Network measures

The synhronisation index obtained for the max and min states for eah

ondition and frequeny bandwillbe usedas aweightedadjaeny matrix in-

diatingthedegreeofonnetivitybetweennodesorEEGeletrodes. Thiswill

the temporalstabilityof eah state. The proess is desribed in hapter4.

Figures 6.10 - 6.12 show the resulting onnetivity maps for the max (left

olumn)and min(rightolumn)statesforeahondition andfrequeny band:

the

α

,

β

and

γ

bands respetively. The rows represent the dierent tasks: (from top to bottom) R hand, Left hand, Feet and Relax linked to happy,

sad, neutralandsurprised faestimulirespetively. Theresultingonnetivity

graphsforthe maxandtheminstatesareshownwithonly5%ofthestrongest

onnetions retained for further analysis. This was performed to maintain a

similar average degree distribution between subjets, sine the betweenness

entralityanbeaetedbythedegreeofanetwork[221℄. Asimilarthreshold

was seleted to perform the analysis as detailed in hapter 4. A range of

thresholds from 3 to 10% was examined as they are the most widely used

in the literature. However, those thresholds were not found to signiantly

hange the properties of the examined onnetivity networks.

Following the same riteria than as in previous hapters, the olours and

sizes ofthenodes arebasedontheirdegree, meaningthatabiggerdiameterof

thenodeontainsahighernumberoflinksonnetedtothis node. Similarly,a

reddisholoursofthenodes,signiesmorelinksonnetedtothenode;Bluish,

alowernumberoflinksonnetedtoit. Furthermore,theedge thiknesses are

basedontheweightedvaluesofthesynhrostateindexmatries. Consequently,

thikerlinesonnetingeletrodepairsmeanhighervaluesthanthinneronesin

theadjaenymatrix. Asmentionedinhapter4,thenodessituatedoutsideof

the brain ontour orrespond tothe nodes labelled asTp9and Tp10, situated

by the ears. This is due to the way EEGNET software plots the onnetivity

graphs.

AninterestingobservationfromFigures6.10to6.12isthattherelaxtaskfor

allfrequeny bandsshows, ingeneral,theless loalisedonnetivityompared

to the other taskslinked toan imaginedmovement. Inaddition, the strength

of the edges represented by the thikness of the onnetions between nodes is

lower in relation to the other three stimuli. It may mean that the number of

proessesongoingwithinthebrainwhenthisstimulusispresentedtotheuseris

lower, onsequently the needof speialised informationintegration operations

are less that those needed for a motorimaginary task.

From theonnetivityguresitanalsobenotiedthat thereisaleardif-

ferenebetweenthe maxandminstatesforthe motorrelatedtasks, espeially

in the

α

and

γ

bands. The min state is the one presenting more segmented onnetivity in relation to the max state. This may mean that most of the

Figure 6.10: Brain onnetivity plots from the maximum and minimum number of o-

urrenes synhrostates in the

α

band for all four onditions: R hand movement, L hand movement,Feetmovementandrelax. Theoloursandsizesofthenodesarebasedontheir

degree: biggerdiametersandreddisholoursofthenodemeaningalargernumberoflinks.

Figure 6.11: Brain onnetivity plots from the maximum and minimum number of o-

urrenes synhrostates in the

β

band for all four onditions: R hand movement, L hand movement,Feetmovementandrelax. Theoloursandsizesofthenodesarebasedontheir

Figure 6.12: Brain onnetivity plots from the maximum and minimum number of o-

urrenes synhrostates in the

γ

band for all four onditions: R hand movement, L hand movement,Feetmovementandrelax. Theoloursandsizesofthenodesarebasedontheir

Therefore, minstate quantitativeharaterisation may leadtoa higherpower

of disrimination between tasks. This is aording tothe results found in the

previous hapters. This phenomenon is not as lear in the

β

band where the dierenes between max and min states are less intuitive from a visual point

of view.

It an be said that there are some similarities in the

α

band (gure 6.10) between the min states for the three motor imaginary tasks (the top three

topographies of the right olumn), speially between L hand and both feet

movements. Thisbehaviourannot beobserved inthe otherfrequeny bands.

This may beexplainedby the inuene ofother proessesourring simultan-

eously in the brain that an be frequeny spei [134℄. Therefore, aeting

dierentlyto the range of frequenies observed.

Fromthe aboveonnetivity plots,asmentionedinhapter 4,objetivein-

formationannotbeobtained,onlysomevisualonjeturesandamorequant-

itativeevaluationmehanismisneeded toyieldanysolid onlusions. Forthis

reason,inadditiontothe onnetivitynetworktopographies showed ingures

6.10 to6.12 aset of onnetivity metriswasalsoobtained for eah frequeny

band, ondition and max/minsynhrostates. Fora more exhaustive explana-

tionabout themeaningandformulationofeahoneof thegraphtheorybased

measures please refer to hapter 4. Tables 6.1 and 6.2 show two examples of

the network measures alulated for the

α

and

γ

bands respetively.

Table6.1: Example of two network metris, harateristi path length (CPL) and global

eieny (G.E.), for the max and min synhrostates for the four onditions (R hand, L

hand,Feet,Relax)forthe

α

band.

measures

Rhand Lhand Feet Relax

maxstate minstate maxstate minstate maxstate minstate maxstate minstate

CPL 4.402 4.29 3.515 3.868 4.451 4.50 4.017 4.52 G.E. 151*10 -3 165*10 -3 130*10 -3 129*10 -3 156*10 -3 136*10 -3 114*10 -3 143*10 -3

Table6.2: Example of twonetwork metris,modularity and transitivity, forthe maxand

minsynhrostatesforthefouronditions(Rhand,Lhand,Feet,Relax)forthe

γ

band.

measures

Rhand Lhand Feet Relax

maxstate minstate maxstate minstate maxstate minstate maxstate minstate

Q 61.4*10 -2 71.8*10 -2 63.7*10 -2 70.0*10 -2 66.4*10 -2 71.6*10 -2 70.9*10 -2 70.3*10 -2 T 66.5*10 -2 86.9*10 -2 61.4*10 -2 80.6*10 -2 59.9*10 -2 84.3*10 -2 70.0*10 -2 69.3*10 -2

From table 6.1 it an be seen that the max state for the L hand and Feet

movement tasks present higher GE values and lower CPL than their orres-

for the R hand tasks the max state boasts this feature. Furthermore, despite

some states leading the integration apaity of the brain for the spei task,

allofthemotortaskshaveasimilarvalueofthosefeatures,meaningthatallof

themperform toaomparablelevelof integrationproess. Thisan explained

by the notion that at the end, although dierent stimuli have been shown to

the partiipant, all belong to the same ognitive task. It means that similar

information integration spei to this task and stimuli is performed. This

behaviouralsoexplainsthesimilaritiesbetween theheadplottopographiesfor

the dierent states withina frequeny band.

Also notieable are the higher values of the transitivity in table 6.2 of the

minstateomparedwiththemaxstateforthemotorimaginarytasks(feetand

R and L hands),This is not the ase for the Relaxondition, where minand

max states show similar values for both onnetivity metris. These results

orroborate the previously mentioned regarding the proposition that the min

state an lead the segregated speialisedproessing.

In document Connectivity analysis from EEG phase synchronisation in emotional BCI (Page 148-154)