CHARACTERISATION AND CROSS FREQUENCY

COUPLING (CFC) ANALYSIS IN EEG SIGNALS

NITHIN A
M160124EE

Under the Guidance of

Dr. SUBHA D.P

DEPARTMENT OF ELECTRICAL ENGINEERING

NATIONAL INSTITUTE OF TECHNOLOGY CALICUT

June 29, 2018

1 / 55
OVERVIEW

Introduction
Objective
Methodology
Brain regions and functions
Mean vector length analysis
Kullback-leibler analysis
Mutual information analysis
Conclusion
References

2 / 55
INTRODUCTION

Cross frequency coupling is a technique to study the interaction

between oscillations in different frequency bands.
CFC can be utilized to detect the neuronal oscillations in EEG signals.
CFC basically studies the interaction between low frequency bands of
delta,theta and alpha with high frequency band gamma.
It is associated with cognitive events and task performance.
Increased coupling strength indicates improved neuronal
computation,communication,learning and memory.
Linear or non linear dependency between time series can be studied by
using mutual information.
Cross mutual information(CMI) : Information transfer between two
brain regions.
Auto mutual information(AMI) : Predicting information transfer
within same brain region.
3 / 55
OBJECTIVES

To study the neuronal oscillations taking place in the brain using cross
frequency coupling.
Frequency to be analyzed in the same region of the brain under
different frequency bands.
Measures for the calculation of CFC
Mean vector length (MVL)
Kullback-Leibler modulation index (KL-MI) method
To study the information transmission in the brain using mutual
information.
Cross mutual information (CMI)
Auto mutual information (AMI)

4 / 55
LITERATURE REVIEW

Sl
Author and publication Title Remarks
No.
1. It presents a new method for assessing
Tort, A.B., Komorowski, R., Measuring phase-amplitude
phase amplitude cross frequency coupling
Eichenbaum, H. and Kopell, N. coupling between neuronal
1 2. It also reviews seven other CFC measures.
Journal of neurophysiology, oscillations of different
3. This method is applied on the hippocampal
104(2), pp.1195-1210. (2010) frequencies.
recording of rats.
1.Mutual information provides a measure of
both the linear and nonlinear statistical
dependencies between two time series.
Jeong, J., Gore, J.C. and 2.Cross-mutual information (CMI) is used to
Mutual information analysis
Peterson, B.S., 2001. quantify the information transmitted from
2 of the EEG in patients with
Clinical neurophysiology, one time series to another.
Alzheimer’s disease.
112(5), pp.827-835. 3.Auto mutual information (AMI) in a time series
estimates how much on average the value of the
time series can be predicted from values of the
time series at preceding points.

5 / 55
EEG data acquisition

Records the electrical activity of the brain

Amplitude : 1µV to 100µV
Frequency : Slightly less then 1Hz to over 100Hz.
Control data : 15 male and 12 female.
Patient data : 9 male and 7 female
Recording protocol: eyes closed (EC), eyes Open (EO), mental
arithmetic eyes closed (MAEC) and mental arithmetic eyes open
(MAEO).
Sampling frequency: 400Hz
data length of 120000
Twenty three electrode locations.
The age group: 50 to 80 years.
EEG data of patients and controls are obtained from Sree Chithira
Thirunal Institute of Medical Sciences, Trivandrum, Kerala, India.

6 / 55
CROSS FREQUENCY COUPLING

There are three types of cross frequency coupling :

Phase-amplitude coupling
Phase-phase coupling
Amplitude-amplitude coupling
Phase amplitude coupling (PAC) is usually adopted because of these
reasons :
Behavioural tasks can modulate in PAC
Potentially involved in sensory integration, memory process
The phase of the low frequency delta (.5 to 4Hz) rhythm modulates power
in the high frequency gamma (24 to 100Hz) band.

7 / 55
MEASURING CROSS FREQUENCY COUPLING
Several algorithms available.
Each proper for a particular case.
No single method has been elected as a preferred standard so far.

Sl
Algorithm Limitation
No
The envelope to Sensitive to coupling phase
1 ESC
signal correlation Potentially not capable of calculating coupling intensity
Phase-Locking
2 PLVMI Potentially not capable of calculating coupling intensity
based measure
3 MVL Mean vector length Need long data length
The general linear
4 GLM Potentially not capable of calculating coupling intensity
model measure
Amplitude power
5 APSD Potentially not capable of calculating coupling intensity
spectral density
6 CV Coherence value Potentially not capable of calculating coupling intensity
Kullback-leibler
7 KL-MI based modulation Need long data length
index
Event related phase Potentially not capable of calculating coupling intensity
8 ERPAC
amplitude coupling Only works on event-related datasets
8 / 55
METHODOLOGY
Unfiltered raw signal.

Figure: 1.Unfiltered signal

The low frequency phase is extracted from the raw signal in order to
create the real valued band pass filtered signal xδ (t). (.5-4Hz)

Figure: 2.Delta band filtered signal

9 / 55
Raw signal was filtered from 24-100Hz in order to create the
real-valued band pass filtered gamma signal xγ (t)

Figure: 3.Gamma band filtered signal

Hilbert transform
Applications of the Hilbert transform in the analysis of biomedical
signals are connected with the concepts of the analytic signal,
envelope, instantaneous phase and instantaneous frequency.
R ∞ x (τ )
The Hilbert transform of a signal x̂ (t) = H[x (t)] = 1/π −∞
t −τ
z(t) = x (t) + i x̂ (t)
zi (t)
Where Instantaneous phase angle is φ(t) = <z(t) = arctan
zp
r (t)
The amplitude envelope can be obtained as follows, A(t) = zr2 + zi2
10 / 55
Composite analytic signal
The Composite analytic signal can be formed by combining the
gamma amplitude and phase of the delta signal.
The composite analytic signal is shown below.
Z (t) = AG (t)exp iφδ (t) (1)

Figure: 4.Composite analytic signal

Courtesy : Reference [18]

11 / 55
Mean vector length

The mean vector length is a measure to assess the PAC strength.

