Transcript Presentation 1 - Translational Neuromodeling Unit

Introduction to
EEG: Oscillations
Saee Paliwal
Translational Neuromodeling Unit
Overview
Background
 History of EEG
 Measurement techniques
What is an EEG?
 Components of an EEG
 Sample healthy data
 Patient data
Oscillations and Synchrony
 Characteristics of brain oscillations
 Oscillations and synchrony
 The function of oscillations
Interesting applications
 Phasic artificial neural networks
A brief history of the
electroencephalogram (EEG)
1875: English physician Richard Caton observes EEG from
the exposed brains of rabbits and monkeys.
1924: German Neurologist Hans Berger uses radio
equipment to amplify electrical activity measured from the
human scalp. He also coins the term electroencephalogram.
1934: Adrian and Matthews published the paper verifying
concept of “human brain waves” and identified regular
oscillations around 10 to 12 Hz which they termed “alpha
rhythm”
Bronzino, 1995
Modern Implementation
Electrode placement for a 10-20 electrode EEG looks as follows:
Two modes of recording--Differential and Referential:
 Differential: two inputs to each differential amplifier are from
two electrodes
 Referential: one or two reference electrodes are used
What is EEG?
 A medical imaging technique that reads scalp electrical
activity generated by brain structures.
 The EEG itself is defined as “electrical activity of an
alternating type recorded from the scalp surface after
being picked up by metal electrodes and conductive
media” (Niedermeyer et al, 1993)
How does this work?
 When the brain is activated, local current flows are
produced
 EEG measures currents that flow during the dendritic
activity of cortical pyramidal neurons.
 The electric potentials are a result of are a result of
graded potentials from pyramidal cells creating electrical
dipoles between the soma and apical dendrites.
 Current is read non-invasively through the skin, skull, etc.
Signals are amplified and normalized during post-
Components of an EEG
 Brain waves are sinusoidal, generally in the range of 0.5 to 100
μV in amplitude
 Through Fourier transforms on the brain signal, a power
spectrum from the raw EEG signal is derived. Brain waves have
been categorized into four basic groups
 beta (>13 Hz)
 alpha (8-13 Hz)
 theta (4-8 Hz)
 delta (0.5-4 Hz).
Alpha and Beta waves
Alpha
 Frequency: 8–13 Hz
 Location: Posterior half of the head and are usually found over
the occipital region of the brain.
 Interpretation: Indicate both a relaxed awareness, may be is
nothing but a waiting or scanning pattern produced by the visual
regions of the brain. Eliminated by opening the eyes, by hearing
unfamiliar sounds, by anxiety, or mental concentration or
attention.
Beta
 Frequency: 14–26 Hz
 Location: the frontal and central regions
 Interpretation: Associated with active thinking, active attention,
focus on the outside world, or solving concrete problems, and is
found in normal adults. A high-level beta wave may be acquired
when a human is in a panic state.
Delta and Theta waves
Delta
 Frequency: 0.5–4 Hz.
 Location: Cortex or thalamus
 Interpretation: Associated with deep stage 3 of slow-wave
sleep and help characterize the depth of sleep.
Theta
 Frequency: 4–7.5 Hz.
 Location: Thalamus, Hippocampus, cortex
 Interpretation: Two types: type 1 associated with voluntary
behavior and during REM sleep, and type 2 appears during
immobility and anasthesia. A theta wave is often accompanied
by other frequencies and seems to be related to the level of
arousal. Larger contingents of theta wave activity in the waking
adult are abnormal and are caused by various pathological
problems.
Data
Example EEG recording, 10 channels
Dementia
EEG recording of a patient with Creutzfeldt–Jakob diseases
(CJD).
CJD is characterized with aphasia, apraxia, and agnosia, and can
be diagnosed using EEG signals. One sees a slowing of the delta
and theta wave activities and, after approximately three months of
the onset of the disease, periodic sharp wave complexes are
generated that occur almost every second, together with a decrease
in the background activity
Epilepsy
Example EEG recording of a patient with epilepsy.
Onset of a clinical seizure characterized by a sudden change of
frequency in the EEG, in the alpha wave frequency band with a slow
decrease in frequency (but increase in amplitude) during the seizure
period. It may or may not be spiky in shape. Bursts of higher
frequencies can be seen in sections of the EEG.
(a) seizure activity
mal seizure
(b) grand
What exactly are we
measuring?
