Stimulated electrodeRecording electrode

Conclusions

Depts of 1Neurosurgery and 2Neurology , Hofstra University School of Medicine, North Shore LIJ Health System, New Hyde Park, NY3Feinstein Institute for Medical Research, Manhassat, NY

4Albert Einstein College of Medicine, Bronx, NY, 5National Institute of Neuroscience, Budapest, Hungary

A.D. Mehta1,3, L. Entz1,3,5, S. Bickel1,3,4, C.J. Keller1,4, D. Groppe1,3, S. Jain2, I. Ulbert5

�e Relationship of Resting State fMRI Correlation and Anticorrelation to Electrically-Evoked Potentials in the Human Brain

IntroductionFunctional MRI studies carried out during rest (R-fMRI) suggest a functional architecture of spatially distributed networks that is represented in low-frequency(<0.1Hz) spontaneous �uctuations of the blood oxygen level-dependent (BOLD)signal. Functional connectivity analysis of this signal reveals both correlated and anti-correlated areas. In order to examine the neurophysiological underpinnings related to the direction of correlation, we investigated the relationship of these correlations to electrophysiological measures. Studies were carried out using direct electrical stimulation of the cerebral cortex with intracranial electrodes in patients undergoing invasive electrode implantation for seizure monitoring.

MethodsWe performed systematic bipolar stimulation of all electrodes by administering a single pulse electrical current (10mA, 0.5Hz, 0.2 msec pulse width, 20 trials per electrode pair) on six patients undergoing intracranial monitoring for intractable epilepsy. Electrodes were localized using post-operative CT and MRI and aligned upon the reconstructed cortical surface of a pre-operative MRI scan. Resting state fMRI scans (3T) were performed prior to electrode implantation.

Probing Sensorimotor Areas

We would like to acknowledge Michael Milham, Clare Kelly, G. Klein, A. Dykstra, and F. Lado for their assistance with this work. �is work was supported by the Page and Otto Marx Jr Foundation and the Epilepsy Foundation of America.

Spectral analysis of CCEP and RSFCResults

Correlated �uctuations of the BOLD signal at rest reliably predict electrically-evoked potentialsPositively correlated regions exhibit higher CCEP power in all frequencies during the N1 andlow frequencies (<11Hz) during N2Anticorrelated regions exhibit higher CCEP power in high frequencies (>25Hz) during N1 / N2Further investigation is warranted with regard to the spatial and temporal relationship betweenCCEPs, RSFC, and spontaneous ECoG

CCEPs were thresholded (Z>6 above baseline during N2) to determine signi�canceResting state correlation values (RSFC)within voxels underlying electrodes were extracted and averagedMatrices from each methodwere analyzed

-1 -.3 +.3 +1-1 -.3 +.3 +1

-1 -.3 +.3 +1

No signi�cant di�erence between CCEPwaveforms whose region had negative RSFC or non-signi�cant RSFC

Negative RSFCPositive RSFCNon−significant RSFC

Negative RSFC

Nonsignificant RSFC

Positive RSFC

0

10

20

30

40

50

CC

EP

z-s

core

ANOVA

0

5

10

15Negative RSFCPositive RSFC

Mea

n C

CE

P

All >10 >25 >50 >75 >90 >95th percentile

Mea

n C

CE

P Z

-val

ue

S1 S2 S3 S6S4 S5

0

All

Ele

ctro

des

Group

5

10

15

0

5

10

15

Significant Negative RSFCNon-significant RSFCSignificant Positive RSFC

250ms

+20uv

S1 S2 S3

S4 S5 S6

Regions with more positive RSFC showed stronger CCEPsbut regions with more negative RSFCs did not

Co-localization of CCEPs and RSFC across brainIn 5/6 subjects, signi�cant CCEPs exhibited higher RSFC than non-signi�cant CCEPsA similar �nding was observed when local electrodes (within 3cm of stimulation site) were removed from analysis