The frequencies of interest for this particular study: delta and gamma
frequency bands.
These frequency bands are used to create an analytic signal defined in
the complex plane by AfA e ifp where AfA is the envelope of gamma
oscillation and φfP is the phase of slow oscillation.
Frequency for slow oscillation: Nesting frequency -(0.5-4)Hz
Frequency for the higher amplitude: Nested frequency-(25-100)Hz.
The instantaneous gamma amplitude can be found out by taking the
length of the complex vector measured from center (0,0).
The angle measured from the center for this vector will give the delta
phase.

12 / 55
If there is no PAC, the plot of AfA e iφfp time series in complex plane
will be a uniform circular density distribution symmetric around zero.
If there is modulation of amplitude for some phases, the coupling will
be more showing unsymmetrical distribution around zero, which in
turn will lead to a bump in its polar plot.
Extend of coupling can be done by taking the mean over all points in
the complex plane.
Coupling is high ⇒ Higher mean vector length and vice-versa.
Mean over all points will be almost zero for symmetric distribution,
showing lesser coupling strength.

13 / 55
Kullback-Leibler method

Adaptation of Kullback-Leibler distance.

Estimates the deviation of an empirical amplitude distribution like
function from the uniform distribution.
The phases are binned in complex analytical signal and mean AfA is
computed over each phase bin and is denoted by < AfA > φfp (j)
where j is the phase bin.
The mean amplitude < AfA > φfp is normalized by dividing each bin
value by the sum over the bins.
< AfA > φfp
P(j) = PN (2)
k=1 < AfA > φfp (k)

where N is the number of phase bins. Here N is taken as 10.

The deviation of the amplitude distribution P from the uniform
distribution in a phase amplitude plot characterizes PAC.
14 / 55
Kullback-Leibler distance:-Deviation of P from a uniform distribution.
KL distance of a discrete function P from a distribution Q is,
N
X P(j)
DKL (P, Q) = P(j)log (3)
j=1
Q(j)

Uniform distribution of mean amplitude over the phase indicates lack

of coupling, and gives zero modulation index.
As MI increases, P moves away from Q. modulation index tracks the
intensity of coupling.
Modulation index method is able to examine two frequency ranges
[fp fA ].

15 / 55
MUTUAL INFORMATION

It will provide linear or non linear dependency between two time series.
Two types :
Cross mutual information (CMI)
Auto mutual information (AMI)
Cross mutual information
CMI is used to quantify the information transmitted from one time
series to another.
X
H(X ) = − PX (xi )log2 PX (xi ) (4)
xi

P
H(X |Y ) = − PXY (xi , yj )log2 [PXY (xi , yj )/PY (yj )] (5)
xi ,yj

i.e., H(X |Y ) = H(X , Y ) − H(Y ) (6)

where (7)
P
H(X , Y ) = − PXY (xi , yj )log2 [PXY (xi , yj )] (8)
xi ,yj

16 / 55
While the a priori uncertainty on X is H(X), the a posteriori uncertainty on
X, given a measurement of y, is H(X—Y). Hence the amount that a
measurement of y reduces the uncertainty of x is

IXY = H(X ) − H(X |Y ) (9)

i.e., IXY = H(X ) + H(Y ) − H(X , Y ) (10)
P PXY (xi , yj )
i.e., IXY = PXY (xi , yj )log2 (11)
xi ,yj PX (xi )PY (yj )

Auto mutual information

AMI estimates, on average, the degree to which x (t + τ ) can be
predicted from x(t)
AMI between x(t) and x (t + τ ) is,

X PXXτ (x (t), x (t + τ ))
IXXτ = PXXτ (x (t), x (t + τ ))log2 (12)
PX (x (t))PXτ X (t + τ )
x (t),x (t+τ )

17 / 55
BRAIN REGIONS AND FUNCTIONS
The human brain can be broadly classified to three regions.
Cerebrum
Associated with the main functions including analyzing,processing and
overall coordination.
Mainly divided into four parts: frontal lobe,parietal lobe,temporal lobe
and occipital lobe.
Cerebellum
Known as the compact brain which is almost similar to the cerebrum.
Main function of the cerebellum is motor control,balance and
coordination.
Any injury to cerebellum may cause lack of balance, shaking and slower
movements.
Brain stem
Stem-like part of the base of the brain
Attached to the spinal cord.
Controls the flow of messages between the brain and rest of the body
Controls basic body functions such as breathing, heart rate etc.
18 / 55
Frontal lobe
Largest part of the mammalian brain.
Main function of the frontal lobe is to control the voluntary movements
like walking.
Broca’s area is a main region in the frontal lobe which helps to convert
our thoughts to speech and language.
Responsible for controlling our behavior, personality and emotions.
If the frontal lobe is damaged it will also effect the problem solving
skill, planning judgement, concentration and intelligence.
Parietal lobe
One of the main regions of cerebral cortex which is located behind the
frontal lobe and above the occipital lobe.
Several areas of the parietal lobe is important in language processing.
Sensory informations like temperature , taste and touch are processed
in this region.
Spatial and temporal perceptions are also carried out by parietal lobe.

