodes at F3, F4, Fz, C3, C4, Cz, P3, P4 and POZ locations as shown in Fig.1.
e data were collected for 5 minutes from all 21 subjects during dierent trial sessions each under three con-
ditions: vigilant task (VT), eyes open (EO), and eyes closed (EC). During VT condition, the subject was highly
engaged by choosing between a primary vs. secondary or tertiary task every 1.5 to 3 seconds. During EO con-
dition, the subject goes through a low engagement state by responding to an optical probe every 2 seconds. e
EC state creates a distraction status, and the subject has to respond to an audio tone every 2 seconds. e same
procedure was followed at all dierent trials for all subjects.
EEG Data Analysis. e EEG data were rst high pass ltered above 1 Hz and then low-pass ltered below
40 Hz, and thus a 1–40 Hz (24 dB/octave) band-pass lter was formed. e rst and last 10 s of each 5-min record-
ing during EO, EC, and VT conditions were rejected to eliminate state transitions. e EEG data were then vis-
ually inspected to determine clean EEG data and randomly occurring large amplitude with power 3 standard
deviations with respect to the mean value of the clean EEG was removed. Aerward, the stereotypical noise like
eye movements, eye blinks, muscular activity, line noise, motion related signal, and heart signals was cleaned by
Concussed
Participants Number of
concussion Loss of
consciousness Confusion Amnesia
Post-
concussion
RTP days
Days from concussion incident to data
collection
From
incident 1 From
incident 2 From
incident 3
1 2 No Yes Yes 14 263 216 —
2 1 No Yes Yes 21 118 —
3 1 No No No 7 267 —
4 3 No Yes Yes 10 462 297 162
5 2 No Yes Yes 25 92 65 —
6 1 No No No 10 127 —
7 1 No No Ye s 15 12 —
Table 1. Concussed participants demographic information.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/
4
Scientific REPORTS | 7: 17221 | DOI:10.1038/s41598-017-17414-x
using well-established Independent Component Analysis procedure of EEGLAB detailed previously31,32. Any
other nonstereotyped or residual artifact was removed through visual inspection of the raw data.
e clean EEG data was then segmented into 1-second epochs containing 256 data points. Power spectral den-
sity (PSD) was determined by computing Fast Fourier Transformations (FFT) with a 10% Hanning window on
each segment to determine spectral power (μV2) for 1 to 40 Hz frequency bins of each EEG channels. e PSD of
the individual bins were then averaged and logged to calculate PSD of conventional EEG frequency bands named
delta (1–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), beta (12–30 Hz) and gamma (30–40 Hz). Although gamma band
is considered as a pattern of neural oscillation with a frequency between 30 Hz to 100 Hz nowadays, we have used
30–40 Hz as gamma band in this study since the gamma band was previously dened up to 40 Hz33. Numerous
studies reported that the gamma band is most apparent at the frequency of 40 Hz3438. While most gamma oscilla-
tion study emphasized frequencies around 40 Hz, electrocorticographic recordings (ECoG) in patients enduring
epilepsy have suggested that the functional activation may be more consistently connected with the higher fre-
quencies, typically greater than 60 Hz and may extend up to 200 Hz and beyond39. In our study, the 30 Hz to 40 Hz
was chosen as gamma band since all the analysis of this study is based on EEG and also the interpretation of the
earlier EEG based concussion studies suggested that the EEG spectrum contains some characteristic waveforms
associated with concussion which primarily fall within the frequency band of 1 to 40 Hz15,4047. Moreover, the use
of high-frequency gamma band is still controversial as studies showed that the change in the higher frequency
gamma-component might be a result of the higher amount of artifact from the electromyographic activity4850.
Aer calculating the PSD for each channel and bands, overall PSD was calculated by calculating the mean
PSD across all nine referential channels for both individual frequency bins and ve frequency bands. Linear
and nonlinear features were then extracted from the ve frequency bands and also from each of 1 to 40 Hz EEG
frequency bins.
is innovative analysis achieved a new range of frequencies with signicant dierences between healthy
and concussed groups even when the band base analysis was not adequate to reveal the decits. Moreover, in
this paper, we present an exploration of the usefulness of several features for use in concussion detection, which
aims at providing accurate feedback as early as possible. Along with the traditionally used band power estimates,
we computed some time domain as well as nonlinear features from each EEG frequency band and then again
computed all the features from each individual frequency bins. e parameters extracted from EEG signal are
explained as follows.
Linear Features. Power spectral density analysis was performed to extract the linear features from the signal. e
extracted features were; (i) average spectral power for ve frequency bands and (ii) the spectral power for each of
the individual frequency from 1 Hz to 40 Hz.
Time domain Feature. Most popular features used for concussion analysis are EEG band based power spectral
density. In this paper, we introduce new features for concussion assessment called Time Domain Parameters that
are also known as Hjorth parameters. e features are inspired by the fact that they have been previously used
in EEG based experiments like Vidaurre et al. used Hjorth parameters, in their brain-computer interface (BCI)
Figure 1. Experimental setup for EEG data collection. (a) Data collection set up for a participant, (b) Map of 9
Electrodes locations. e locations were plotted using EEGLAB31.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/
5
Scientific REPORTS | 7: 17221 | DOI:10.1038/s41598-017-17414-x
study51 whereas Cecchin et al. used Hjorth parameters for seizure assessment from raw scalp EEG signals52. e
parameters introduced by Hjorth53 are three features dened as follows by equation1 to 3:
=Activity xt varxt(()) (());(1)
=
()
Mobility xt var
varxt
(()) (()) ;
(2)
dx t
dt
()
=
()
Complexity xt Mobility
Mobility xt
(()) (())
(3)
dx t
dt
()
The first parameter, Activity, calculates the alteration of time signal and characterizes the signal power.
Mobility is computed by calculating the square root of the variance of the rst derivative of the signal divided by
the activity and thus species the average frequency or proportion of standard deviation of the spectral power.
Complexity describes the change in frequency by comparing the Mobility of the rst derivative of the signal
with the signal’s mobility, and for more resemblance between the signals, the value converges to one. ese three
parameters consider the frequency component of the signal itself and thus remain more robust against the errors
due to overtting or non-stationarities of the signal52. To reduce the complexity of calculation, these three param-
eters were calculated in a stationary mode of signal separately for each EEG channel of the entire signal. us, the
extracted parameters were three features per channel and, as a whole, a feature vector for each parameter. A total
of 27 features (3 features for nine channels) were extracted and then averaged for all channels.
Nonlinear Features. Dierent nonlinear parameters have been shown signicantly useful in the diagnosis of
neurological disorders. Nonlinear parameters like approximate entropy (ApEn), Hurst exponent, and Correlation
dimension have been used for automatic diagnosis of seizure onset and reported as a promising approach in dif-
ferentiating normal, pre-ictal and epileptic seizure from EEG signals54.
In the eld of cortical neuronal dynamic study, the existence of long-range temporal correlation (LRTC) is
considered a potential observed phenomenon as it is proven to be gradually reduced with the power-spectrum55.
e LRTC property of an amplitude-time signal has vital importance as it is found to have a relationship with
the distributed neural network33. Poil et al. reported the coexistence of LRTC property of amplitude time series
with neuronal avalanche activity56, and thus recommended a relationship between oscillatory activity detected
in the EEG and the criticality hypothesis56,57. Using these hypotheses, Shew et al. suggested a possible connection
between optimal functioning and LRTC in the amplitude of oscillations35. Moreover, the signicance of the LRTC
property has also been proven in numerous clinical studies linking a number of neuronal diseases (including
schizophrenia58, Alzheimer’s disease59, major depressive disorder60, and epilepsy61) with altered LRTC properties.
To quantify the degree of change in LRTC property in a signal, the Hurst exponent (H), (explained in a later para-
graph) is measured55. Hurst exponent was used by Holler et al. for the disorder of consciousness studies62 whereas
Culic et al. reported this property to be important to dierentiate epileptic patients63.
Another nonlinear parameter that was calculated was ApEn. ApEn is a widely known mathematical algorithm
which computes the predictability of time series data by quantifying the regularity and complexity of the signal.
ApEn quanties the logarithmic likelihood of the patterns in the signal that remain close on next incremental
comparisons64.
Values of the ApEn parameter have been reported signicantly dierent between EEGs collected from epilep-
tic seizure patients and normal EEG signals65. Guo et al. present a method based on approximate entropy for clas-
sifying the EEG regarding the existence and absence of seizures using the neural network with 99.85% accuracy66.
Inspired by these publications, we tested the ecacy of these features to distinguish healthy and concussed
athletes in this study. Approximate entropy (ApEn) and Hurst exponent were extracted as the nonlinear features
to measure synchrony and complexity of the EEG signal as explained in the following sections.
Approximate Entropy. ApEn was calculated for each frequency (1 to 40 Hz) and for each of ve frequency bands
of EEG data for all three dierent conditions in order to nd out if there was any relationship between the ran-
domness of EEG data along with a concussion. A lower value of approximate entropy species that the EEG data
is more deterministic whereas a higher value of ApEN determines the data is more random. is feature was
calculated using the ApEn function provided by Kijoon Lee in the MATLAB central le exchange67. e tolerance
chosen for ApEn calculation was two standard deviations.
Hurst Exponent. e Hurst exponent (H) calculates the extent information presented by a signal is related to
the history of the signal. e value of H varies from 0 to 1; 0 < H < 0.5 indicates the samples in the signals are
far apart and independent and thus the signal is short-range dependent. However, if 0.5 < H < 1, then the value
is said to contain LRTC, with higher values of H representing a stronger LRTC property55. e Hurst exponent
is thus known as the index of long-range dependence68. e value of H was calculated for each channel over the
entire EEG signal. A total of 9 components for 9 EEG channels were extracted for each signal.