Regressing out distance from stimulation site yielded similar results

0 5 10 15−0.5

0

0.5

1

CCEP

RS

FC

R=0.63

0 0.05 0.1 0.15−1

−0.5

0

0.5

1

Distance (mm )

RS

FC

R=0.73

-1

1 ROI, 128 sites

0 5 10 15

0

CCEP

Res

idua

ls

R=0.62

−1

−0.5

0

0.5

1

20 010

2030

00.040.080.12−0.5

0

0.5

1.0

1.5

2.0

CCEP

Distance (mm )-1

RSFC

20

r = 0.31

CCEP

S5

r = 0.32

S6

10 20 0 10 200n = 25n = 20

r = 0.24

S3

0 10 20

r = 0.21

S4

0 10 20n = 39 n = 35

r = 0.28 r = 0.29

S1 S2

0 10 20 0 20−0.2

0

0.2

0.4

10

RS

FC

n = 44 n =15

0

0.05

0.10

0.15

0.20

Mea

n R

SFC

Z-v

alue

S1 S2 S3 S6

+ - + - + -

S4

+ -

S5

+ - + - + -

+ -

0

0.05

0.10

0.15

0.20

0

0.05

0.10

0.15

0.20

All

Ele

ctro

des

0

0.05

0.10

0.15

0.20

Mea

n R

SFC

Z-v

alue

+ - + - + -CCEP Response

+ - + - + -

Non

-Loc

al E

lect

rode

s

Group

+ Significant CCEP - Non-significant CCEP

+ -0

0.05

0.10

0.15

0.20

Mea

n R

SFC

Z-v

alue

N1

+ -0

0.05

0.10

0.15

0.20

N2

+ -0

0.05

0.10

0.15

0.20

Mea

n R

SFC

Z-v

alue

+ -0

0.05

0.10

0.15

0.20

All Electrodes Non-local Electrodes

Group analysis for N1 (10-70ms) and N2 (70-500ms)

Spectral analysis of CCEP and RSFC

We observed no di�erence between the amplitude of evoked response between regions with negative and non-signi�cant RSFC2.However, changes may exist in the evoked spectral properties between groups. Power in delta and gamma frequencies of spontaneous ECoG have been shown to correlate with RSFC1. Anticorrelations are thought to represent segregated networks and may communicate with each other via neuronal oscillations at certain frequency bands. �erefore, we applied time-frequency decomposition to CCEPs to further investigate this relationship. EEGLAB was used to calculate event related spectral pertubation (ERSP) and inter-trial coherence (ITC) from 0.5-200Hz.

CCEPs converted to frequency domain and power during N1 (10-70ms) and N2 (70-500ms) was compared to RSFC correlation

N1: Signi�cantly higher CCEP power in all frequencies for positive RSFC when compared to non-signi�cant and negative RSFC

1-4Hz

RSFC < -0.3-0.3 < RSFC < +0.3 (non-significant)RSFC > 0.3

N1

4-8Hz 8-11Hz

11-25Hz 25-70Hz

1-200Hz

Nor

mal

ized

Pow

er

665 2981 1007

70-200Hz

1-4Hz

RSFC < -0.3-0.3 < RSFC < +0.3 (non-significant)RSFC > 0.3

N2

4-8Hz 8-11Hz

11-25Hz 25-70Hz

1-200Hz

665 2981 1007

70-200Hz

0

1

p < 0.05

N1: Low gamma (25-70Hz) shows higher CCEP power for negative RSFC than non-signi�cant RSFCN2: Higher CCEP power in lower frequencies (<11Hz) for positive RSFC compared to non-sign�cnant and negative RSFCN2: Negative RSFC shows signi�cantly higher CCEP power in higher frequencies (>11Hz) compared to non-signi�cant RSFC