19 / 55
Temporal lobe
Situated beneath the lateral sulcus on each cerebral hemisphere of the
human brain.
Limbic system structures including olfactory cortex, amygdala, and
hippocampus are situated within this region.
Any damage to temporal lobe may lead to problems with memory,
understanding language and maintaining emotional control.
Occipital lobe
Smallest lobe among these five lobes.
Located in the back side of the cerebral cortex.
Main function of this lobe includes optical-spatial action , image and
color recognition.
If this portion of the brain gets damaged, processing of optical signals
and identification of objects and geometrical shapes gets hindered.

20 / 55
Limbic system : located in the cerebrum itself, which deals with the
emotions, memory and stimulation. It has mainly four regions:
thalamus , hypothalamus , hippocampus and amygdala.
Thalamus
Situated in the fore brain and is having numerous functions.
It is having the duty of processing sensory information as well as
relaying it.
Injury to thalamus might cause trauma.
Hypothalamus
Very small in size and situated at the base of the brain.
Main function is to control hormone level, thirst, emotions, sleep,
homeostasis, temperature control etc.
If the hypothalamus is not working properly, then it is very hard to find
out because it has wide range of functions.
The main causes of dysfunction of hypothalamus are head injuries,
radiation, tumors and genetic problems.

21 / 55
Hippocampus
Situated in the medial temporal lobe of the human brain.
Main function is memory
Associated with both short term and long term memory, spatial
navigation, learning and emotion.
Alzheimer disease will begin from hippocampal area.
Amygdala
Positioned in the temporal lobe of human brain.
The main functions of the amygdala are hormonal secretion , memory
emotional response and arousal.

22 / 55
MEAN VECTOR LENGTH ANALYSIS

Eyes closed (EC) condition

0.04
Patient
0.035 Control

0.03

Coupling Strength 0.025

0.02

0.015

0.01

0.005

0
Fp1Fp2F3 F4 C3 C4 P3 P4 O1 O2 T1 T2 F7 F8 T3 T4 T5 T6 Fz Cz Pz A1 A2
Electrode Location

Figure: 5.PAC strength for EC

1 Significant difference in coupling strength is observed in F3 , C4 , T2 ,

T5 , A1 and A2.
2 It is seen that frontal temporal and central lobes are affected.
3 Clinical studies show that mild cognitive impairment will affect these
regions.
4 Proven using PAC under CFC studies.
23 / 55
Eyes open (EO) condition
0.04
Patient
0.035 Control

0.03

Coupling Strength
0.025

0.02

0.015

0.01

0.005

0
Fp1Fp2F3 F4 C3 C4 P3 P4 O1 O2 T1 T2 F7 F8 T3 T4 T5 T6 Fz Cz Pz A1 A2
Electrode Location

Figure: 6.PAC strength for EO

1 Significant difference in coupling strength is noted in F3,C4,T2,T6,A1

and A2 locations.
2 The left part of the frontal region shows less PAC strength for patients.
3 The right part of the temporal region and central region is also showing
less PAC strength for patients.

24 / 55
Mental arithmetic eyes closed (MAEC) task
0.04
Patient
0.035 Control

0.03

Coupling Strength
0.025

0.02

0.015

0.01

0.005

0
Fp1Fp2F3 F4 C3 C4 P3 P4 O1 O2 T1 T2 F7 F8 T3 T4 T5 T6 Fz Cz Pz A1 A2
Electrode Location

Figure: 7.PAC strength for MAEC

1 Most of the electrode locations are affected and significant variations

are seen in F3 , C4 , T1 , T2 , F7 , T4 , T5 , T6 and A1.
2 Left part of the frontal lobe (electrodes placed at F3 and F7) is
showing notable difference in coupling strength.
3 Under cognitive task condition whole of the temporal lobe is also
affected.
4 Frontal lobe which is responsible for intelligence concentration and
problem solving skills.
5 Temporal lobe is responsible for memory and understanding language.
25 / 55
Mental arithmetic eyes open (MAEO) task
0.04
Patient
0.035 Control

0.03

Coupling Strength
0.025

0.02

0.015

0.01

0.005

0
Fp1Fp2F3 F4 C3 C4 P3 P4 O1 O2 T1 T2 F7 F8 T3 T4 T5 T6 Fz Cz Pz A1 A2
Electrode Location

Figure: 8.PAC strength for MAEO

1 Whole of the frontal lobe (F3,F7,F4 and F8) is affected.

2 Left side of parietal and occipital lobes also shows variation in PAC
strength at electrode locations P3 and O1 respectively.
3 Parietal and occipital lobes are associated with perception.
4 The temporal lobe situated on the left side of the cerebral hemisphere
is affected in regions T1 and T5 and on the right side of hemisphere in
regions T2 , T4 and T6

26 / 55
KULLBACK-LEIBLER ANALYSIS

Eyes closed condition

0.05
Patient
0.045 Control

0.04

Coupling Strength 0.035

0.03

0.025

0.02

0.015

0.01

0.005

0
Fp1Fp2F3 F4 C3 C4 P3 P4 O1 O2 T1 T2 F7 F8 T3 T4 T5 T6 Fz Cz Pz A1 A2
Electrode Location

Figure: 9.PAC strength for EC

1 Temporal regions in both, right and left hemispheres (T1,T2,T5 ) also

show considerable variation in Phase amplitude coupling strength.
2 Frontal lobe(F3) in the left hemisphere is affected.
3 The delta-gamma coupling is mostly affected in temporal lobe and
frontal lobe in patients with mild Alzheimer’s.

