David M. Groppe1, Taufik A. Valiante2,3, Christopher J. Honey11Dept. of Psychology, University of Toronto 2Division of Fundamental Neurobiology, Toronto Western Research Institute 3Division of Neurosurgery, Dept. of Surgery, University of Toronto

Hypothesis •Sub-50 Hz iEEG rhythms act as a regulatory mechanism that reduces neuronal firing. Strong regulatory rhythms will inhibit the spread of ictal activity.

1. The relative power of putative regulatory frequencies does not robustly predict the relative speed of ictal propagation. This may be due to factors we have not controlled for (e.g., the strength of the ictal activity in areas immediately upstream from each region) and the difficulty with measuring rhythm power immediately preceding high gamma ictal propagation.2. The preictal sub-50 Hz correlation with high gamma power reliably predicts if a channel will exhibit high gamma ictal activity. This observational effect might stem from the inhibitory effect of ictal regulatory rhythms. However it might also be due to IIDs or an accident of anatomy. Future work will investigate these possibilities.

Inhibitory effects of pre-ictal brain rhythms on seizure propagation

Introduction Types of Propagation

Propagation Times

The latency of propagation to each channel was correlated with the mean regulatory frequency power in the 30 sec prior to ictal onset.

Contact: david.m.groppe@gmail.com

Identifying Regulatory Frequencies

Predicting Propagation Times Predicting Ictal Involvement

Conventional Propagation Latency Pre-Ictal High Gamma Power Correlations Predict Ictal High Gamma Activity

Pre-Ictal Baselined Spectrograms

L. Hippocampal Onset Channel

L. Posterior Temporal Channel

L. Central Temporal Channel

Hippocampal Onset

Conventional Propagation

High Gamma Propagation

Pre-ictal power timecourses were computed for all frequencies in the 1-50 Hz range (2 Hz bins) and were correlated with preictal high gamma power . Outlier high gamma points rejected to avoid IIDs.

50 100 150 200 250

−2

0

2

4

Z−score

Seconds

High Gamma (70-100/200 Hz) Log Power12-14 Hz Log PowerDiscarded Outliers

L. Posterior Temporal Channel

Pearson’s r

1-50 Hz x High Gamma Power Correlations, All Channels, Patient 1

=Frequency with correlation closest to -1 Hz

Power (dB

)Pear

son’

s r

Channel LPT2-3, Patient 1

Hz10 20 30 40 50

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.2

Channel LPT2-LPT3Correlation with High Gamma PowerPower Spectral Density -40

-44

-48

-52

-56

-60

-64

-68Minimum r

The 1-50 Hz frequency that shows the minimal correlation with high gamma power is the putative regulatory frequency for that channel.

0

10

20

30

40

50

Seco

nds

<−Left Right−>

LHD

1−2

LHD

2−3

LHD

3−4

LAT1−2

LAT2−3

LAT3−4

LMT1−2

LMT2−3

LMT3−4

LPT1−2

LPT2−3

LPT3−4

RH

D1−

2

RH

D2−

3

RH

D3−

4

RA

T1−2

RA

T2−3

RA

T3−4

RM

T1−2

RM

T2−3

RM

T3−4

RPT

1−2

RPT

2−3

RPT

3−4

Day 11, Seizure 2Day 12, Seizure 1Day 13, Seizure 1Day 13, Seizure 2Day 14, Seizure 1Day 15, Seizure 1Day 15, Seizure 2Day 1, Seizure 1Day 3, Seizure 1Day 3, Seizure 2Day 5, Seizure 1

Latency of first spectral power change in each channel following left hippocampal seizure onset in Patient 1. Purple bars=SD. Red horizontal lines=Mean.