Statistical Analysis. e decits between healthy and concussed groups were veried using statistical anal-
ysis, and the measurements were performed without knowledge of groups. e Shapiro-Wilk test was applied
to ascertain the normality of the data. For normally distributed data, a two-tailed Student t-test, followed by
Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/
6
Scientific REPORTS | 7: 17221 | DOI:10.1038/s41598-017-17414-x
Bonferroni’s post hoc test when applicable was implemented; otherwise, Wilcoxon rank sum test was considered.
e values in the manuscript are presented as mean ± standard deviation format with statistical signicance level
set at (p = 0.05). e test of signicance was performed using the MATLAB Statistical Toolbox69.
Data Availability. e database generated and/or analyzed during the current study will be available from
the corresponding author upon request.
Results
BESS Test. Postural decits in terms of the BESS associated with concussion showed no signicant dierence
between healthy and concussed group. Average sway per second was calculated using a modied Wii balance
board during the BESS assessment for healthy group (group average sway = 3.28 ± 0.69 cm) and concussed group
(group average sway = 3.00 ± 0.72 cm). e number of average BESS errors reported by the healthy group were
thirty compared to thirty-four reported by the concussed group. ough the average sway scores exhibited by
both groups were quite similar, the concussed group reported more errors than their healthy matched controls.
e t-test resulted in no signicant dierences (Average sway: p-value = 0.33, Number of errors: p-value = 0.39)
between the groups regarding average sway and number of errors.
K-D Test. K-D test measures the deciencies of attention and eye movements by capturing the speed of rapid
number naming. e athletes who sustained concussions required slightly more time to complete the task than
their peers in the healthy group (by approximately 0.1%), but the decits did not reach a level of signicance
(Healthy group: 53.20 ± 10.33, Concussed group: 53.74 ± 10.29; p-value = 0.966).
ImPACT Test. e healthy and concussed groups were not signicantly dierent with regard to age but were
signicantly dierent based on the number of prior concussions. A two-tailed t-test was performed to evaluate
the dierences in neuropsychological test performance regarding ImPACT battery between the concussed and
control groups. Table2 presents the detailed descriptive statistics for verbal and visual memory, processing speed,
and reaction time composite scores.
ough a number of studies reported the ability of the ImPACT to dierentiate healthy and concussed groups,
our analysis revealed no signicant dierence in any composite scores between the groups.
Neuronal Deficits in Terms of EEG Band-Power following Concussion. The EEG analysis was
conducted to extract the neuronal decits following a concussion. Athletes in the concussed group exhibited
an increase in delta and theta bands, and a decrease in alpha, beta and gamma frequencies compared to their
uninjured peers during all three testing conditions (Table3). As indicated in Table3, the dierence reached the
signicance level for the increase in delta band and decreased in alpha, beta and gamma frequency bands for all
three conditions.
Neuronal Decits in Terms of EEG Individual Frequency Power following Concussion. is anal-
ysis considered individual EEG frequencies to nd gaps between healthy and concussed groups. Figure2 shows
the results of both frequency band and individual frequency analysis for three experimental conditions (EO, EC,
and VT). e dashed black line shows the condence level of p = 0.05. e solid red lines show the p-value for
each frequency band (delta, theta, alpha, beta, and gamma bands). e bars in each frequency band show the
p-value for individual frequencies. To highlight the subject to subject variance for each group, a supplementary
table with the mean and standard deviation of power of each signicant frequency bins for each group is added
to the manuscript.
e athletes who sustained a concussion had a range of frequencies with a signicant dierence from the
healthy group during EO condition (1–3 Hz, 9–10 Hz, 20–24 Hz, 27–30 Hz, and 33–38 Hz) as shown in Fig.2(a).
A very similar, but not all range of signicance was exhibited during EC condition (1–3 Hz, 9–10 Hz, 15–18 Hz,
20–24 Hz, 28–30 Hz, and 35–38 Hz) as shown in Fig.2(b). e signicant individual frequencies exhibiting the
decits between healthy and concussed groups during VT condition were (1–3 Hz, 6–7 Hz, 9–10 Hz, 19–30 Hz,
34–38 Hz) and were much consistent with EO condition as shown in Fig.2(c).
Neuronal Decits in terms of Nonlinear Features from EEG Individual Frequency following
Concussion. In the nal analysis, nonlinear features were calculated in order to nd out if new features
Composite Scores
Healthy Group Concussed Group
F value p-Va lueMean ± SD Mean ± SD
Verbal Memory Index 89.93 ± 7.87 87.57 ± 9.25 0.58 0.58
Visual Memory Index 86.43 ± 5.37 82.57 ± 7.79 0.25 0.27
Motor Speed Index 40.36 ± 5.81 37.28 ± 5.37 0.75 0.20
Reaction Time Index 0.62 ± 0.09 0.65 ± 0.12 0.38 0.59
Impulse Control Index 5.86 ± 3.21 6.14 ± 2.9 0.86 0.84
Total Symptom Score Index 2.93 ± 2.13 3.14 ± 3.53 0.12 0.88
Table 2. Group means and standard deviations for ImPACT composite scores of healthy and concussed groups.
e test of signicance was performed with statistical signicance level of 0.05.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/
7
Scientific REPORTS | 7: 17221 | DOI:10.1038/s41598-017-17414-x
extracted from the EEG data can tabulate the deciencies due to a concussion. e extracted features were
approximate entropy, activity, mobility, complexity and Hurst exponent features. While calculating these fea-
tures for EEG frequency bands (delta, theta, alpha, beta, and gamma), no signicant decits were found between
healthy and concussed athletes. But when the analysis was done for individual frequencies instead of frequency
bands, interesting outcomes were exhibited. A set of individual frequencies was found for each nonlinear feature
which can reveal signicant decits between healthy and concussed athletes as reported in Fig.3. As shown in
Fig.3(a) for EO condition, the frequencies indicating signicant decits between healthy and concussed groups
in terms of 2 or more nonlinear features are 1–2 Hz, 8 Hz, 19 Hz, 21 Hz, 23 Hz, 25–26 Hz, 31 Hz, 34 Hz and 37 Hz.
For EC condition, Fig.3(b), the range of frequencies with decits in two or more features was for 1–2 Hz, 13 Hz,
16 Hz, 34 Hz and 37 Hz, and for VT condition, from Fig.3(c), the range was 1–2 Hz, 8 Hz, 15–16 Hz, 23 Hz, 26 Hz,
31 Hz, 34–35 Hz and 37 Hz. e most ecient nonlinear features to reveal deciency following concussion were
approximate entropy, activity and Hurst exponent feature.
Discussion
Residual damage to the brain due to concussion can oen evade clinical detection. Enhancing ways in which con-
cussion is assessed is pivotal, specically in susceptible individuals such as adolescent athletes where functional
decits can be elusive and seriously underreported. Better assessment is also essential since early identication of
the signs of a concussion can progress positive outcomes and thus suggests that there is a clear need for an eec-
tive evaluation approach to eciently assess and quantify high-risk individuals such as athletes who may have
already sustained a concussion. e current study aims to test the hypothesis that the concussion disrupts the
normal brain activities of a person. To detect these decits, we combined the BESS, K-D test, ImPACT, and EEG
analysis to capture the postural, suboptimal, neurophysiological and neuronal decits following a concussion.
Evidence from the previous studies29,40,70 shows that the cognitive impairment regarding the BESS is most
pronounced during the time of injury and 24 hours post injury but appears to resolve by day ve aer a concus-
sion incident. e balance decit through the BESS in our research resulted in no signicant dierence between
the healthy and concussed group and thus strengthened the already established hypothesis26,40,70 that the postural
decits resolve within a brief period post-injury and therefore may suggest that the BESS is not sensitive enough
to interpret any residual decits associated with long-term concussion history.
As expected, the K-D test, which is mainly a rapid screen tool and typically used immediately aer concussion25,
was unable to detect any decits in our study. is can be explained by the fact that the related visual decits due
to a concussion were resolved during the several months’ time gap between the concussion incident and data
collection.
e ImPACT was reported by multiple sports-related concussion studies as a potential tool to detect the
impaired neurocognitive functioning due to concussion2628. Also, some studies showed neuropsychological base-
line assessment models like ImPACT could assist the diagnosis of subtle neurocognitive deviations in athletes
aer a concussion incident26,27. ough several studies demonstrated that a history of concussion is associated
with poorer performance in ImPACT71, the role of concussion history remains a controversial issue, with various
studies yielding no relationship between concussion history and ImPACT performance28. e results of this man-
uscript suggest that there is no signicant eect of a history of concussion associated with performance measured
by ImPACT, which is understandable, as ImPACT is an immediate post-concussion paradigm, and due to the
long time gap between concussion incident and data collection, the sensitivity of the test deteriorates with time.
To capture the signature neuronal decits exhibited by concussed athletes that distinguish them from their
healthy peers, we evaluated several approaches utilizing a set of linear, time-frequency based features along with
nonlinear features extracted from EEG signals. In conjunction with band base analysis, this study undertook a
systematic exploration to nd out the decits within specic frequency bins from 1 to 40 Hz. e system works by
following four main steps: data acquisition, data preprocessing, feature extraction (power spectral, time domain
and nonlinear) and statistical analysis (functional decits detection).
For band base analysis, EEG was divided into traditional frequency bands (delta, theta, alpha, beta, and
gamma). Aer normalization, power spectral density analysis revealed a signicant dierence between healthy
and concussed athletes. ere are several ndings of interest. First, the PSD features collected from frequency
sub-bands played an important role in distinguishing concussed individuals. Discriminative features were
observed in delta, alpha, beta and gamma frequency bands. A dierence was also noted in theta frequency band.