27 / 55
Eyes open condition
0.05
Patient
0.045 Control

0.04

0.035

Coupling Strength
0.03

0.025

0.02

0.015

0.01

0.005

0
Fp1Fp2F3 F4 C3 C4 P3 P4 O1 O2 T1 T2 F7 F8 T3 T4 T5 T6 Fz Cz Pz A1 A2
Electrode Location

Figure: 10.PAC strength for EO

1 PAC strength is more in fronto-parietal (Fp2) and frontal region in the

left part of the cerebral cortex (F3)
2 Similar observations are made in central line of the brain (C4), parietal
lobe located in the left part of the brain (P3) and temporal lobe
located in the right side of the cerebral cortex (T2,T6).
3 Similar results are obtained in MVL.

28 / 55
Mental arithmetic eyes closed (MAEC) task
0.05
Patient
0.045 Control

0.04

0.035

Coupling Strength
0.03

0.025

0.02

0.015

0.01

0.005

0
Fp1Fp2F3 F4 C3 C4 P3 P4 O1 O2 T1 T2 F7 F8 T3 T4 T5 T6 Fz Cz Pz A1 A2
Electrode Location

Figure: 11.PAC strength for MAEC

1 Considerable difference in PAC is viewed in frontal lobe (F3) located in

the left side of the cerebral cortex.
2 Temporal lobe located in both sides of the cerebral hemisphere is
affected (T1,T2,T3,T4,T5 and T6).

29 / 55
Mental arithmetic eyes open (MAEO) task
0.05
Patient
0.045 Control

0.04

0.035

Coupling Strength
0.03

0.025

0.02

0.015

0.01

0.005

0
Fp1Fp2F3 F4 C3 C4 P3 P4 O1 O2 T1 T2 F7 F8 T3 T4 T5 T6 Fz Cz Pz A1 A2
Electrode Location

Figure: 12.PAC strength for MAEO

1 Significant variation is seen in right and left half of the frontal lobe
(F3,F7,F4,F8) which is having important role in
concentration,intelligence and problem solving skill.
2 Temporal region (T1,T5,T2,T4) is also highly affected that is, the
temporal lobes situated in the right and left part of the brain is
affected.
3 The left part of the parietal lobe (P3) is also slightly affected and left
part of the occipital region is also affected as these regions are
associated with vision.
30 / 55
MVL vs KL-MI

Mean vector length

EC EO MAEC MAEO
P C P C P C P C
Max. PAC 0.025 0.035 0.024 0.025 0.024 0.026 0.024 0.027
Flow 2.11 2.11 2.11 2.11 0.5 0.5 1.3 1.3
Fhigh 35.52 35.52 35.52 35.52 28.62 28.62 24.89 24.89

KL-MI Method
EC EO MAEC MAEO
P C P C P C P C
Max. PAC 0.039 0.043 0.034 0.037 0.026 0.029 0.028 0.039
Flow 2.5 2.5 2.5 2.5 0.5 0.5 1.93 1.93
Fhigh 40 40 32 32 58 58 47 47

31 / 55
1 The result obtained in our analysis is matching with the clinical
studies.
2 PAC strength is lesser in frontal and temporal lobes.
3 The neurons in frontal and temporal lobes are damaged, that is
neuronal death occurs.
4 Learning,communication and computation are affected.
5 The range of variation of PAC strength is larger in KL-MI method
than MVL method.
6 KL-MI method is more sensitive than MVL.

32 / 55
Resting condition vs cognitive task
Mean vector length
Ele.Location EC MAEC EC P MAEC P EO MAEO EO P MAEO P
Fp1 0.0191 0.0194 0.0197 0.0202 0.0203 0.0208 0.0202 0.0231
Fp2 0.0216 0.0225 0.0223 0.0223 0.0222 0.0238 0.0224 0.024
F3 0.0230 0.0232 0.0206 0.0209 0.0234 0.0244 0.0211 0.0232
F4 0.0226 0.0246 0.0240 0.024 0.023 0.0262 0.024 0.023
C3 0.0190 0.0195 0.0187 0.0187 0.0193 0.0195 0.0187 0.0193
C4 0.0236 0.0232 0.0198 0.0204 0.023 0.023 0.0187 0.0206
P3 0.0207 0.021 0.0194 0.0194 0.0214 0.0211 0.0188 0.0192
P4 0.0215 0.0222 0.0225 0.0232 0.021 0.0216 0.0215 0.0220
O1 0.0191 0.0187 0.0167 0.0181 0.018 0.0194 0.0164 0.0168
O2 0.0215 0.0216 0.0216 0.0223 0.022 0.0215 0.0217 0.0213
T1 0.0219 0.0224 0.0214 0.019 0.0208 0.0216 0.0215 0.0198
T2 0.0224 0.0220 0.0189 0.0193 0.022 0.023 0.0192 0.0208
F7 0.0197 0.0207 0.0191 0.0193 0.0204 0.0224 0.019 0.019
F8 0.0213 0.0233 0.0210 0.0221 0.0213 0.0243 0.0206 0.0211
T3 0.018 0.0186 0.0177 0.0181 0.0187 0.0184 0.0178 0.0177
T4 0.0206 0.0224 0.0202 0.0205 0.0206 0.024 0.02 0.0203
T5 0.0357 0.0261 0.026 0.0241 0.025 0.027 0.0246 0.0243
T6 0.0224 0.023 0.0205 0.0211 0.0229 0.023 0.0208 0.0211
Fz 0.0211 0.0224 0.020 0.0212 0.0214 0.0225 0.0201 0.0218
Cz 0.0230 0.0234 0.0224 0.0218 0.023 0.0237 0.0223 0.0220
Pz 0.0183 0.0182 0.0193 0.0192 0.0186 0.0192 0.0191 0.0188
A1 0.0238 0.026 0.0198 0.0187 0.0215 0.0234 0.0186 0.0192
A2 0.0221 0.0220 0.018 0.0193 0.021 0.0230 0.0184 33 / 55
0.0193
1 Controls are showing higher PAC strength for eyes closed cognitive
task as compared to eyes closed resting condition except regions
C4,T2,T5 and A1.
2 Patients are showing Lesser PAC strength for eyes closed cognitive
task compared with eyes closed resting task in locations T1,T5,Cz,Pz
and A1.
3 Controls are showing higher PAC strength for eyes open cognitive task
as compared to eyes open resting condition except regions P3 and T3.
4 Patients are showing Lesser PAC strength value for eyes open
cognitive task compared with eyes open resting condition in locations
F4,T1,T5,Cz, and Pz.

