Comodulation and Coherence in Normal and Clinical Populations

David A Kaiser, Ph.D.Rochester Institute of Technology

Chagall

Birthday

Defining my terms

• Raw EEGs are voltages across time

– In time domain, we estimate • amplitude (positive and negative values)

– at a sample rate (only positive)

– In frequency domain, we estimate• magnitude (only positive) • phase (positive and negative)

– at a frequency

• Spectral power is magnitude squared

How similar are two signals?

• Cross-correlation reveals similarities in time between signals. (e.g., Barlow, 1951; Brazier & Casby, 1951)

• Cross-spectral analysis reveals similarities in frequency… [next slide]

How similar are two signals?

• Cross-spectral analysis reveals similarities in frequency. – Signals may be similar in phase, magnitude, or both

• phase analysis: coherence (Goodman, 1957; Walter, 1961)

• magnitude analysis: comodulation (Pearson, 1896; Kaiser, 1994)

For example, does cortical activity become more or less similar after treatment?

Cross-spectral analysis

Coherence estimates phase consistency Comodulation estimates magnitude consistency …between signals at each specified frequency across time

Coh = average normalized cross-spectrum amplitude2

Comod = average normalized cross-product amplitude

Coh range from 0.0 to 1.0Comod range from -1.0 to 1.0

Confusing point: Tukey called “coherency” the square root of coherence

Comodulation was invented to examine low spatial resolution concerns of EEG topography

(e.g., volume conduction, Nunez, 1990)

Does surface EEG reflect cortical potentials well?

- if not, all neighbors will be equally correlated with each other

- if so, correlations will be stronger within functionally-related areas

…coherent if their phase relationship is stable

…comodulated if their magnitude relationship is stable

Signals are …

• Coherence analysis provides– Coherence (Coh)– Phase delay (+/-180o)

• Comodulation analysis provides– Comodulation (Comod)– Proportion: Site 1/Site 2

Functional model for dominant frequency

...suggests common generatorSingle network organizes neural activity

...suggests common responseMultiple networks (related but dissimilar) organize neural activity

• Coherent but not comodulated– Pacemaker network unified

– Primitive recruitment• Synchronization

• Comodulated but not coherent

– Pacemaker network partly segregated by cortical feedback

– Complex recruitment

• Coordination

Fastforward systemDrivers

Thalamocortical projectionsFast, focal, strong

(Momentary) Consciousness

Feedback system

Modulators

Corticothalamic projections

Slow, diffuse, weak

• Sustained consciousness (i.e., self-)

Why comodulation analysis is performed on magnitude (V) and not on power (V2)

• Brief history of power spectral analysis in EEG– Dietsch (1932) analyzed 7 EEG signals using

Fourier (1831).

– Cooley & Tukey (1965) invented FFT algorithm, reducing computer workload, allowing practical spectral applications

– Dumermuth & Fluhler (1967) applied FFT to EEG

• BUT ...

• Why assume brain rhythms and mental activity are related by a power function? Are changes in brain behavior actually associated with larger changes in mental behavior (i.e. reason for using the power spectrum)?

• Might brain and mind activity be more linearly related at this level of investigation (i.e., reason for using the magnitude spectrum)?

Comodulation versus Coherence Data sets

Group N= Females Right handed Age

mean

Age range

ADHD 7 1 All 9.0 6-11

Asperger 11 3 All 10.6 7-16

Child 10 5 All 10.4 5-13

Adult 20 10 All 28.2 23-39

THANKS to Jolene Ross & Jim Caunt for ADHD, some of the AS data; Coralee Thompson for normal children

EEG Comodulation and Coherence values are often very similar!

• Dark bars = Comod Red/green bars = Coherence

Eyes Closed Replications

Within subject, n=20 EC1 v 2:

r = .91 Coh

r = .84 Comod

Being more reliable also can mean less sensitive to state differences

Rho DATA Z-SCORE from norms

How to read our Comod & Coh maps

Typical Atypical(25 college students) (1 college student)

If you build it (adult pattern of frontal lobe myelination), it still takes time for them to come…

MTBI patient Rage Disorder

Autobiography dysfunction associated with reduced right anterior temporal pole

connectivity (Asperger’s, Schizophrenia)

Hypermodulations after stroke, 74yF (seen during math task only!)

Autism as severe global disconnectivity

Asperger v Child Norms8-12 Hz (in Std Error)

ADHD v Child Norms 8-12 Hz (in Std Error)

Global Comodulation by age

19-site mean of 18 comparisons each site

Life is about making connections...

…but not too fast!...

Beta hypercoherence between occipital and medial frontal cortex, esp. right-sided, during rest for Asperger children

(resembles adult pattern)

…in fact, slowing down the rate of connections at some times in your life may even do you some good!

One form of Intelligence (neotenous) resists the natural neural integration trajectory

Neural & “behavioral” (Ph.D) indices of neoteny