One subject, all stimulation sites, all electrode responses

Avg of ERSP

Freq

Hz

-200 0 200 4000359

17295298

200

-5

0

5

-200

0

200

uV

ERP Inter-trial coherence

-200 0 200 4000359

17295298

200

0

0.5

1

1-200Hz

0 100 200 300 400 500

80-200Hz

Tria

l

1-200Hz

5

10

15

20 -200

-100

0

100

20080-200Hz

-200 0 200 4000

100

200

300

Time (ms)

-200 0 200 400

-200 0 200 400

0 100 200 300 400 500

Tria

l

5

10

15

20

5

10

15

20

dB

1 He BJ, Snyder AZ, Zempel JM, Smyth MD, Raichle ME. Electrophysiological correlates of the brain’s intrinsic large scale functional architecture. PNAS, 2008; 39-44.2 Keller CJ, Bickel S, Entz L, Ulbert I, Milham MP, Kelly C, Mehta AD. Intrinsic functional architecture predicts electrically evoked responses in the human brain. PNAS, 2011; 308-13.

Comparison of positive and negative RSFC and CCEP

Examples of relationship between RSFC and CCEP

Seed Time Series

Cortico-cortical evoked potentials (CCEP)

Resting state functional connectivity (RSFC)

-1 -.3 +.3 +1

Z-value

Z-score of CCEPat each electrode

RSFC (Fisher z-transformed r-value) at each electrode

0 .5 1s .2 .1 .3 .4

N1

N2

Stimulation

0 1s

Correlate

Significant response

Non-significant response

Stimulated electrodes

30s

0

% B

OLD

cha

nge

Correlated Time Series

-200uV

Freq

Hz

0

3

5

9

17

29

52

98200

-200

0

200

uV

Tria

l

5

10

15

20

0

3

5

9

17

29

52

98200

Avg of ERSP

-200 0 200 400

ERP Inter-trial coherence

-200 0 200 400

1-200Hz

0 100 200 300 400 500

80-200Hz

1-200Hz 80-200Hz

-200 0 200 400

Time (ms)

-200 0 200 400

-200 0 200 400

0 100 200 300 400 500

Tria

l

5

10

15

20

-5

0

5

0

0.5

1

-200

-100

0

100

200

0

100

200

300

dB

Stimulated electrodeRecording electrode

N1 N2

Each trendline represents 1 stimulation site (80-128 response sites)

Consistent high frequency (80-200Hz) oscillations following stimulationStereotyped ERP with high inter-trial coherence

No clear distinction between regionsof positive / negative RSFC and ERPor spectral properties of CCEP

Sample sizes for each groupare noted in white text

Probing Language Areas

Stimulation

-1 -.3 +.3 +1

Probing Broca’s Area

0 10 20 30 40 50 -0.5

0

0.5

1.0

r = 0.55

0 10 20 30 40 50 -0.5

0

0.5

1.0

r = 0.56

All Electrodes Significant Electrodes

CCEP

RS

FC

RS

FC

CCEP0

0.1

0.2

0.3

0.4

0.5

0.6

+ -

Mea

n R

SF

C Z

-val

ue

+ Significant CCEP - Non-significant CCEP

Stimulation

- 200 uV

500 ms

-1 -.3 +.3 +1

Probing Wernicke’s Area

r = 0.47

0 5 10 15 20 25 30-0.5

0

0.5

1.0

0 5 10 15 20 25 30 -0.5

0

0.5

1.0

r = 0.63

CCEP

RS

FC RS

FC

CCEP

All Electrodes Significant Electrodes

0

0.1

0.2

0.3

0.4

0.5

0.6

+ -

Mea

n R

SF

C Z

-val

ue

+ Significant CCEP - Non-significant CCEP

Stimulated electrodeSigni�cant CCEPNon-signi�cant CCEP

Stimulated electrodeSigni�cant CCEPNon-signi�cant CCEP

0 .5 1s .2 .1 .3 .4

N1

N2

Stimulation

-200uV

0 .5 1s .2 .1 .3 .4

N1

N2

Stimulation

-200uV

Stim