34 / 55
 Table: Resting condition Vs Cognitive task using KL-MI

KL-MI
Ele.location EC MAEC EC P MAEC P EO MAEO EO P MAEO P
Fp1 0.0218 0.0269 0.0194 0.0233 0.0212 0.0243 0.0205 0.0201
Fp2 0.0232 0.0258 0.0244 0.0231 0.03 0.0196 0.0153 0.0376
F3 0.024 0.0376 0.018 0.0211 0.0192 0.0227 0.0177 0.016
F4 0.0269 0.0255 0.0212 0.0231 0.0255 0.0192 0.0174 0.0255
C3 0.0233 0.0244 0.0197 0.0227 0.0209 0.02 0.0181 0.0195
C4 0.0272 0.0277 0.023 0.0269 0.028 0.0277 0.0194 0.0268
P3 0.0189 0.023 0.0201 0.0245 0.0258 0.0217 0.0155 0.0236
P4 0.0196 0.0239 0.023 0.0211 0.0214 0.0229 0.0196 0.021
O1 0.0246 0.0238 0.0182 0.0218 0.0217 0.0243 0.017 0.0226
O2 0.0205 0.026 0.0185 0.0205 0.0198 0.0226 0.0171 0.0198
T1 0.0205 0.0416 0.0194 0.0275 0.0211 0.0215 0.01575 0.0174
T2 0.0294 0.0436 0.0263 0.0391 0.0369 0.0374 0.0291 0.0354
F7 0.01593 0.0266 0.0144 0.0315 0.0169 0.0091 0.0076 0.0174
F8 0.022 0.0322 0.0181 0.0269 0.0171 0.0173 0.0141 0.0169
T3 0.0165 0.0307 0.0132 0.035 0.0087 0.0173 0.0095 0.0152
T4 0.0244 0.0239 0.0224 0.0219 0.0199 0.0223 0.0174 0.0212
T5 0.015 0.0147 0.0172 0.0125 0.01 0.0173 0.0231 0.0198
T6 0.018 0.0171 0.0176 0.0152 0.0153 0.0177 0.0147 0.0164
Fz 0.0162 0.0196 0.0159 0.0183 0.012 0.0137 0.0129 0.0128
Cz 0.0173 0.0218 0.0164 0.0201 0.0173 0.0181 0.0136 0.0186
Pz 0.0107 0.0158 0.0124 0.0158 0.0082 0.0103 0.008 0.0118
A1 0.0222 0.0318 0.019 0.0212 0.013 0.0243 0.0111 0.0215
A2 0.0139 0.0325 0.0134 0.0291 0.0093 0.0146 0.0087 35 / 55
0.0132
1 Controls are showing higher PAC strength for eyes closed cognitive
task as compared to eyes closed resting condition except regions
F3,F4,O1,T4 and T6.
2 Patients are showing Lesser PAC strength for eyes closed cognitive
task compared with eyes closed resting task in locations
Fp2,P4,T3,T4,T5 and T6.
3 Controls are showing higher PAC strength for eyes open cognitive task
as compared to eyes open resting condition except regions
Fp2,F4,P3,F7.
4 Patients are showing Lesser PAC strength for eyes open cognitive task
compared with eyes open resting condition in locations Fp1,F3,T5.

36 / 55
 CMI CMI
F F
p p
1 1

0
2
4
6
8
0
2
4
6
8

10
12
14
16
10
12
14
16

- -
F F7 F F7
p p
2 2
- -
F F8 F F8
p p
1 1
- -
F F3 F F3
p p
2 2
-F -F
P 4 P 4
3 3
-T -T
P 3 P 3
2 2
-P -P
P 4 P 4
3 3
Cross mutual information

-T -T
P 5 P 5
2 2
- -
P T6 P T6
3 3
-F -F
P p1 P p1
4 4
-F -F
p p
P 2 P 2
3 3
-F -F
P 7 P 7
4 4
-F -F
P 8 P 8
3 3
Patient

-F -F
Control

Electrode Location
Electrode Location

P 3 P 3
4 4
- -
MUTUAL INFORMATION ANALYSIS

T F4 T F4
5 5
-F -F
Patient
Control

T p1 T p1
6 6
Figure: 13.CMI Variation For EC

Figure: 14.CMI Variation For EO

-F -F
p p
T 2 T 2
5 5
-F -F
T 7 T 7
6 6
-F -F
T 8 T 8
5 5
-F -F
T 3 T 3
6 6
-F -F
4 4
37 / 55
 16
Patient
Control
14