Condition Participants
Delta (μV2)eta (μV2)Alpha (μV2)Beta (μV2)Gamma (μV2)
Mean ± SD Mean ± SD Mean ± SD Mean ± SD Mean ± SD
EO Condition Healthy 4.33 ± 0.25 3.47 ± 0.33 3.14 ± 0.36 2.48 ± 0.30 1.95 ± 0.15
Concussed 4.81 ± 0.34*3.67 ± 0.38 2.69 ± 0.25*2.14 ± 0.18*1.58 ± 0.21*
EC Condition Healthy 4.21 ± 0.33 3.43 ± 0.28 3.22 ± 0.23 2.46 ± 0.26 1.92 ± 0.13
Concussed 4.66 ± 0.30*3.59 ± 0.33 2.85 ± 0.34*2.13 ± 0.33*1.50 ± 0.34*
VT Conditi on Healthy 4.22 ± 0.24 3.38 ± 0.49 3.09 ± 0.40 2.47 ± 0.27 1.97 ± 0.17
Concussed 4.68 ± 0.45*3.58 ± 0.34 2.65 ± 0.28*2.08 ± 0.29*1.52 ± 0.21*
Table 3. EEG band power decits between healthy and concussed group for eyes open (EO); eyes closed (EC)
and vigilant task (VT) conditions. (*Denotes signicant dierences between healthy and concussed group at
statistical signicance level of 0.05).
Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/
8
Scientific REPORTS | 7: 17221 | DOI:10.1038/s41598-017-17414-x
It should be pointed out that similar frequency bands were targeted in some previous EEG studies of concus-
sion47,72,73. An increase in delta and theta frequency and a decrease in beta frequency were also reported by
McCrea et al.40 and Slobounov et al.72. e discrimination at reported by dierent frequency bands can indicate
signicant neuronal dysfunction. According to Demos et al.73, an increase in delta frequency may indicate brain
injuries, learning problems, or diculties with cognition. e decrease in alpha band power exhibited through
the analysis partially overlaps with the results reported by atcher et al. in a previously conducted mTBI based
study where coherence, phase, and power analysis was performed on EEG data collected from 130 participants43.
e decrease in alpha power exhibited by concussed athletes compared to control peers may be interpreted as a
reection of reduced cortical excitability74. A substantial decrease in beta and gamma power was also revealed
by the analysis. Certain levels of beta waves allow easy focus and involvement in conscious thought and logical
Figure 2. P-value vs. frequency plot. A set of individual frequencies from EEG data exhibits power spectral
density decits between healthy and concussed athletes. e X-axis in the gure shows the individual
frequencies and Y-axis shows the level of signicance. e color of bars is dierent based on each frequency
band, and the level of signicance for each EEG frequency band is shown by red lines. e p-value vs. frequency
is shown during three conditions (a) eyes open (EO) (b) eyes closed (EC), and (c) vigilant task (VT). All the test
of signicance was performed with statistical signicance level of 0.05.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/
9
Scientific REPORTS | 7: 17221 | DOI:10.1038/s41598-017-17414-x
thinking, whereas a decrease in beta waves may point to poor cognition, diculty in concentration73. Moreover, a
movement plan based study in terms of reaction time and endpoint error reported that a decrease in beta power is
correlated with higher end point error75. A study conducted by Kwon et al. demonstrated a reduced gamma power
by schizophrenia patients and concluded that the decit might reveal a less eective local neuronal synchroni-
zation to external stimuli in the thalamic sensory oscillations or in the sensory cortex76. A decrease in gamma
power was also reported to be correlated with lower consciousness in the anesthesia study conducted by Pritchett
et al.77. Several studies also reported that a decrease in gamma power is frequently related to an increase in the
low-frequency range (delta frequency band) power78,79 and interpreted to be related to lower neuronal activity of
the brain region that operates to generate behavior80. All these specic power increases in the slower frequency
band (delta), combined with the decrease of power in faster frequency bands (alpha, beta, gamma) exhibited by
concussed athletes may imply that their neurological status is not as sound as their healthy matched peers in the
control group.
ough a lot of studies revealed signicant dierences in EEG sub-bands, there is no signature prole to
indicate increase or decrease of band powers associated with concussion. at’s why the pathophysiology of con-
cussion is considered heterogeneous and not yet completely understood. To reinforce our EEG-based functional
decits hypothesis, in an innovative approach, the PSD based analysis for each of the EEG individual frequencies
was conducted. Aer analyzing 189 cases, i.e., three dierent trials in three dierent conditions (EO, EC, VT) for
21 subjects as shown in Fig.2, it was concluded that four ranges of frequencies are more ecient in highlighting
decits following a concussion. ese ranges are slow delta (1–2 Hz), slow alpha (9–10 Hz), fast beta (20–30 Hz)
and fast gamma (34–39 Hz). A similar individual frequency-based analysis conducted by us on eyes closed EEG
collected from a dierent dataset of 20 healthy and 20 immediate concussed athletes also resulted in a nearly simi-
lar range of frequencies (1–2 Hz of delta band, 8–10 Hz of the alpha band, 24–29 Hz of the beta band and 34–36 Hz
range within the gamma band)41. To date, no individual frequency based study was conducted for concussion
Figure 3. Frequencies with a signicant dierence in approximate entropy, activity, mobility, complexity and
Hurst exponent between healthy and concussed athletes for three conditions: (a) eyes open (EO), (b) eyes
closed (EC), and (c) vigilant task (VT). All the test of signicance was performed with statistical signicance
level of 0.05.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/
10
Scientific REPORTS | 7: 17221 | DOI:10.1038/s41598-017-17414-x
assessment, and more collaborative research is needed to establish a direct relationship of these frequency bins
with a concussion. e decrease in alpha band frequency bins exhibited through individual frequency analysis
partially overlaps with the results reported by atcher et al. in a previously conducted mTBI based study43. An
increase in theta band frequency bins during VT task may be associated with ADHD, depression, hyperactivity,
impulsivity, and inattentiveness51. e individual frequency-based analysis also revealed signicant dierences in
the upper level of beta bands compared to the lower level frequency bins. Oscillatory activity in the beta band was
previously reported to reect the presence of inhibition of the process of ongoing motor task81.
Elgendi et al. demonstrated an Alzheimer disease (AD) study and reported that new optimized frequency
ranges (4–7 Hz, 8–15 Hz, 19–24 Hz) resulted in better classication accuracy than the traditional frequency bands
for the diagnosis of AD82. Similarly, if we consider the neurological decits observed in individual frequency bins,
as well as in the conventional frequency bands as a whole, the most reliable interpretation is that these decits
may be a consequence of their injury and can possibly be used as a concussion assessment index to identify the
concussed athletes at the time of injury or during the post-concussion recovery period.
In the second phase of this study, a set of time-domain and nonlinear features were extracted. ese features
have been proven to be suitable to characterize neurological disorders like epilepsy, attention-decit/hyperactivity
disorder (ADHD) and Alzheimer disease in the literature83. It was hypothesized that the time domain and nonlin-
ear feature based study could reveal new aspects and provide more information regarding the complex and cha-
otic nature of the EEG data. As reported by Mohammadi et al.84, quantitative measures of chaos and non-linear
features are convenient descriptive tools to characterize electrophysiological abnormalities in neuropsychiatric
disorders that are not evident in linear analysis. To show the eectiveness of these features for a concussion, in
a similar approach to power analysis, the features were calculated for both frequency bands and individual EEG
frequencies. ough the concussed athletes exhibit dierent values for Hjorth time domain parameters and non-
linear parameters like approximate entropy and Hurst exponent, none of the parameters showed a signicant
dierence compared to their healthy peers for traditional EEG frequency bands. But, when the analysis was done
for each frequency, it was noted that signicant dierences were observed for certain frequencies as shown in
Fig.3(a–c).
e observation of signicantly dierent nonlinear features also revealed important notions about concussed
athletes. e concussed athletes exhibited a decrease in Hjorth complexity and mobility. It has been reported by
Pezard et al.85 that depressive subjects tend to display lower complexity than controls. Moreover, Hamida et al.86,86
reported the decreased complexity and mobility are associated with insomniac subjects. Approximate entropy
quanties the amount of regularity in data by calculating the upcoming amplitude values of the signal based
on the knowledge of the preceding amplitude values87. Sohn et al.88 reported a signicantly lower approximate
entropy for a group of ADHD subjects compared to matched controls and hypothesized that the patients might
not have sucient levels of cortical activation to reach the requirements of attention-demanding tasks. Following
their hypothesis, a signicant decrease in approximate entropy exhibited by concussed athletes may point out that
their cortical information processing is altered compared to healthy athletes. Moreover, many pathological dis-
order studies like schizophrenia, posttraumatic stress disorder, panic disorder, and epilepsy reported lower com-
plexity in pathological states compared to healthy subjects89. e notion claimed by the authors is that the lower
EEG complexity is attributed to the abnormal neural integration in the above-mentioned mental disorders58 and
thus a lower value of ApEn demonstrated by concussed athletes in our study implies that they may still have some
irregularity in their neural integration.
Another nonlinear feature with a signicant dierence was the Hurst exponent. Higher values of Hurst expo-
nent indicate a stronger long-range temporal correlation of amplitude uctuations of EEG55. In accordance with
the result reported by Geng et al.90 in their epileptic study, a decreased Hurst exponent exhibited by concussed
athletes in our study implies that the degree of anti-correlation of concussed athletes is larger than that of healthy
athletes.
The most efficient frequencies indicating the deficits were found to be 1–3 Hz, 21–24 Hz, 28–30 Hz and
35–38 Hz. Among the EEG task condition, EO and VT conditions were found to be more ecient in identifying
hidden decits due to a concussion. ough conventional band base analysis revealed no signicant dierence
between healthy and concussed athletes regarding time domain and nonlinear features, individual frequency
analysis was ecacious to exhibit these hidden discrepancies. ese dierences at specic frequencies would
remain unnoticed if only conventional frequency bands were considered. Ultimately, this study exposed the fact
that EEG analysis for each frequency is equally as important as conventional bands to evaluate the neurological
dysfunction following a concussion.
