a review of neural networks & theoretical ... - about ansa
Post on 26-Mar-2022
3 Views
Preview:
TRANSCRIPT
A Review of
Neural Networks
& Theoretical
ApplicationsDAVID S. CANTOR, PH.D., MS, FNAN, QEEG-D, BCN
MIND AND MOTION DEVELOPMENTAL CENTERS OF GEORGIA, LLC
WWW.MINDMOTIONCENTERS.COM
1
Acknowledgments
*DICK GENARDI, PH.D. (PRESENTATION CONTRIBUTOR)
*RICHARD SOUTAR, PH.D. (PRESENTATION CONTRIBUTOR)
STEPHEN STAHL, MD, PH.D.
LUKASZ KONOPKA, PH.D.
ADRIAN VAN DEUSEN
THOMAS COLLURA, PH.D.
RON BONNSTETTER, PH.D.
ROBERT CHABOT, PH.D.
LESLIE PRICHEP, PH.D.
Chaos is the science of surprises, of the nonlinear and the unpredictable. It
teaches us to expect the unexpected. While most traditional science deals with
supposedly predictable phenomena like gravity, electricity, or chemical reactions,
Chaos Theory deals with nonlinear things that are effectively impossible to predict
or control, like turbulence, weather, the stock market, our brain states, and so
on. These phenomena are often described by fractal mathematics, which captures
the infinite complexity of nature. Many natural objects exhibit fractal properties,
including landscapes, clouds, trees, organs, rivers etc, and many of the systems in
which we live exhibit complex, chaotic behavior. Recognizing the chaotic, fractal
nature of our world can give us new insight, power, and wisdom. For example, by
understanding the complex, chaotic dynamics of the atmosphere, a balloon pilot
can “steer” a balloon to a desired location. By understanding that our ecosystems,
our social systems, and our economic systems are interconnected, we can hope
to avoid actions which may end up being detrimental to our long-term well-being.
CHAOS….
4
“As far as the laws of mathematics refer to reality, they are not certain, and as far as they are certain, they do not refer to reality.”
-Albert Einstein
Day 1: Basics
of Neural
NetworkI. GENERAL CONCEPTS AND TERMS
5
Brain Complexityhttps://youtu.be/OCpYdSN_kts
Brain Rhythms: Functional Brain Networks Mediated by Oscillatory Neural Coupling
Micrograph of the cell body
with synaptic inputs.
As many as 30-50K inputs per
cell in the mammalian brains
are seen.
Cerebellar cells may have as
many as 100K inputs.
Connection Basics -
Outline
◆ Explain most common network terms found in the functional imaging literature.
◆ Discuss parcellation and types of networks.
◆ Review key regions of interest
◆ Outline the Development of Networks and how they facilitate perception and cognition (adaptation and ontogenetic development of behaviors)
8
Connection Basics
Current purpose in studying the connectome of the brain is to
discover
◆ The mechanisms of communication between one structural
system/network to others, assuming each operating at
different/multiple frequencies, orchestrating activity from
specific to integrative which underlie the full range of human
behavior. This transfers occurs from ventral to dorsal levels,
between areas within and across hemispheres.
◆ How successive integration allows adaptation to events both
external and internal to the organism which occur on time
scales ranging from at least milliseconds to spanning several days.
9
Connection Basics 10
The Golden Ratio 11
Connection Terms
https://youtu.be/f1E2lStdrl8
Phi: Spirit Science, The Golden Ratio
Connection
Basics- Terms
Graph TheoryA graph can mathematically represent a network as a series of vertices and connecting edges.
The adjacency matrix lists each vertex and its neighbors.
Directionality, weighting-strength of connection, could be represented
Graph theory is used to describe structure and operational characteristics of a network, detect and
quantify subsystems; mapped to functions. Straaten & Stam, 2013
13
M.van den Hueval & O. Sporns,2013
14
Connection Basics-Terms
◆ Node or vertex is an area of the brain as small as
a neuron or much larger region such as a
Brodmann area which is connected to other
nodes which together form a network(s).
◆ A module ( many neurons) is a group of nodes
with a large number of mutual connections within
the group and few connections to nodes out of
the group.
◆ Outside connections are to other modules which
may serve different functions
15
Connection Basics-Terms
◆ The role/importance of a node is described by
various measures called centrality measures.
◆ Nodal degree- “K” is a key centrality measure
referring to the number of edges or connections
from a node to other nodes(neighbors). Can
indicate when a node is a point of relative importance in a network. Hubs are high degree
nodes. An average K can also be calculated for
a whole network.