12

10

CMI
8

0
F 7

F F8

F F3

P 4

T p1

T 2
P p1

P 2

T 7

T 8

T 3
4
P 3

P 5
P T6

P 7

P 8

P 3
T F4
P 4
-F

-F

p
p

-F

-F

-F

-F
-T

-T

-F

-F

-F
-P

-F

-F
-F

-F
-

-
1

6
3

4
2
p

6
3

4
F

Electrode Location

Figure: 15.CMI Variation For MAEC

1 Similar observations has been seen in EC,EO and MAEC tasks.

2 Information transfer in controls is more than that of patients.
3 Maximum information transfer is seen in frontal-parietal and
parietal-temporal.
4 Information transfer doesn’t effect frontoparietal-frontal and
frontoparietal-Temporal

38 / 55
 16
Patient
Control
14

12

10
CMI
8

0
F 7

F F8

F F3

P 4

T p1

T 2
P p1

P 2

T 7

T 8

T 3
4
P 3

P 5
P T6

P 7

P 8

P 3
T F4
P 4
-F

-F

p
p

-F

-F

-F

-F
-T

-T

-F

-F

-F
-P

-F

-F
-F

-F
-

-
1

6
3

4
2
p

6
3

4
F

Electrode Location

Figure: 16.CMI Variation For MAEO

1 Mutual information of patient’s are greater in P3-F3 and P4-F4.

2 Mutual information of patients are slightly greater in region where the
frontoparietal is involved.

39 / 55
Resting condition vs cognitive task

Cross mutual information

Ele.Location EC MAEC EC P MAEC P EO MAEO EO P MAEO P
Fp1-F7 12.75921 12.9212 11.86635 11.6659 12.7854 13.35202 11.32866 11.75101
Fp2-F8 12.82423 12.96507 11.84183 11.64493 12.8511 13.41334 11.32734 11.91604
Fp1-F3 13.18328 13.18411 12.92863 12.64412 13.1581 12.87337 12.63524 13.49328
Fp2-F4 13.18854 13.29863 13.26371 13.00718 13.2061 12.90942 12.85942 13.69508
P3-T3 13.1748 13.14385 12.89185 12.71103 13.1032 12.81938 12.78219 13.61589
P2-P4 12.74077 12.84126 11.38652 11.66034 12.6878 12.81299 11.06332 11.77115
P3-T5 12.76589 12.94243 11.84851 11.70219 12.7362 12.95382 11.50011 11.89221
P2-T6 14.68816 14.94903 12.85854 13.14432 14.6822 15.33306 12.61115 12.83185
P3-Fp1 13.17552 13.34498 13.29529 12.98927 13.1638 12.51705 12.84928 13.54399
P4-Fp2 13.20856 13.34338 13.25113 12.9408 13.2108 12.92937 12.8432 13.59421
P3-F7 12.75808 12.8884 11.85404 11.72099 12.7408 12.91041 11.49889 11.89253
P4-F8 12.79462 12.91627 11.81839 11.70871 12.7754 12.9304 11.50333 11.82058
P3-F3 13.1748 13.14385 12.89185 12.71103 13.1032 12.81938 12.78219 13.61589
P4-F4 13.16428 13.24853 13.28255 13.05745 13.1211 12.82083 13.07609 13.63302
T5-Fp1 12.76435 12.95791 11.88497 11.65005 12.7731 13.36078 11.33615 11.28634
T6-Fp2 12.81092 12.98855 11.86929 11.59539 12.8223 13.42925 11.33667 11.89059
T5-F7 14.6655 14.97185 13.31123 13.18523 14.7327 15.4122 13.04469 12.34325
T6-F8 14.71382 15.04026 13.29207 13.22149 14.7884 15.4769 13.06194 12.85032
T5-F3 12.75523 12.74832 11.46075 11.33735 12.7076 12.88748 11.25326 11.34193
T6-F4 12.74689 12.87859 11.81997 11.68411 12.7266 12.91464 11.50093 11.85464

40 / 55
1 For controls the value of cross mutual information is higher for eyes
closed cognitive tasks except P3-T3,P3-Fp1,P3-F3.
2 For patients the value of CMI is lesser for eyes closed cognitive tasks
compared with resting state except P2-P4,P2-T6,P3-F3.
3 For controls the value of cross mutual information is higher for eyes
open cognitive tasks except
Fp1-F3,Fp2-F4,P3-T3,P3-Fp1,P4-Fp2,P3-F3,P4-F4.
4 For patients the value of CMI is greater for eyes open cognitive tasks
compared with resting state except T5-Fp1,T5-F7,T6-F8.
5 It mainly affect on frontal and parietal lobes.