Conclusions
is study suggests that EEG analysis is more sensitive compared to cognitive testing to decipher persistent seque-
lae of sport-related concussion. For the rst time, a set of time domain and nonlinear EEG features was utilized in
addition to the standard frequency band features to highlight neuronal decits following a concussion. Also, the
approach of analysis using individual frequencies of EEG was conducted for the rst time to study concussion.
is innovative approach combined with novel features opens a new door to interpret subtle post-concussion
decits. While no previous work was done to nd the post-concussive decits in individual frequency level, the
result demonstrated a new range of frequency which is more successful to reveal the discrepancies. In sum, accu-
mulated evidence from this study suggests that the proposed approach of EEG analysis was successful to identify
that the athletes with a history of concussive injury still exhibited neurological alterations, despite reporting to be
symptom-free by standard postural, visual or neurophysiological tests.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/
11
Scientific REPORTS | 7: 17221 | DOI:10.1038/s41598-017-17414-x
Limitations and Future Work. Although the current manuscript has several strengths, some limitations
should be considered. First, the analysis was cross-sectional, and it is always possible that some unmeasured vari-
able may add to the current group alterations. is probability is minimalized, however, as the study groups were
co-players cautiously matched for weight, height, age, years of education, and sport. Although we repeated the
experiment in three separate sessions and using three dierent conditions, the data set is small and was limited
to male athletes only. As such, the conclusions drawn from the current dataset should be used to guide similar
studies on larger datasets and other age groups. However, this is an ongoing project, and we are collecting data
from more participants so that more rigorous quantitative and qualitative analysis can be performed with a larger
data set consisting of recordings from a large number of subjects in the future. Future work would also include
applying the proposed methodology for the classication of two classes, namely healthy and concussed, to detect
and predict the concussion from EEG signals for the normal and abnormal condition. erefore, our ndings
will engender more comprehensive evaluations towards clinical applicability of concussion assessment for proper
diagnosis and prevention through accurate RTP decision, as well as managing the treatment and rehabilitation
ecacy post-concussion.
References
1. McCrory, P. et al. Consensus Statement on Concussion in Sport-e 4th International Conference on Concussion in Sport Held in
Zurich. November 2012. PM  5, 255–279 (2013).
2. Langlois, J. A., utland-Brown, W. & Wald, M. M. e epidemiology and impact of traumatic brain injury: a brief overview. J. Head
Trauma ehabil. 21, 375–378 (2006).
3. Daneshvar, D. H., Nowinsi, C. J., Mcee, A. C. & Cantu, . C. e epidemiology of sport-related concussion. Clin. Sports Med. 30,
1–17, vii (2011).
4. Bleiberg, J. et al. Duration of cognitive impairment aer sports concussion. Neurosurgery 54, 1073-78-80 (2004).
5. Hinton-bayre, A. D., Geen, G. & McFarland, . Mild head injury and speed of information processing: A prospective study of
professional rugby league players. J. Clin. Exp. Neuropsychol. 19, 275–289 (1997).
6. Basford, J. . et al. An assessment of gait and balance decits aer traumatic brain injury. Arch. Phys. Med. ehabil. 84, 343–349
(2003).
7. Davis, G. A., Iverson, G. L., Gusiewicz, . M., Ptito, A. & Johnston, . M. Contributions of neuroimaging, balance testing,
electrophysiology and blood marers to the assessment of sport-related concussion. Br. J. Sports Med 43, i36–i45 (2009).
8. Parer, T. M., Osternig, L. ., Van Donelaar, P. & Chou, L.-S. Gait Stability following Concussion. Med. Sci. Sport. Exerc 38,
1032–1040 (2006).
9. Parer, T. M., Osternig, L. ., Lee, H.-J., Donelaar, P. van & Chou, L.-S. e eect of divided attention on gait stability following
concussion. Clin. Biomech. 20, 389–395 (2005).
10. Iverson, G. L., Broos, B. L., Collins, M. W. & Lovell, M. . Tracing neuropsychological recovery following concussion in sport.
Brain Inj. 20, 245–252 (2006).
11. McCrea, M., Hammee, T., Olsen, G., Leo, P. & Gusiewicz, . Unreported concussion in high school football players: implications
for prevention. Clin. J. Sport Med. 14, 13–7 (2004).
12. Broshe, D. ., De Marco, A. P. & Freeman, J. . A review of post-concussion syndrome and psychological factors associated with
concussion. Brain Inj. 29, 228–237 (2015).
13. Stuart, S., Hicey, A., Morris, ., O’Donovan, . & Godfrey, A. Concussion in contact sport: A challenging area to tacle. Journal of
Sport and Health Science https://doi.org/10.1016/j.jshs.2017.03.009 (2017).
14. Baillargeon, A., Lassonde, M., Leclerc, S. & Ellemberg, D. Neuropsychological and neurophysiological assessment of sport
concussion in children, adolescents and adults. Brain Inj. 26, 211–220 (2012).
15. Balan, O., Member, S., Virji-babul, N., Miyaoshi, M. & Maeig, S. Source-domain Spectral EEG Analysis of Sports-elated
Concussion via Measure Projection Analysis? In Engineering in Medicine and Biology Society (EMBC), 2015 37th Annual
International Conference of the IEEE 4053–4056 (IEEE, 2015).
16. Garg, S. et al. A Comparison of EEG Power Spectral and Wavelet Features in Concussed Cohorts Using Support Vector Machine.
IEEE Eng. Med. Biol. Soc 29, 2014 (2014).
17. Zhou, Y. et al. Mild Traumatic Brain Injury : Longitudinal egional Brain Volume Changes. adiology 267, 880–890 (2013).
18. Prichep, L. S., McCrea, M., Barr, W., Powell, M. & Chabot, . J. Time Course of Clinical and Electrophysiological ecovery Aer
Sport-elated Concussion. J. Head Trauma ehabil. 28, 266–273 (2013).
19. orn, A., Golan, H., Melamed, I., Pascual-Marqui, . & Friedman, A. Focal cortical dysfunction and blood-brain barrier disruption
in patients with Postconcussion syndrome. J. Clin. Neurophysiol. 22, 1–9 (2005).
20. CDC. Get the Stats on Traumatic Brain Injury in the United States. at https://www.cdc.gov/traumaticbraininjury/pdf/BlueBoo_
factsheet-a.pdf
21. Signs and Symptoms | Concussion | Traumatic Brain Injury | CDC Injury Center. at https://www.cdc.gov/traumaticbraininjury/
symptoms.html
22. Giza, C. C. et al. Summary of evidence-based guideline update: Evaluation and management of concussion in sports: eport of the
Guideline Development Subcommittee of the American Academy of Neurology. Neurology 80, 2250–2257 (2013).
23. Chapter 15.1-18.2 Concussion Management for athletes and teacher support program. North Daota Century Code 1–2 at http://
www.legis.nd.gov/cencode/t15-1c18-2.pdf. (2011).
24. Bell, D. ., Gusiewicz, . M., Clar, M. A. & Padua, D. A. Systematic review of the balance error scoring system. Sports Health 3,
287–95 (2011).
25. Howitt, S. et al. e utility of the ing-Devic test as a sideline assessment tool for sport-related concussions: a narrative review. J.
Can. Chiropr. Assoc 60, 322–329 (2016).
26. Lovell, M. . et al. ecovery from mild concussion in high school athletes. J. Neurosurg. 98, 296–301 (2003).
27. Schatz, P., Pardini, J. E., Lovell, M. ., Collins, M. W. & Podell, . Sensitivity and specificity of the ImPACT Test Battery for
concussion in athletes. Arch. Clin. Neuropsychol. 21, 91–99 (2006).
28. Iverson, G. L., Gaetz, M., Lovell, M. . & Collins, M. W. elation between subjective fogginess and neuropsychological testing
following concussion. J. Int. Neuropsychol. Soc. 10, 904–6 (2004).
29. Munia, T. T. . et al. Preliminary results of residual decits observed in athletes with concussion history: Combined EEG and
cognitive study. Proc. Annu. Int. Conf. IEEE Eng. Med. Biol. Soc. EMBS 2016–Octob, 41–44 (2016).
30. B-alert x10: User manual. Carlsbad, CA. Advanced Brain Monitoring. at http://www.advancedbrainmonitoring.com/xseries/x10/
(2010).
31. Delorme, A. & Maeig, S. EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent
component analysis. J. Neurosci. Methods 134, 9–21 (2004).
32. Urigüen, J. A. & Garcia-Zapirain, B. EEG artifact removal—state-of-the-art and guidelines. J. Neural Eng. 12, 31001 (2015).
Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/
12
Scientific REPORTS | 7: 17221 | DOI:10.1038/s41598-017-17414-x
33. ay, S., Crone, N. E., Niebur, E., Franaszczu, P. J. & Hsiao, S. S. Neural correlates of high-gamma oscillations (60-200 Hz) in
macaque local eld potentials and their potential implications in electrocorticography. J. Neurosci. 28, 11526–36 (2008).
34. Gold, I. Does 40-Hz Oscillation Play a ole in Visual Consciousness? Conscious. Cogn. 8, 186–195 (1999).
35. Cric, F. & och, C. Towards a neurobiological theory of consciousness. Semin. Neurosci. 2, 263–275 (1990).
36. Cric, F. & och, C. A framewor for consciousness. Nat. Neurosci. 6, 119–126 (2003).
37. Cardin, J. A. et al. Driving fast-spiing cells induces gamma rhythm and controls sensory responses. Nature 459, 663–667 (2009).