High Gamma Propagation Latency

0

20

40

60

80

100

Seco

nds

<−Left Right−>

LHD

1−2

LHD

2−3

LHD

3−4

LAT1−2

LAT2−3

LAT3−4

LMT1−2

LMT2−3

LMT3−4

LPT1−2

LPT2−3

LPT3−4

RH

D1−

2

RH

D2−

3

RH

D3−

4

RA

T1−2

RA

T2−3

RA

T3−4

RM

T1−2

RM

T2−3

RM

T3−4

RPT

1−2

RPT

2−3

RPT

3−4

Day 11, Seizure 2Day 12, Seizure 1Day 13, Seizure 1Day 13, Seizure 2Day 14, Seizure 1Day 15, Seizure 1Day 15, Seizure 2Day 1, Seizure 1Day 3, Seizure 1Day 3, Seizure 2Day 5, Seizure 1

Latency of first increase in high gamma power in each channel following left hippocampal seizure onset in Patient 1. Purple bars=SD. Red horizontal lines=Mean.

Conventional Propagation

High Gamma Propagation

Although these data fit expectations, across patients the effect is not robust: MN (SE) r=0.04 (0.16)

Although these data fit expectations, across patients the effect is not robust: MN (SE) r=0.16 (0.17)

−3 −2 −1 0 1 2 3

−2

−1

0

1

2

3

Postonset to Prepropagate Reg. Freq. Power (Z−score)

Hig

h G

amm

a Pr

opag

atio

n La

tenc

y (Z−s

core

)

LHD3−4LAT1−2LAT2−3LAT3−4LMT1−2LMT2−3LMT3−4LPT1−2LPT3−4RHD1−2RHD2−3RHD3−4RAT1−2RAT2−3RAT3−4RMT1−2RMT2−3RMT3−4

r=0.38

Patient 1

−3 −2 −1 0 1 2

−2

−1

0

1

2

3

Postonset to Prepropagate Reg. Freq. Power (Z−score)

Prop

agat

ion

Late

ncy

(Z−s

core

)

LHD3−4LAT1−2LAT2−3LAT3−4LMT1−2LMT2−3LMT3−4LPT1−2LPT2−3LPT3−4RHD1−2RHD2−3RHD3−4RAT1−2RAT2−3RAT3−4RMT1−2RMT2−3RMT3−4RPT1−2RPT2−3RPT3−4

r=0.25

Patient 1 1-50 Hz x High Gamma Power Correlations, All Channels, Patient 1

Hz

•6 Engel 1 patients (5 MTLE, 1 MT+FLE)•Mean [SD] seizures: 8 [4]•Mean [SD] iEEG electrodes: 27 [3]•Mean [SD] preictal data: 25 [13] minutes

MotivationIn sensory and motor cortex, the power of sub-50 Hz iEEG rhythms (e.g., beta in motor cortex) is anti-correlated with high gamma band (70-200 Hz) power, which is a robust correlate of local action potentials. This appears to be a general property of most of neocortex (Honey et al., in prep). We think that these sub-50 Hz rhythms may act as a regulatory mechanism that decrease the excitability of neocortex. If true, then strong regulatory rhythms should (1) inhibit the spread of ictal activity into an area and (2) prevent marked increases in mean population firing rate.

Conclusions

No

or In

frequ

ent H

igh

Gam

ma

Icta

l Act

ivity

Freq

uent

Hig

h G

amm

a Ic

tal A

ctiv

ity

Pearson’s r

Channels that do not show high gamma ictal activity tend to have low high gamma x regulatory frequency power correlations in the preictal period.

0.5

0.6

0.7

0.8

0.9

1

Area Underthe Curve

Proportion Correct(Threshold P(HGIA)>0.5)

Boxplots (n=6)

Utility of predictions for left out patient are evaluated with area under the curve (AUC, maximum accuracy=1, chance accuracy=0.5) and proportion of correct predictions (where we predict high gamma ictal activity if probability exceeds 50%). AUC is reliably greater than 0.5 (one-tailed permutation test, p=0.02, t(5)=5.58).

Median propagation latency across all channels for each seizure was subtracted from channel propagation times to control for differences in seizure strength. Latency and power were converted to z-scores to control for interchannel differences.

The latency of high gamma propagation to each channel was correlated with the mean regulatory frequency power in the 30 sec prior to ictal onset. Propagation latencies and regulatory frequency power were normalized to control for differences across seizures and channels as in the conventional propagation analysis described above.