41 / 55
Auto mutual information

0.085
Patient
Control
0.08

0.075
AMI
0.07

0.065

0.06

0.055
Fp1
Fp2F3 F4 C3C4 P3 P4O1O2 T1 T2 F7 F8 T3 T4 T5 T6 Fz Cz Pz A1 A2
Electrode Location

Figure: 17.AMI Variation For EC

0.085
Patient
Control
0.08

0.075
AMI

0.07

0.065

0.06

0.055
Fp1
Fp2F3 F4 C3C4 P3 P4O1O2 T1 T2 F7 F8 T3 T4 T5 T6 Fz Cz Pz A1 A2
Electrode Location

Figure: 18.AMI Variation For EO

42 / 55
 0.085
Patient
Control
0.08

0.075

AMI
0.07

0.065

0.06

0.055
Fp1
Fp2F3 F4 C3C4 P3 P4O1O2 T1 T2 F7 F8 T3 T4 T5 T6 Fz Cz Pz A1 A2
Electrode Location

Figure: 19.AMI Variation For MAEC

0.085
Patient
Control
0.08

0.075
AMI

0.07

0.065

0.06

0.055
Fp1
Fp2F3 F4 C3C4 P3 P4O1O2 T1 T2 F7 F8 T3 T4 T5 T6 Fz Cz Pz A1 A2
Electrode Location

Figure: 20.AMI Variation For MAEO

43 / 55
1 AMI for all tasks are showing similar variation.
2 AMI is very less in the right part of the cerebral hemicortex compared
to the left part.
3 In EC,EO and MAEC the T1 region is showing some variation for
patients compared to controls.
4 In MAEO T1,T5 and F7 is showing a considerable difference in AMI.

44 / 55
Resting condition vs cognitive task

Auto mutual information

Ele.Loc. EC MAEC EC P MAEC P EO MAEO EO P MAEO P
Fp1 0.060691 0.061365 0.0667 0.064786 0.060535 0.061026 0.066136 0.065821
Fp2 0.0817 0.082126 0.080167 0.078314 0.081239 0.079987 0.079414 0.07975
F3 0.056791 0.057209 0.061717 0.059729 0.056739 0.055457 0.061107 0.060107
F4 0.07867 0.079143 0.078592 0.077286 0.078626 0.077361 0.078679 0.078593
C3 0.059613 0.059978 0.064192 0.062207 0.059474 0.058187 0.06375 0.063129
C4 0.081026 0.081617 0.082083 0.0805 0.080861 0.079643 0.082157 0.082007
P3 0.060039 0.060265 0.063292 0.062086 0.060039 0.0591 0.063271 0.062879
P4 0.081222 0.080078 0.080242 0.078621 0.079496 0.0784 0.080221 0.079907
O1 0.063087 0.063422 0.067517 0.069729 0.0629 0.061761 0.067129 0.06665
O2 0.0792 0.079635 0.078442 0.076507 0.079048 0.077787 0.077943 0.078036
T1 0.06027 0.060417 0.067625 0.066421 0.060374 0.059183 0.067364 0.069514
T2 0.078022 0.078435 0.078825 0.077493 0.077922 0.076565 0.078671 0.078736
F7 0.062213 0.06237 0.067733 0.065879 0.061861 0.060587 0.066679 0.066593
F8 0.081404 0.081578 0.079475 0.07795 0.081017 0.07897 0.079271 0.079279
T3 0.062457 0.063304 0.065158 0.063 0.061487 0.061409 0.063986 0.063929
T4 0.080109 0.08163 0.077383 0.076029 0.0796 0.078483 0.077293 0.077407
T5 0.06007 0.06147 0.063725 0.061943 0.060022 0.0594 0.063107 0.067143
T6 0.081443 0.082922 0.079633 0.077871 0.081265 0.080204 0.079286 0.079286
Fz 0.059278 0.059796 0.06415 0.062386 0.059278 0.058491 0.065693 0.063443
Cz 0.07893 0.079978 0.077083 0.0756 0.078826 0.077857 0.079214 0.076936
Pz 0.061387 0.061804 0.065108 0.063293 0.061283 0.060239 0.066779 0.064264
A1 0.079478 0.079887 0.076858 0.075186 0.079174 0.077704 0.07635 0.076421
A2 0.061339 0.0603 0.062592 0.061207 0.059755 0.058783 0.061957 0.062579

45 / 55
1 For controls AMI value is greater in cognitive condition compared
with resting state in almost all electrode locations.
2 AMI value is lesser for the patients in cognitive condition compared
with resting condition almost all electrode locations.

46 / 55
CONCLUSION

CFC analysis has been conducted on twenty seven controls and fifteen
patients.
Protocol: Eyes closed,eyes open, mental arithmetic eyes closed and
mental arithmetic eyes open.
Frequency band: Delta and Gamma.
Electrode locations : 23
Measures used : MVL and KLMI
Both measures are giving almost same result.
In resting state some parts of frontal and temporal lobes are affected.
Under cognitive task condition complete temporal and frontal lobes
are affected

47 / 55
Temporal lobe and frontal lobes are mostly affected. Because the
limbic system is located in this region.
Frontal lobe is associated with intelligence, planning and judging.
Kullback-Leibler method is more sensitive than Mean Vector Length.
Slight variation in coupling strength can be identified using
Kullback-Leibler method.
Cognitive task condition is showing significant variation in PAC
strength for patients.

48 / 55
In cross mutual information analysis the communication is higher in
frontal-parietal and parietal-temporal regions.
EC,EO and MAEC tasks are showing almost same variation for CMI.
MAEO showing considerable variation of CMI for patients in
parietal-frontal.
CMI doesn’t affect frontoparietal-frontal and frontoparietal-temporal.
AMI is very less in the right part of the cerebral hemicortex.
Mild Alzheimer slightly affects temporal region for EC,EO,MAEC
tasks.
It slightly affects temporal and frontal region for MAEO task.

49 / 55
Contribution

KL-MI and MVL in human beings.