38. Iaccarino, H. F. et al. Gamma frequency entrainment attenuates amyloid load and modies microglia. Nature 540, 230–235 (2016).
39. Crone, N. E., Sinai, A. & orzeniewsa, A. High-frequency gamma oscillations and human brain mapping with electrocorticography.
159, 275–295 (2006).
40. McCrea, M. Acute eects & recovery aer sport related concussion: A Quantitative Brain Electrical Activity Study. J. Head Trauma
ehabil. 25, 283–293 (2010).
41. Munia, T. T. ., Haider, A. & Fazel-ezai, . Evidences of Brain Functional Decits Following Sport-elated Mild Traumatic Brain
Injury. In IEEE Engineering in Medicine and Biology Society 3212–3215. (IEEE, 2017).
42. Munia, T. T. . et al. Preliminary results of residual decits observed in athletes with concussion history: Combined EEG and
cognitive study. In Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS
2016–Octob, 41–44 (2016).
43. Thatcher, . W., Waler, . A., Gerson, I. & Geisler, F. H. EEG discrimination of mild head injury. Electroencephalogr. Clin.
Neurophysiol. 73, 94–106 (1989).
44. ompson, J. W. G. In Foundations of Sport-elated Brain Injuries 341–374 (Springer US, 2006). https://doi.org/10.1007/0-387-
32565-4_15
45. Nuwer, M. ., Hovda, D. A., Schrader, L. M. & Vespa, P. M. outine and quantitative EEG in mild traumatic brain injury. https://doi.
org/10.1016/j.clinph.2005.05.008 (2005).
46. Cao, C. & Slobounov, S. Application of a novel measure of EEG non-stationarity as ‘Shannon- entropy of the pea frequency shiing’
for detecting residual abnormalities in concussed individuals. Clin. Neurophysiol. 122, 1314–21 (2011).
47. Teel, E. F., ay, W. J., Geronimo, A. M. & Slobounov, S. M. esidual alterations of brain electrical activity in clinically asymptomatic
concussed individuals: An EEG study. Clin. Neurophysiol. 125, 703–707 (2014).
48. Whitham, E. M. et al. ining activates EMG in scalp electrical recordings. Clin. Neurophysiol. 119, 1166–1175 (2008).
49. Whitham, E. M. et al. Scalp electrical recording during paralysis: Quantitative evidence that EEG frequencies above 20Hz are
contaminated by EMG. Clin. Neurophysiol. 118, 1877–1888 (2007).
50. Yuval-Greenberg, S., Tomer, O., eren, A. S., Nelen, I. & Deouell, L. Y. Transient Induced Gamma-Band esponse in EEG as a
Manifestation of Miniature Saccades. Neuron 58, 429–441 (2008).
51. Vidaurre, C., rämer, N., Blanertz, B. & Schlögl, A. Time Domain Parameters as a feature for EEG-based Brain–Computer
Interfaces. Neural Networs 22, 1313–1319 (2009).
52. C ecchin, T. et al. Seizure lateralization in scalp EEG using Hjorth parameters. Clin. Neurophysiol. 121, 290–300 (2010).
53. Hjorth, B. EEG analysis based on time domain properties. Electroencephalogr. Clin. Neurophysiol. 29, 306–310 (1970).
54. umar, Y., Dewal, M. L. & Anand, . S. Epileptic seizure detection using DWT based fuzzy approximate entropy and support vector
machine. Neurocomputing 133, 271–279 (2014).
55. Blythe, D. A. J., Haufe, S., Müller, .-. & Niulin, V. V. e eect of linear mixing in the EEG on Hurst exponent estimation.
Neuroimage 99, 377–387 (2014).
56. Poil, S.-S., Hardstone, ., Mansvelder, H. D. & Linenaer-Hansen, . Critical-State Dynamics of Avalanches and Oscillations
Jointly Emerge from Balanced Excitation/Inhibition in Neuronal Networs. J. Neurosci. 32, 9817–9823 (2012).
57. Friedman, N. et al. Universal Critical Dynamics in High esolution Neuronal Avalanche Data. Phys. ev. Lett. 108, 208102 (2012).
58. Niulin, V. V., Jönsson, E. G. & Brismar, T. Attenuation of long-range temporal correlations in the amplitude dynamics of alpha and
beta neuronal oscillations in patients with schizophrenia. Neuroimage 61, 162–169 (2012).
59. Montez, T. et al. Altered temporal correlations in parietal alpha and prefrontal theta oscillations in early-stage Alzheimer disease.
Proc. Natl. Acad. Sci. USA 106, 1614–9 (2009).
60. Linenaer-Hansen, . et al. Breadown of Long-ange Temporal Correlations in eta Oscillations in Patients with Major
Depressive Disorder. J. Neurosci. 25, 10131–10137 (2005).
61. Monto, S., Vanhatalo, S., Holmes, M. D. & Palva, J. M. Epileptogenic Neocortical Networs Are evealed by Abnormal Temporal
Dynamics in Seizure-Free Subdural EEG. Cereb. Cortex 17, 1386–1393 (2007).
62. Höller, Y. et al. Comparison of EEG-Features and Classification Methods for Motor Imagery in Patients with Disorders of
Consciousness. PLoS One 8, e80479 (2013).
63. Ćulić, M., Stojadinović, G., Martać, L. & Soović, M. Use of the Hurst Exponent for Analysis of Electrocortical Epileptiform Activity
Induced in ats by Administration of Camphor Essential Oil or 1,8-Cineole. Neurophysiol. Neiroziologiya/Neurophysiology 42,
64–69 (2010).
64. Puthanattil, S. D. & Joseph, P. . Analysis of EEG Signals Using Wavelet Entropy and Approximate Entropy : A Case Study on
Depression Patients 8, 420–424 (2014).
65. Vijith, V. S., Jacob, J. E., Iype, T., Gopaumar, . & Yohannan, D. G. Epileptic seizure detection using non linear analysis of EEG. In 2016
International Conference on Inventive Computation Technologies (ICICT) 1–6, https://doi.org/10.1109/INVENTIVE.2016.7830193
(IEEE, 2016).
66. Guo, L., ivero, D. & Pazos, A. Epileptic seizure detection using multiwavelet transform based approximate entropy and articial
neural networs. J. Neurosci. Methods 193, 156–163 (2010).
67. i, Chon, Scully, C. & Sheng, Lu Approximate entropy for all signals. IEEE Eng. Med. Biol. Mag. 28, 18–23 (2009).
68. Hurst, H. E. Long-Term Storage Capacity of eservoirs. Trans. Am. Soc. Civ. Eng 116, 770–799 (1951).
69. Statistics and Machine Learning Toolbox Documentation. at https://www.mathwors.com/help/stats/
70. iemann, B. L. & Gusiewicz, . M. Eects of Mild Head Injury on Postural Stability as Measured rough Clinical Balance Testing.
J. Athl. Train. 35, 19–25 (2000).
71. Schatz, P. Long-Term Test-etest eliability of Baseline Cognitive Assessments Using ImPACT. Am. J. Sports Med. 38, 47–53 (2010).
72. Slobounov, S., Sebastianelli, W. & Hallett, M. esidual brain dysfunction observed one year post-mild traumatic brain injury:
Combined EEG and balance study. Clin. Neurophysiol. 123, 1755–1761 (2012).
73. Demos, J. N. Getting Started with Neurofeedbac. WW Norton Co https://doi.org/10.1016/j.jpsychores.2005.08.007 (2005).
74. Nunez, P. L. & Srinivasan, . Electric Fields of the Brain :e Neurophysics of EEG. (Oxford University Press, Inc, 2006). at https://
brainmaster.com/soware/pubs/brain/Nunez2ed.pdf (2006).
75. Yang, L., Leung, H., Plan, M., Snider, J. & Poizner, H. Alpha and beta band power changes predict reaction time and endpoint error
during planning reaching movements. In 2014 7th International Conference on Biomedical Engineering and Informatics 264–268,
https://doi.org/10.1109/BMEI.2014.7002782 (IEEE, 2014).
76. won, J. S. et al. Gamma Frequency–ange Abnormalities to Auditory Stimulation in Schizophrenia. Arch. Gen. Psychiatry 56, 1001
(1999).
77. Pritchett, S. et al. Power analysis of gamma frequencies (30–47Hz), adjusting for muscle activity (80–97Hz), in anesthesia: A
comparison between young adults, middle-aged and the elderly. In 2008 30th Annual International Conference of the IEEE
Engineering in Medicine and Biology Society 825–830, https://doi.org/10.1109/IEMBS.2008.4649281 (IEEE, 2008).
Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/
13
Scientific REPORTS | 7: 17221 | DOI:10.1038/s41598-017-17414-x
78. Crone, N. E., Miglioretti, D. L., Gordon, B. & Lesser, . P. Functional mapping of human sensorimotor cortex with
electrocorticographic spectral analysis. II. Event-related synchronization in the gamma band. Brain 121(Pt 12), 2301–15 (1998).
79. Crone, N. E. et al. Functional mapping of human sensorimotor cortex with electrocorticographic spectral analysis. I. Alpha and beta
event-related desynchronization. Brain 121(Pt 12), 2271–99 (1998).
80. Swann, N. C. et al. Motor System Interactions in the Beta Band Decrease during Loss of Consciousness. https://doi.org/10.1162/
jocn_a_00884
81. Pfurtscheller, G. Event-related synchronization (ES): an electrophysiological correlate of cortical areas at rest. Electroencephalogr.
Clin. Neurophysiol. 83, 62–9 (1992).
82. Elgendi, M. et al. Optimization of EEG frequency bands for improved diagnosis of Alzheimer disease. In 2011 Annual International
Conference of the IEEE Engineering in Medicine and Biology Society 2011, 6087–6091 (IEEE, 2011).