◆ Edges are structural connections between nodes.
16
Connection Basics-Terms
◆ Degree correlation- the extent to which nodes of the same or different degree are connected to each other.
◆ Assortative- when nodes show tendency to connect with nodes of the same degree
◆ Disassortative- when nodes show tendency to connect with nodes of different degree.
◆ At macroscopic level( EEG, MEG, fMRI) brain is assortatively organized; disassortative at neuronal level.
17
Connection Basics-Terms
◆ Hubs are nodes/vertices which play a central role in the overall
function of a network. Two major hub types are connector hubs
and provincial hubs.
◆ Connector hubs are hubs which have a broad array of
connections between modules within a network or connections between networks.
◆ Provincial hubs are also high degree nodes but which connect
to other nodes within the same module.
18
Connection Basics-Terms
◆ Rich club is example of nodes with high assortativity.
◆ Level of parcellation influences this metric. Complex
modular networks are likely assortative
◆ High degree nodes serve as connector hubs for
integration of locally processed information.
◆ At macroscopic level assortativity may arise from the
way the network is formed: by modulation based on
synchronization rather than growth. ( Stam,2010)
19
Connection Basics-Terms
◆ P(k) is the degree distribution of a graph, module,
network or system. It is a measure of the
probability that a randomly chosen node will
have the probability “k.”
◆ The presence of hubs is indicated by a power law
degree distribution, i.e. some nodes have an very
high degree, and most others have a low degree.
20
Connection
Basics-Terms
“L”
CPL-characteristic path length/ simply path length-“L” is number of vertices involved to get from one node to another; equals number of nodes crossed (involved?) -not actual length of axons or spatial distribution of the nodes. Shorter “L” is held as more efficient; inversely related to IQ
Shortest path from vertex 1
to 12 is 4. L = 4.
21
Connection
Basics-Terms
“C”
Clustering Coefficient
Is a local connectivity measure indicating to what extent the neighboring nodes, to which a node is connected, are connected to each other. Expressed as
ratio of actual number of edges (connections) to the number of total possible connections between a nodes neighbors
Vertex 8 has neighbors 5 & 11. “C” is
calculated by number of edges between
neighbor vertices divided by the total
possible number of edges. Thus, C= 1/1= 1.
22
Connection Basics-Terms
◆ Closeness centrality is the length of the shortest paths
between a node and the rest of the network.
◆ Betweeness centrality is the number of short communication
paths a node participates in. It is defined as the number of
shortest paths going through a node or edge. The betweeness
centrality is high when the node or edge is used for many
shortest paths. This measure can be normalized by dividing it
by its maximum value (the total amount of shortest paths in
the network). A relative draw- back of this method is that
computation time can be long, especially in networks with
many nodes.
23
Connection Basics-Terms
◆ Eigenvector centrality allocates a value to each
node in the network in such a way that the node
receives a large value if it has strong connections
with many other nodes that have themselves a
central position within the network (Lohmann et
al., 2010).
◆ Thus, connections to important nodes count
more, making the nodes with relatively few edges
to very important nodes also important (maybe more important than nodes with many edges to
less important nodes).
24
Connection Basics -Terms
◆ Small worldness (SW)/Sigma is the balance between network integration and differentiation; the ratio of local clustering and the characteristic path length (CPL) of a node relative to the same ratio in a randomized (hypothetical) network.
◆ Global efficiency -the average shortest path length –i.e. smaller the CPL the more efficient the network. Some authors dispute this finding.
◆ Regional efficiency is global efficiency computed for each node; related to the clustering coefficient.
25
Connection Basics-Terms
Review
Most frequently referred to:
K- degree
P(k) degree distribution
L –path length
C- clustering coefficient
Also: Sigma ( a cumulative metric) = the ratio of normalized clustering coefficient to the characteristic path length, a measure of small-world organization.
26
Two modules-
groups of vertices
with high
interconnectivity
with connector
hub (black
square).
Modularity (MOD) is the degree a system can be divided into smaller networks.
27
M.van den Hueval & O. Sporns,2013
28
Connection Basics-Terms
◆ Modules/motifs(brief spatio-temporal patterning) -sub grouping of nodes-edges (graph) can be momentarily accessed (microstate) for a particular function
◆ Modules/subsystems can be hierarchically organized; subsystems within subsystems allowing for functional specialization and finer differentiation.