Introduction of the Kullback-Leibler method in Alzheimer disease.
Study between the resting state and cognitive state in cross frequency
coupling.
Cross mutual information transfer and auto mutual information
transfer on resting condition and cognitive task.

50 / 55
Future scope

Analysis can be conducted on different location same frequency.

Cross frequency coupling studies can be extended to patients with
moderate or severe cognitive impairment.
Mutual information studies can be extended to moderate or severe
Alzheimer patients.

51 / 55
REFERENCE
Binetti, Giuliano, E. Magni, A. Padovani, S. F. Cappa, A. Bianchetti,
and M. Trabucchi. ”Executive dysfunction in early Alzheimer’s
disease.” Journal of Neurology, Neurosurgery & Psychiatry 60, no. 1
(1996): 91-93.
Besthorn, C., R. Zerfass, C. Geiger-Kabisch, H. Sattel, S. Daniel, U.
Schreiter-Gasser, and H. FÃűrstl.”Discrimination of Alzheimer’s disease
and normal aging by EEG data.” Electroencephalography and Clinical
Neurophysiology 103, no. 2 (1997): 241-248.
Pucci, E., N. Belardinelli, G. Cacchio, M. Signorino, and F. Angeleri.
”EEG power spectrum differences in early and late onset forms of
Alzheimer’s disease.” Clinical Neurophysiology 110, no. 4 (1999):
621-631.
Jelles, B., J. H. Van Birgelen, J. P. J. Slaets, R. E. M. Hekster, E. J.
Jonkman, and C. J. Stam. ”Decrease of non-linear structure in the
EEG of Alzheimer patients compared to healthy controls.” Clinical
Neurophysiology 110, no. 7 (1999): 1159-1167. 52 / 55
Chan, Dennis, Nick C. Fox, Rachael I. Scahill, William R. Crum,
Jennifer L. Whitwell, Guy Leschziner, Alex M. Rossor, John M.
Stevens, Lisa Cipolotti, and Martin N. Rossor. ”Patterns of temporal
lobe atrophy in semantic dementia and Alzheimer’s disease.” Annals of
neurology 49, no. 4 (2001): 433-442.
Pijnenburg, Y. A. L., Y. Vd Made, AM Van Cappellen Van Walsum,
D. L. Knol, Ph Scheltens, and C. J. Stam. ”EEG synchronization
likelihood in mild cognitive impairment and Alzheimer’s disease during
a working memory task.” Clinical neurophysiology 115, no. 6 (2004):
1332-1339.
Latchoumane, Charles-Francois Vincent, Emmanuel Ifeachor, Nigel
Hudson, Sunil Wimalaratna, and Jaeseung Jeong. ”Dynamical
nonstationarity analysis of resting EEGs in Alzheimer’s disease.” In
International Conference on Neural Information Processing, pp.
921-929. Springer, Berlin, Heidelberg, 2007.
Yang, Albert C., Shuu-Jiun Wang, Kuan-Lin Lai, Chia-Fen Tsai,
Cheng-Hung Yang, Jen-Ping Hwang, Men-Tzung Lo, Norden E.
Huang, Chung-Kang Peng, and Jong-Ling Fuh. ”Cognitive and 53 / 55
Canolty, Ryan T., Erik Edwards, Sarang S. Dalal, Maryam Soltani,
Srikantan S. Nagarajan, Heidi E. Kirsch, Mitchel S. Berger, Nicholas
M. Barbaro, and Robert T. Knight. . ”High gamma power is
phase-locked to theta oscillations in human neocortex.”science 313
,no. 5793 (2006): 1626-1628.
Tort, Adriano BL, Robert Komorowski, Howard Eichenbaum, and
Nancy Kopell. ”Measuring phase-amplitude coupling between neuronal
oscillations of different frequencies.” Journal of neurophysiology , 104,
no. 2 (2010): 1195-1210.
Cohen, Michael X. ”Assessing transient cross-frequency coupling in
EEG data.” Journal of neuroscience methods 168, no. 2 (2008):
494-499.
Penny, W. D., E. Duzel, K. J. Miller, and J. G. Ojemann. ”Testing for
nested oscillation.” in Journal of neuroscience methods, 174, no. 1
(2008): 50-61.
Canolty, Ryan T., and Robert T. Knight, ”The functional role of
cross-frequency coupling.,” in Trends in cognitive sciences, 14, no. 11
(2010): 506-515 54 / 55
Tort, Adriano BL, Robert W. Komorowski, Joseph R. Manns, Nancy J.
Kopell, and Howard Eichenbaum. ”Theta-gamma coupling increases
during the learning of item-context associations.” Proceedings of the
National Academy of Sciences . 106, no. 49 (2009): 20942-20947.
Tort, Adriano BL, Mark A. Kramer, Catherine Thorn, Daniel J.
Gibson, Yasuo Kubota, Ann M. Graybiel, and Nancy J. Kopell.
”Dynamic cross-frequency couplings of local field potential oscillations
in rat striatum and hippocampus during performance of a T-maze
task.” in Proceedings of the National Academy of Sciences , 105, no.
51 (2008): 20517-20522.
Belluscio, Mariano A., Kenji Mizuseki, Robert Schmidt, Richard
Kempter, and GyÃűrgy BuzsÃąki. ”Cross-frequency phase-phase
coupling between theta and gamma oscillations in the hippocampus.”
in Journal of Neuroscience , 32, no. 2 (2012): 423-435.
Samiee, S. and Baillet, S. (2017). ”Time-resolved phase-amplitude
coupling in neural oscillations.”NeuroImage,, 159, pp.270-279.
55 / 55