83. Stam, C. J. Nonlinear dynamical analysis of EEG and MEG: eview of an emerging eld. Clin. Neurophysiol. 116, 2266–2301 (2005).
84. Mohammadi, M. . et al. EEG classication of ADHD and normal children using non-linear features and neural networ. Biomed.
Eng. Lett 6, 66–73 (2016).
85. Pezard, L. et al. Depression as a dynamical disease. Biol. Psychiatry 39, 991–9 (1996).
86. Hamida, S. T. B, Ahmed, B. & Penzel, T. A novel insomnia identication method based on Hjorth parameters. 2015 IEEE Int. Symp.
Signal Process. Inf. Technol. ISSPIT 2015 548–552, https://doi.org/10.1109/ISSPIT.2015.7394397 (2016).
87. Bruhn, J., öpce, H. & Hoe, A. Approximate entropy as an electroencephalographic measure of anesthetic drug eect during
desurane anesthesia. Anesthesiology 92, 715–26 (2000).
88. Sohn, H. et al. Linear and non-linear EEG analysis of adolescents with attention-decit/hyperactivity disorder during a cognitive
tas. Clin. Neurophysiol. 121, 1863–1870 (2010).
89. Taahashi, T. Complexity of spontaneous brain activity in mental disorders. Prog. Neuro-Psychopharmacology Biol. Psychiatry 45,
258–266 (2013).
90. Geng, S., Zhou, W., Yuan, Q., Cai, D. & Zeng, Y. EEG non-linear feature extraction using correlation dimension and Hurst exponent.
Neurol. es. 33, 908–912 (2011).
Acknowledgements
Financial support from University of North Dakota Faculty and Collaborative Seed Money is gratefully
acknowledged. We also gratefully acknowledge Jerey Gendreau and Jessica I. Herren for their assistance in data
collection. In addition, we would like to thank Dr. Colin Combs and Dr. Kumi Nagamoto-Combs from School
of Medicine and Health Sciences of University of North Dakota for their suggestions in the initial study design.
Author Contributions
T.T.K.M., A.H., M.R. and R.F-R. conceptualized the research. M.R. and R.F-R. designed the experiment. C.S. and
M.R. performed the experiment; T.T.K.M., and A.H. conducted the formal analysis; T.T.K.M.,A.H., and R.F-R.
discussed and interpreted the results;T.T.K.M., A.H., and R.F-R. wrote the original dra; All authors reviewed
and edited the nal manuscript.
Additional Information
Supplementary information accompanies this paper at https://doi.org/10.1038/s41598-017-17414-x.
Competing Interests: e authors declare that they have no competing interests.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional aliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Cre-
ative Commons license, and indicate if changes were made. e images or other third party material in this
article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons license and your intended use is not per-
mitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the
copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
© e Author(s) 2017
Content courtesy of Springer Nature, terms of use apply. Rights reserved
1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com
... This explanation may also apply to the three participants in the control group who were misclassified as concussed. Longitudinal neuroimaging studies show that in some individuals, mTBI induced changes in the brain can persist for several months and even past a year [16][17][18][19][20][21][22][23][24] . These persistent alterations are seen in EEG, fMRI and DTI data. ...
... We did not further filter or otherwise clean the data to remove artefacts due to line noise, eye blinks and motion, and electromyogram (EMG) contamination as is typically done (c.f. Porter et al. 88 , Rotem-Kohavi et al. 52,53 , Munia et al. 20 ). The resting state EEG data were acquired in binary simple (.RAW ) format 89 , and converted to Matlab 79 (.mat) format using EEGLab 90 for further processing and analysis. ...
... Our choice to use 90-second long segments reflects a compromise between wanting to use a short EEG sample on the one hand, and on other hand, ensuring that the sample is sufficiently long to capture low frequency brain activity, i.e. oscillations with frequencies down to 0.1 Hz. The latter constraint was motivated by two considerations: (1) EEG studies find that, compared to non-concussed controls, concussed athletes exhibit altered activity in the delta band (0.5-4 Hz) 20,34,91 , with increasing divergences observed near the lowest frequencies. ...
Article
Full-text available
Concussion is a global health concern. Despite its high prevalence, a sound understanding of the mechanisms underlying this type of diffuse brain injury remains elusive. It is, however, well established that concussions cause significant functional deficits; that children and youths are disproportionately affected and have longer recovery time than adults; and that individuals suffering from a concussion are more prone to experience additional concussions, with each successive injury increasing the risk of long term neurological and mental health complications. Currently, the most significant challenge in concussion management is the lack of objective, clinically- accepted, brain-based approaches for determining whether an athlete has suffered a concussion. Here, we report on our efforts to address this challenge. Specifically, we introduce a deep learning long short-term memory (LSTM)-based recurrent neural network that is able to distinguish between non-concussed and acute post-concussed adolescent athletes using only short (i.e. 90 s long) samples of resting state EEG data as input. The athletes were neither required to perform a specific task nor expected to respond to a stimulus during data collection. The acquired EEG data were neither filtered, cleaned of artefacts, nor subjected to explicit feature extraction. The LSTM network was trained and validated using data from 27 male, adolescent athletes with sports related concussion, benchmarked against 35 non-concussed adolescent athletes. During rigorous testing, the classifier consistently identified concussions with an accuracy of > 90% and achieved an ensemble median Area Under the Receiver Operating Characteristic Curve (ROC/AUC) equal to 0.971. This is the first instance of a high-performing classifier that relies only on easy-to-acquire resting state, raw EEG data. Our concussion classifier represents a promising first step towards the development of an easy-to-use, objective, brain-based, automatic classification of concussion at an individual level.
... Electrophysiological assessments of concussions also show promise. Electroencephalography (EEG) serves as a non-invasive method to measure electrical activity, providing insight into brain activity connected to concussion pathology 3 and has been used to identify functional changes in the brain following a concussion [28][29][30][31][32][33] . EEG is more manageable and inexpensive than many other brain imaging techniques 32 and although still needing special training, it is more accessible for researchers and clinicians. ...
... Age mean (SD) 30 www.nature.com/scientificreports/ Virtual reality experiment. ...
... The power spectral density (PSD) was computed for each epoch with 'Welch's method, with the following frequency bands: delta (1-4 Hz), theta (4-8 Hz), alpha (8)(9)(10)(11)(12)(13), beta (13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30), low gamma (30)(31)(32)(33)(34)(35)(36)(37)(38)(39)(40)(41)(42)(43)(44)(45). The relative power of each band was then computed obtaining a total of 5 EEG-related features. ...
Article
Full-text available
Current diagnosis of concussion relies on self-reported symptoms and medical records rather than objective biomarkers. This work uses a novel measurement setup called BioVRSea to quantify concussion status. The paradigm is based on brain and muscle signals (EEG, EMG), heart rate and center of pressure (CoP) measurements during a postural control task triggered by a moving platform and a virtual reality environment. Measurements were performed on 54 professional athletes who self-reported their history of concussion or non-concussion. Both groups completed a concussion symptom scale (SCAT5) before the measurement. We analyzed biosignals and CoP parameters before and after the platform movements, to compare the net response of individual postural control. The results showed that BioVRSea discriminated between the concussion and non-concussion groups. Particularly, EEG power spectral density in delta and theta bands showed significant changes in the concussion group and right soleus median frequency from the EMG signal differentiated concussed individuals with balance problems from the other groups. Anterior–posterior CoP frequency-based parameters discriminated concussed individuals with balance problems. Finally, we used machine learning to classify concussion and non-concussion, demonstrating that combining SCAT5 and BioVRSea parameters gives an accuracy up to 95.5%. This study is a step towards quantitative assessment of concussion.
... In a sample of neuropathic CP patients vs. controls, Sarnthein et al. [52] report significant increases in absolute delta, theta, alpha, and beta oscillatory power . Importantly, the authors included another group of CP patients who were not taking centrally acting medication and found the same increases in power (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18), except not in the high beta range (18)(19)(20)(21)(22)(23)(24)(25). Sarnthein [52] also found that these elevations of power and pain attenuated after receiving a therapeutic central lateral thalamotomy. ...
... Delta (1-4 Hz) 0.35 p < 0.01 Theta (4)(5)(6)(7) 0.31 p < 0.05 Alpha (7)(8)(9)(10)(11)(12)(13) 0.32 p < 0.05 Low Beta (13)(14)(15) 0.32 p < 0.05 High Beta (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30) 0.06 n.s Gamma (30)(31)(32)(33)(34)(35)(36)(37)(38)(39)(40)(41)(42)(43)(44)(45) −0.06 n.s Correlations between absolute power and duration of symptoms (time since MVA) in patients with PCS + CP.. ...
... Delta (1-4 Hz) 0.35 p < 0.01 Theta (4)(5)(6)(7) 0.31 p < 0.05 Alpha (7)(8)(9)(10)(11)(12)(13) 0.32 p < 0.05 Low Beta (13)(14)(15) 0.32 p < 0.05 High Beta (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30) 0.06 n.s Gamma (30)(31)(32)(33)(34)(35)(36)(37)(38)(39)(40)(41)(42)(43)(44)(45) −0.06 n.s Correlations between absolute power and duration of symptoms (time since MVA) in patients with PCS + CP.. ...