The beating brain
The mechanism:
 The fundaments of the EEG signal lie in in the following
mechanism:
 Brain cells are activated and local current flows are
produced
 EEG measures the currents that flow during synaptic
excitations of dendrites of pyramidal neurons in cortex.
 Large populations of neurons generate EEG-readable
activity
Some characteristics:
 The data itself is continuous and periodic. Even a single neuron
can oscillate!
 Neighboring frequency are associated with different brain states
and compete each other, but several rhythms can coexist at the
same time.
The power of the beat
 The power density of EEG inversely proportional to frequency (f
).
 Perturbations in the slower frequencies can cause a cascade of
energy dissipation at higher frequencies—the slower waves are
the moderators.
 Properties of neuronal oscillators are incumbent on the physical
architecture of neuronal networks, and are limited by the speed
of neuronal communication (axon conduction and synaptic
delays)
 The period of neuronal oscillations is constrained by the size of
the neuronal pool involved in the signal (smaller pools generate
higher frequencies)
Oscillations and Synchrony
 Integration of information requires “synchrony” across different
neural populations
 Synchrony is defined by the temporal window within which some
trace of an earlier event is retained, which then alters the
response to a subsequent event.
 Successive events that evoke identical responses are deemed
nonsynchronous.
 Large scale neuronal oscillators behave like relaxation
oscillators—their activity is phase dependent, so the “duty cycle”
of the oscillator is separated from the receiving phase, and can
synchronize robustly.
The function of oscillations part
1
Input selection and plasticity:
The neuronal (single or assembly) response to a strong input is an
oscillation, the frequency of which is a function of two things: the
leak conductance and capacitance of the neuronal membrane
(responsible for low-pass filtering) and the voltage-gated currents
(high-pass filtering).
 This makes the neuron a band-pass filter!
These resonant frequencies allow neurons to select inputs based on
their frequency characteristics and set network dynamics.
Binding cell assemblies:
 Information in the brain thought to be stored in distributed,
flexible pools of neurons
 Oscillatory synchrony could bind them—cost effective (unlike
chemical synaptic change)
 As long as the frequencies of the coupled oscillators remain
similar, synchrony can be sustained even with very weak
synaptic links
 So basically, activated neuronal groups in distant cortical
The function of oscillations part
2
Consolidation and combination of learned information:
 The resting state of the brain consists of self-governed
oscillations (particularly of thalamocortical loops)
 These oscillations represent information acquired during the
day.
 This replay of information learned allows for “compiling”
Representation by phase information:
 For any oscillator, the coupling strength is proportional to the
magnitude of phase shift (phase advancement).
 Exploiting this property allows for rapid, short-term storage in
the forward phase shift of action
 Oscillators can exploit STDP through several temporal iterations
of a pattern—this kind of learning disappears when theta waves
are disrupted.
An interesting application…
Synchrony in ANNs
Artificial Neural Networks benefit from
synchrony
 Reichart et al, 2014 proposed a novel approach to Deep
Learning—neural networks with phase synchrony
 Essentially, he added a phase component to the transfer
function
 The most striking result is that he was able to disentangle
overlapping images, demonstrating the idea that each image
concept was stored in phasic information
Thank you!!
Questions?
References
J. D. Bronzino. 1995. Principles of Electroencephalography. In: J.D. Bronzino ed. The
Biomedical Engeneering Handbook, pp. 201-212, CRC Press, Florida.
Schnitzler A, Gross J (2005) Normal and pathological oscillatory communication in the brain.
Nat Rev Neuro 6: 285–296.
Schyns PG, Thut G, Gross J (2011) Cracking the Code of Oscillatory Activity. PLoS Biol 9(5):
e1001064. doi:10.1371/journal.pbio.1001064
E. Niedermeyer, F. H. Lopes da Silva. 1993. Electroencephalography: Basic principles,
clinical applications and related fields, 3rd edition, Lippincott, Williams & Wilkins, Philadelphia.
Reichert, David P., and Thomas Serre. "Neuronal Synchrony in Complex-Valued Deep
Networks." arXiv preprint arXiv:1312.6115 (2013).
Buzsáki, György, and Andreas Draguhn. "Neuronal oscillations in cortical networks." Science
304.5679 (2004): 1926-1929.
Buzsaki, Gyorgy. Rhythms of the Brain. Oxford University Press, 2006.