◆ Higher order system can influence a subsystem without affecting its own intrinsic function.
29
Connection Basics-
Network Types
Ordered
◆ Each node only connected to its nearest neighbors
◆ C =high- only connections to neighbors
◆ L= high, takes many steps to reach distant node
Types fr. Watts & Strogatz, 1998
30
Connection Basics-
Network Types ◆ With probability p randomly
rewire a few edges
◆ L is lowered
◆ C remains high
◆ Complex adaptive system
requires Small World network
& noise or unpredictability
◆ Matches neural & social
networks more precisely
than ordered or random
◆ Human cortex is Small World
Small World
31
Connection Basics-
Network Types
Random◆ Increase p and set it equal for every
combination of two nodes
◆ C & L drop to low value
◆ Hippocampus has random organization
◆ Cerebellum has parallel organization- no reenterant loops.
Erdos & Renyi, 1959
32
Connection Basics-
Network Types
Scale Free◆ Probability of new
connection depends on
node degree K
◆ Higher K node has
higher p
◆ Produces a scale free
network or power law
distribution when-Few
nodes with high K, many
nodes with low K
Scale Free Network
33
Connection Basics-
Structure Implications◆ For the brain to function as a complex adaptive system it must have
small world organization and noise/variance.
◆ Some neurofeedback practitioners question whether constraint of overall variance occurs by training excessive metrics with restricted criteria. Should one consider whether a dynamical system would be limited and adaptation to the training conditions degraded when a certain number of training parameters is exceeded?
◆ Question then--- Do we limit metrics and areas targeted for training whether surface or (s)Loreta?
◆ Consider the possible influence of the number of metrics/constraints employed in training for optimal outcome may relate to the degree of controllability that a ROI evidences on associated networks
34
Defining Structural Connectivity
Leading to Functional Connectivity
◆ There is no standardized generally accepted anatomical definition of a node or edge (Meehan & Bressler,2012)
◆ Many “imaging” modalities in identifying nodes use different measures—metabolic (PET), blood (MRI,fMRI,rs-fcMRI), spectral power (LFP, EcoG, EEG, MEG)
◆ Time domain of MRI is several seconds, EEG –milliseconds; hence different resolution due to time dampening
◆ Correlation of heightened activity ( under task) of a node identified by one modality not fully understood how relates to node identified by different modality. Various advantages- model constructions from each ( Zalesky et. a. 2010)
◆ Network not simply defined by co-activated nodes, must also identify edge
◆ Node could be defined not only by co-activity but by increased correlation with other node while activation remains constant
◆ Use of different values for parameters results in different nodes and connectivity patterns, determines to which network(s) a nodal areas is associated
◆ Measure of single neural unit spiking identifies that 30% of neurons associated with network function are found outside of network itself ; suggests that nodal boundaries are more diffuse than defined by fMRI & function more broadly distributed
35
Connection Basics◆ Modern analysis shows that functions are widely distributed via
grouping of areas from local to long distance networks (Cf. Catani & Theibaut de Shotten, 2012)
◆ The number of and variety of networks determined from structural analysis and functional analysis depend on which metrics used and inclusion parameter values set and whether the data was taken under at rest or at task conditions.
◆ Some networks are referred to as resting state networks and others called demand or at task networks. There can be significant overlap but structure does not account for all functional networks (Honey et al., 2009).
◆ Full understanding of network behavior requires an understanding of distribution, timing, and integration of information in the service of cognition, behavior and emotion
36
Basics of
Neural
NetworkII. DEVELOPMENT
37
Organization
◆ Localized neural circuit developmental and
organization “builds” networks which progressively
become more complex and hierarchical over
time to assimilate accumulative knowledge and
promote “precognition” to facilitate “efficiency”.
38
The Basics
Brain
Development
39
40
Timeline of Changes◆ Differentiation of the neural tube occurs from GA 4 to 12
weeks with new neurons formed in proliferative zones
◆ Between GA 12 and 20 weeks, the neurons multiply followed by migration to cortical destinations
◆ Around GA 29 weeks, the process of myelination starts at the brain stem and continues generally in an inferior-to-superior and posterior-to-anterior path
◆ Between 2 and 7 postnatal years, it is unclear whether synaptogenesis is balanced by elimination of cells and synapses
◆ Myelination of proximal pathways tends to occur first, followed by myelination of distal pathways
◆ Cortical myelination seems to mirror functional development
◆ Maturational trajectories, with sensory tracts myelinating before motor tracts accompanied by protracted myelination of association tracts
41
◆ The term ‘connectivity’ encompasses several concepts in
neuroscience. Structural connectivity describes the physical
link—the long range connections formed by white matter
tracts.