Article
Full-text available
(1) Background: Mild traumatic brain injury produces significant changes in neurotransmission including brain oscillations. We investigated potential quantitative electroencephalography biomarkers in 57 patients with post-concussive syndrome and chronic pain following motor vehicle collision, and 54 healthy nearly age- and sex-matched controls. (2) Methods: Electroencephalography processing was completed in MATLAB, statistical modeling in SPSS, and machine learning modeling in Rapid Miner. Group differences were calculated using current-source density estimation, yielding whole-brain topographical distributions of absolute power, relative power and phase-locking functional connectivity. Groups were compared using independent sample Mann-Whitney U tests. Effect sizes and Pearson correlations were also computed. Machine learning analysis leveraged a post hoc supervised learning support vector non-probabilistic binary linear kernel classification to generate predictive models from the derived EEG signatures. (3) Results: Patients displayed significantly elevated and slowed power compared to controls: delta (p = 0.000000, r = 0.6) and theta power (p < 0.0001, r = 0.4), and relative delta power (p < 0.00001) and decreased relative alpha power (p < 0.001). Absolute delta and theta power together yielded the strongest machine learning classification accuracy (87.6%). Changes in absolute power were moderately correlated with duration and persistence of symptoms in the slow wave frequency spectrum (<15 Hz). (4) Conclusions: Distributed increases in slow wave oscillatory power are concurrent with post-concussive syndrome and chronic pain.
... Two of the main limitations of EEG analysis are the time required and the need for specialized expert knowledge to identify features in the EEG signal. Recently, machine learning approaches such as support vector machine (SVM), random forest (e.g., Cao and Slobounov, 2010;Munia et al., 2017;Vergara et al., 2017;Jacquin et al., 2018;Wickramaratne et al., 2020) as well as deep learning, convolution neural network (Boshra et al., 2019) have been used to classify concussion using both resting state and task-based EEG signals with varying degrees of success. The main issue with these previous methods is that they all require significant preprocessing of the time series. ...
... Typically, the acquired EEG data are filtered and cleaned to remove these artifacts (c.f. Munia et al., 2017;Hristopulos et al., 2019), with cleaning strategies in use ranging from fully automated algorithms to mixed schemes where some artifacts are removed via automated algorithms and others are manually identified and removed. ...
Article
Full-text available
Artificial neural networks (ANNs) are showing increasing promise as decision support tools in medicine and particularly in neuroscience and neuroimaging. Recently, there has been increasing work on using neural networks to classify individuals with concussion using electroencephalography (EEG) data. However, to date the need for research grade equipment has limited the applications to clinical environments. We recently developed a deep learning long short-term memory (LSTM) based recurrent neural network to classify concussion using raw, resting state data using 64 EEG channels and achieved high accuracy in classifying concussion. Here, we report on our efforts to develop a clinically practical system using a minimal subset of EEG sensors. EEG data from 23 athletes who had suffered a sport-related concussion and 35 non-concussed, control athletes were used for this study. We tested and ranked each of the original 64 channels based on its contribution toward the concussion classification performed by the original LSTM network. The top scoring channels were used to train and test a network with the same architecture as the previously trained network. We found that with only six of the top scoring channels the classifier identified concussions with an accuracy of 94%. These results show that it is possible to classify concussion using raw, resting state data from a small number of EEG sensors, constituting a first step toward developing portable, easy to use EEG systems that can be used in a clinical setting.
... For many individuals who sustain mTBI, this neural dysfunction and its associated cognitive, physical, and behavioral symptoms are transient and resolve within a matter of weeks or months. However, a minority of patients-5-30%-experience persistent post-concussive symptoms (PPCS) beyond the typical recovery period (4)(5)(6) and these symptoms are often associated with residual neural dysfunction (7)(8)(9)(10). ...
Article
Full-text available
A specific variant of neurofeedback therapy (NFT), Live Z-Score Training (LZT), can be configured to not target specific EEG frequencies, networks, or regions of the brain, thereby permitting implicit and flexible modulation of EEG activity. In this exploratory analysis, the relationship between post-LZT changes in EEG activity and self-reported symptom reduction is evaluated in a sample of patients with persistent post-concussive symptoms (PPCS). Penalized regressions were used to identify EEG metrics associated with changes in physical, cognitive, and affective symptoms; the predictive capacity of EEG variables selected by the penalized regressions were subsequently validated using linear regression models. Post-treatment changes in theta/alpha ratio predicted reduction in pain intensity and cognitive symptoms and changes in beta-related power metrics predicted improvements in affective symptoms. No EEG changes were associated with changes in a majority of physical symptoms. These data highlight the potential for NFT to target specific EEG patterns to provide greater treatment precision for PPCS patients. This exploratory analysis is intended to promote the refinement of NFT treatment protocols to improve outcomes for patients with PPCS.
... While dual-task studies with concurrent neural measures for athletes post-SRC are limited, initial studies indicate reduced frontopolar oxygenation and activation during isolated balance or cognitive tasks in those with symptomatic SRC and/ or a history of SRC [43,[62][63][64]. Electroencephalography studies have further revealed associations between bandwidth power and greater postural instability in athletes with SRC [65][66][67][68]. While these data are unique from prior work of dual-task deficits in those symptomatically and cognitively cleared for RTS, this initial work adds to the evidence of neural activity alterations in the frontal lobe post-SRC Fig. 2 SRC may lead to a fundamental loss of motor stability that contributes to the risk of MSK injury. ...
Article
Full-text available
Current best practices to direct recovery after sports-related concussion (SRC) typically require asymptomatic presentation at both rest and during a graduated exercise progression, and cognitive performance resolution. However, this standard of care results in a significantly elevated risk for musculoskeletal (MSK) injury after return-to-sport (RTS). The elevated risk is likely secondary to, in part, residual neurophysiological and dual-task motor stability deficits that remain despite RTS. These deficits present as a loss of autonomous control of gait and posture and an increased need for cognition for motor stability. Thus, the incorporation of strategies that can enhance motor stability and restore autonomous control of gait and posture during SRC recovery and RTS progression may facilitate a reduction of the elevated risk of secondary MSK injury. We provide a theoretical framework for the application of motor learning principles to restore autonomous gait and postural stability after SRC via incorporation, or targeted manipulation, of external focus, enhanced expectations, autonomy support, practice schedule variability, and dual-task strategies during rehabilitation and RTS training.
... While several studies investigated the changes in the cortical activity during the lower-limb motor tasks (Munia, Haider, Schneider, Romanick, & Fazel-Rezai, 2017;Slobounov, Teel, & Newell, 2013;Wittenberg et al., 2017), the current understanding is still inadequate. Not only should we be able to quantify the functional measures using kinematic factors (e.g., trajectory, the COP, limits of stability, etc.) but also be able to complement the underlying neurophysiological mechanism of balance control through different modalities. ...
Article
Full-text available
Traumatic brain injury (TBI) often results in balance impairment, increasing the risk of falls, and the chances of further injuries. However, the underlying neural mechanisms of postural control after TBI are not well understood. To this end, we conducted a pilot study to explore the neural mechanisms of unpredictable balance perturbations in 17 chronic TBI participants and 15 matched healthy controls (HC) using the EEG, MRI, and diffusion tensor imaging (DTI) data. As quantitative measures of the functional integration and segregation of the brain networks during the postural task, we computed the global graph-theoretic network measures (global efficiency and modularity) of brain functional connectivity derived from source-space EEG in different frequency bands. We observed that the TBI group showed a lower balance performance as measured by the center of pressure displacement during the task, and the Berg Balance Scale (BBS). They also showed reduced brain activation and connectivity during the balance task. Furthermore, the decrease in brain network segregation in alpha-band from baseline to task was smaller in TBI than HC. The DTI findings revealed widespread structural damage. In terms of the neural correlates, we observed a distinct role played by different frequency bands: theta-band modularity during the task was negatively correlated with the BBS in the TBI group; lower beta-band network connectivity was associated with the reduction in white matter structural integrity. Our future studies will focus on how postural training will modulate the functional brain networks in TBI.
... Regression-Based techniques were supported for denoising; however, they are limited by the disadvantage of bidirectional contamination. As a solution to the problem of bidirectional contamination, low pass filtering and adaptive filters were offered before applying the regression (Croft & Barry, 2000;Munia, Haider, Schneider, Romanick, & Fazel-Rezai, 2017;Salis et al., 2013;Suchetha & Kumaravel, 2013) our scope was a comparative analysis of the performance of three standard denoising methods like continuous Empirical Mode Decomposition (EMD. However, adaptive filters require defining reference techniques for modeling. ...
Article
Introduction: Several computer-aided diagnosis systems for depression are suggested for use by clinicians to authorize the diagnosis. EEG may be used as an objective analysis tool for identifying depression in the initial stage to avoid it from reaching a severe and permanent state. However, artifact contamination reduces the accuracy in EEG signal processing systems. Methods: This work proposes a novel denoising method based on Empirical Mode Decomposition (EMD) ( with Detrended Fluctuation Analysis (DFA) and wavelet packet transform. Initially, real EEG recordings corresponding to depression patients are decomposed into various mode functions by applying EMD. Then, DFA is used as the mode selection criteria. Further Wavelet Packets Decomposition (WPD)-based evaluation is applied to extract the cleaner signal. Results: Simulations were conducted on real EEG databases for depression to demonstrate the effects of the proposed techniques. To conclude the efficacy of the proposed technique, SNR and MAE were identified. The obtained results indicated improved signal-to-noise ratio and lower values of MAE for the combined EMD-DFA-WPD technique. Additionally, Random Forest and SVM (Support Vector Machine)-based classification revealed the improved accuracy of 98.51% and 98.10% for the proposed denoising technique. Whereas the accuracy of the EMD- DFA is 98.01% and 95.81% and EMD combined with DWT technique equaled 98.0% and 97.21% for the EMD- DFA technique for RF and SVM, respectively, compared to the proposed method. Furthermore, the classification performance for both classifiers was compared with and without denoising to highlight the effects of the proposed technique. Conclusion: Proposed denoising system results in better classification of depressed and healthy individuals resulting in a better diagnosing system. These results can be further analyzed using other approaches as a solution to the mode mixing problem of the EMD approach.