Example 12-14 Hz & High Gamma Power Pre-Ictal Time Series

and somatotopically distinct for different fingers (A). In contrast,changes in beta power are broadly distributed and essentially thesame for any finger movement. The instantaneous broadbandpower shows a clear temporal correlation with individual fingermovements (C); note that the broadband trace obtained at eachsite correlates with the position of only one of the fingers and notthe others. The spatial overlap of changes for particular pairs ofdigits is summarized in Figure 3B. Overlap (B) is quantified by theproportion of sites showing similar changes with movement of thetwo digits (Methods), and is larger for the beta power (nearcomplete overlap) than broadband power (largely distinct spatial-ly). Similar overlap patterns are documented for the other eightsubjects as well (D). Broadband spectral changes show largely non-overlapping representation between digits, while the distribution of

change in the beta rhythm overlaps almost completely in all cases.Digit-specific electrodes were typically 1 cm from one another(also see Figures S5, S6, S7, S8, S9, S10 in Text S1).

Coupling between low frequency rhythm phase andamplitude of broadband spectral change

The relation between the instantaneous broadband power and

phase of ECoG signal at different frequencies is documented in

Figure 4. The analysis is based on recordings obtained during

finger movements and rest (B). The dynamic broadband (C)

obtained from the raw ECoG potential (D) shows distinct

fluctuations related to index finger movement. The ECoG can

be filtered with a simple wavelet to obtain a temporally varying

Figure 1. Electrocorticographic (ECoG) activity in rolandic cortex during movement and rest (Subject 1). (A) ECoG potential ismeasured from the brain surface. (B) ECoG electrodes in situ, embedded in silastic. (C) Electrode positions on the cortical surface. (D) Traces showsimultaneous finger position color-coded as in inset (top) along with aspects of the potential timeseries. The raw ECoG voltage at bottom (black) isshown from the M1 site marked with green dot in A–C. ‘‘Motor rhythm’’ is 12–20 Hz bandpassed ECoG (gold trace). ‘‘Broadband spectral change’’

(pink) is the timeseries of an estimate of the coefficient A(t) in a power law in the power spectral density of the form P(t)~A(t)1

f x. (E) The power

spectral density during movement (green) and rest (black) reveals a decrease in a peaked process at low frequencies (gray – 12–20 Hz), and abroadband increase across the rest of the frequency range during movement (60 Hz line noise and harmonics omitted). (F) The spatial distribution ofsites showing a decrease in 12–20 Hz power associated with thumb movement. (color represents a signed r2 measurement, scaled to the maximumacross the array: 0.38). In these figures the Central sulcus and Sylvian fissure are shown in yellow. (G) Broadband spectral changes associated withthumb movement are similarly shown (maximum: 0.64).doi:10.1371/journal.pcbi.1002655.g001

Motor Cortical Activity Is Coupled to Rhythm Phase

PLOS Computational Biology | www.ploscompbiol.org 3 September 2012 | Volume 8 | Issue 9 | e1002655

Modified from: Miller et al. (2012) PLOS Comp. Bio.

High Gamma Band

1-50 HzPutative 12-20 Hz RegulatoryFrequency

Correlate of mean population

firing rate

High Gamma Band

Boxplots, All Patient Data (n=1286)

High Gamma Ictal Activity

No Yes

−0.2

0

0.2

0.4

0.6

Min

Sub

50H

z x

Hig

h G

amm

a Po

wer

rM

in. 1

-50

Hz

x H

igh

Gam

ma

Pow

er r

Logistic Regression, 5/6 Patients

−0.25 0 0.25 0.5 0.750

0.2

0.4

0.6

0.8

1

P(H

igh

Gam

ma

Icta

l Act

ivity

)

Min. 1−50 Hz x High Gamma Power r

High gamma ictal activityNo high gamma ictal activityModel fit

To evaluate the ability of preictal high gamma x regulatory frequency power correlations to predict high gamma ictal activity we used logistic regression and leave-one-out cross-validation.