Conference Paper
Traumatic Brain Injury (TBI) is a highly prevalent and serious public health concern. Most cases of TBI are mild in nature, yet some individuals may develop following-up persistent disability. The pathophysiologic causes for those with persistent postconcussive symptoms are most likely multifactorial and the underlying mechanism is not well understood, although it is clear that sleep disturbances feature prominently in those with persistent disability. The sleep electroencephalogram (EEG) provides a direct window into neuronal activity during an otherwise highly stereotyped behavioral state, and represents a promising quantitative measure for TBI diagnosis and prognosis. With the ever-evolving domain of machine learning, deep convolutional neural networks, and the development of better architectures, these approaches hold promise to solve some of the long entrenched challenges of personalized medicine for uses in recommendation systems and/or in health monitoring systems. In particular, advanced EEG analysis to identify putative EEG biomarkers of neurological disease could be highly relevant in the prognostication of mild TBI, an otherwise heterogeneous disorder with a wide range of affected phenotypes and disability levels. In this work, we investigate the use of various machine learning techniques and deep neural network architectures on a cohort of human subjects with sleep EEG recordings from overnight, in-lab, diagnostic polysomnography (PSG). An optimal scheme is explored for the classification of TBI versus non-TBI control subjects. The results were promising with an accuracy of ∼95% in random sampling arrangement and ∼70% in independent validation arrangement when appropriate parameters were used using a small number of subjects (10 mTBI subjects and 9 age- and sex-matched controls). We are thus confident that, with additional data and further studies, we would be able to build a generalized model to detect TBI accurately, not only via attended, in-lab PSG recordings, but also in practical scenarios such as EEG data obtained from simple wearables in daily life.
Chapter
This study aims to discover a possible relationship between electroencephalogram (EEG) signature changes as physiological indicators of one’s current state, and performance on the Vestibular Ocular Motor Screening (VOMS) assessment. A Muse 2 generated a baseline EEG scan for each participant, allowing for the collection of data associated with one’s brain activity. The subjects were then taken through several VOMS domain tests with a continued recording by the device. A comparable analysis was conducted between the participant’s baseline recording and VOMS recording with an intent to identify the consistent correlations in between. In conclusion the findings of this study show potential for characteristic brain activity patterns dependent upon what VOMS domain is being tested. Therefore, when any deviations from those features are observed, the likelihood of the presence of a concussion is much greater.
Article
Full-text available
Objective: The objective of this paper is to review existing literature surrounding the utility of the King-Devick test which is a commonly used sideline assessment tool for sport-related concussions. Methods: A review of the literature was performed using MEDLINE, CINHAL, and SportDiscus databases. The search was performed from the beginning of the record through November 16(th), 2015. Results: This search strategy yielded 27 articles from aforementioned databases. Further searching in The Cochrane Library with King-Devick AND Concuss* search terms yielded one additional article, summing a total of 28 articles. After removal of duplicates and implementation of the inclusion/exclusion criteria, 8 articles for extensively reviewed. Conclusion: This narrative review suggests that the King-Devick test is an efficient sideline assessment tool for sport-related concussions. However, we recommend that the King-Devick should be used as a sideline screening tool, not a concussion diagnosis tool at this time. A proper baseline time including multiple tests may be recommended to negate the learning affect and to have a reliable baseline in which to measure from for future reference. A three second difference appears appropriate to identify the possibility of concussion and to remove an athlete from play. At this time, the athlete should be monitored and further evaluated as symptoms are sometimes delayed. We suggest that further research may be useful to better determine the efficacy of the K-D test in detecting concussions across a broader range of athletes and sports. We also suggest further research may investigate the K-D test a potential return-to-play tool for clinicians and medical personnel.
Conference Paper
Full-text available
EEG contains immense information about the brain activity which cannot be understood completely by visual inspection. Powerful signal processing algorithms in EEG analysis can greatly assist the physicians and neurologists to extract such hidden information. It has been found that EEG being a time-varying and non-stationary signal, can be analyzed by non-linear methods. In this paper we tried to evaluate the non linear features like Approximate Entropy, Sample Entropy and Hurst Exponent in epileptic and normal EEG and has obtained clear discrimination between them. These features extracted during interictal phase EEG are potential parameters for exploring the possibility of epilepsy diagnosis from interictal phase EEG. SVM classifier was implemented for the non linear features extracted from EEG. Classification parameters of classifier based on non linear features were calculated.
Article
Full-text available
Changes in gamma oscillations (20–50 Hz) have been observed in several neurological disorders. However, the relationship between gamma oscillations and cellular pathologies is unclear. Here we show reduced, behaviourally driven gamma oscillations before the onset of plaque formation or cognitive decline in a mouse model of Alzheimer’s disease. Optogenetically driving fast-spiking parvalbumin-positive (FS-PV)-interneurons at gamma (40 Hz), but not other frequencies, reduces levels of amyloid-β (Aβ)1–40 and Aβ 1–42 isoforms. Gene expression profiling revealed induction of genes associated with morphological transformation of microglia, and histological analysis confirmed increased microglia co-localization with Aβ. Subsequently, we designed a non-invasive 40 Hz light-flickering regime that reduced Aβ1–40 and Aβ1–42 levels in the visual cortex of pre-depositing mice and mitigated plaque load in aged, depositing mice. Our findings uncover a previously unappreciated function of gamma rhythms in recruiting both neuronal and glial responses to attenuate Alzheimer’s-disease-associated pathology.
Conference Paper
Sport related mild traumatic brain injury (mTBI), generally known as a concussion, is a worldwide critical public health concern nowadays. Despite growing concern emphasized by scientific research and recent media presentation regarding mTBI and its effect in athletics life, the management, and prevention of mTBI are still not properly done. The evaluation mainly hampered due to the lack of proper knowledge, subjective nature of assessment tools including the fact that the brain functional deficits after mTBI can be mild or hidden. As a result, development of an effective tool for proper management of these mild incidents is a subject of active research. In this paper, to examine the neural substrates following mTBI, an analysis based on electroencephalogram (EEG) from twenty control and twenty concussed athletes is presented. Preliminary results suggest that the concussed athletes have a significant increase in delta, theta and alpha power but a decrease in beta power. We also calculated the power for individual frequencies from 1 Hz to 40 Hz in order to find out the specific frequencies with the highest deficits. The significant deficiencies were found at 1-2 Hz of delta band, 6-7 Hz of theta band, 8-10 Hz of the alpha band, and 16-18 Hz and 24-29 Hz of the beta band. Though there was no significant difference as observed in gamma band, we found the deficit was significant at 34-36 Hz range within the gamma band. The observed deficits at various frequencies demonstrate that even if there is no significant difference in the traditional frequency bands, there may be hidden deficits at some specific frequencies within a frequency band. These preliminary results suggest that the EEG analysis at each unity frequency may be more promising means of identifying the neuronal damage than the traditional frequency band based analysis. Eventually, the proposed analysis can provide an improved approximation to monitor the pathophysiological recovery after a concussion.
Conference Paper
Assessment, treatment, and management of sport-related concussions are a widely recognized public health issue. Although several neuropsychological and motor assessment tools have been developed and implemented for sports teams at various levels and ages, the sensitivity of these tests has yet to be validated with more objective measures to make return-to-play (RTP) decisions more confidently. The present study sought to analyze the residual effect of concussions on a sample of adolescent athletes who sustained one or more previous concussions compared to those who had no concussion history. For this purpose, a wide variety of assessment tools containing both neurocognitive and electroencephalogram (EEG) elements were used. All clinical testing and EEG were repeated at 8 months, 10 months, and 12 months post-injury for both healthy and concussed athletes. The concussed athletes performed poorer than healthy athletes on processing speed and impulse control subtest of neurocognitive test on month 8, but no alterations were marked in terms of visual and postural stability. EEG analysis revealed significant differences in brain activities of concussed athletes through all three intervals. These long-term neurocognitive and EEG deficits found from this ongoing sport-related concussion study suggest that the post-concussion physiological deficits may last longer than the observed clinical recovery.
Article
Purpose Attention-Deficit Hyperactivity Disorder (ADHD) is a neuro-developmental disorder that is characterized by hyperactivity, inattention and abrupt behaviors. This study proposes an approach for distinguishing ADHD children from normal children using their EEG signals when performing a cognitive task. Methods In this study, 30 children with ADHD and 30 age-matched healthy children without neurological disorders underwent electroencephalography (EEG) when performing a task to stimulate their attention. Fractal dimension (FD), approximate entropy and lyapunov exponent were extracted from EEG signals as non-linear features. In order to improve the classification results, double input symmetrical relevance (DISR) and minimum Redundancy Maximum Relevance (mRMR) methods were used to select the best features as inputs to multi-layer perceptron (MLP) neural network. Results As expected, children with ADHD had more delays and were less accurate in doing the cognitive task. Also, the extracted non-linear features revealed that non-linear indices were greater in different regions of the brain of ADHD children compared to healthy children. This could indicate a more chaotic behavior of ADHD children while performing a cognitive task. Finally, the accuracy of 92.28% and 93.65% were achieved using mRMR method and DISR method using MLP, respectively. Conclusions The results of this study demonstrate the ability of the non-linear features to distinguish ADHD children from healthy children.
Article
Visual awareness is a favorable form of consciousness to study neurobiologically. We propose that it takes two forms: a very fast form, linked to iconic memory, that may be difficult to study; and a somewhat slower one involving visual attention and short-term memory. In the slower form an attentional mechanism transiently binds together all those neurons whose activity relates to the relevant features of a single visual object. We suggest this is done by generating coherent semi-synchronous oscillations, probably in the 40-70 Hz range. These oscillations then activate a transient short-term (working) memory. We outfit several lines of experimental work that might advance the understanding of the neural mechanisms involved. The neural basis of very short-term memory especially needs more experimental study.