◆ On the other hand, Functional connectivity describes
statistical association of functional signals between brain
areas observed through various functional imaging
approaches, including functional MRI (fMRI), electro- and
magneto-encephalography (EEG, MEG), and
fluorodeoxyglucose (18F) positron emission tomography
(FDG-PET).
42
◆The differential development of
major white matter tracts is also
accompanied by a developmental
shift in their inter-tract microstructural
correlation from a more random
state to a more organized state,
suggesting refinement of white
matter organization with maturation
(Mishra et al., 2013).
43
Connection Basics -
Developmental
44
Connection Basics -
Developmental
◆ Brain continues to mature through the 20s(Gogtay
et al, 2004) well into the 30s (Lebel et al 2011).
◆ Developmental network remodeling shows
decreased grey mater density, due esp to short
range connection synaptic pruning and increase
myelination occuring throughout life. (Hagmann,
2010), (Bartzokis, 2004).
◆ Remodeling seen in terms of decreased path
length, small worldness, clustering, modularity,
and fiber density—specific to hemisphere; into
adulthood.
45
Connection Basics -
Developmental
◆ Fiber connection, nodal degree, and nodal
efficiency in frontal cortex decreased fr 12-30 yrs;
high increase in temporal and parietal areas.
◆ Consistent with development of executive
functions and long range anterior-posterior
increased communication.
◆ Differential trends-left hemisphere shows increased clustering, modularity, and global
efficiency(shorter characteristic path length).
46
Connection Basics -
Developmental
◆ Right hemisphere the opposite. Decrease small
worldness, ie less areas of differentiation.
◆ The splenium and isthmus connect the left and
right hemisphere temporal, parietal and occipital
cortices.
◆ In the splenium and isthmus (Chung,et al. 2001)
associated with age, there is an increase in the inter-hemispheric level of myelination and /or
axon count.
Brain connectivity in
normally developing
children and adolescentsBudhachandra S. Khundrakpam, John D. Lewis,
Lu Zhao, François Chouinard-Decorte, Alan C.
Evans
MCGILL CENTRE FOR INTEGRATIVE NEUROSCIENCE, MCCONNELL BRAIN IMAGING CENTRE, MONTREAL NEUROLOGICAL INSTITUTE, MCGILL UNIVERSITY, MONTREAL, QUEBEC H3A 2B4, CANADA
NEUROIMAGE (2016),
HTTP://DX.DOI.ORG/10.1016/J.NEUROIMAGE.2016.03.062
48
Measuring Networks with:
1. DTI (Diffuse Tensor imaging)-
employing FA◆ Fractional anisotropy (FA) is a scalar value
between zero and one that describes the degree
of anisotropy of a diffusion process. A value of
zero means that diffusion is isotropic, i.e. it is
unrestricted (or equally restricted) in all directions.
A value of one means that diffusion occurs only
along one axis and is fully restricted along all
other directions. FA is a measure often used in
diffusion imaging where it is thought to reflect
fiber density, axonal diameter, and myelination in
white matter. The FA is an extension of the concept of eccentricity of conic sections in 3
dimensions, normalized to the unit range.
49
Measuring Networks with:
1. DTI (Diffuse Tensor imaging)-
employing FA◆ One common approach is the use of DTI
tractography to delineate major white matter tracts, and compute measures such as FA (indicator of microstructural integrity) of the traced tracts at several developmental ages. The resulting age-related changes show the microstructural changes for the traced white matter tracts during development. In one such study, Lebel and Beaulieu used DTI based tractography and region of interest analyses on longitudinal scans of 103 healthy subjects aged 5–32 years (each volunteer scanned twice) to describe the exponential changes in the microstructural development of white matter (using FA) from childhood to adulthood (Lebel and Beaulieu, 2011).
50
Research findings indicate increasing
integration and decreasing
segregation of structural connectivity
with age indicating network-level
refinement mediated by white matter
development.
51
Measuring Networks with:
2. rs-fMRI◆ Resting state fMRI (rs-fMRI) is a functional imaging
technique that permits measurement of spontaneous, low-frequency (0.1 Hz), and high-amplitude fluctuations while subjects are ‘at rest’ (not performing any overt task)
◆ Functional connectivity as assessed by the correlation of rs-fMRI signals has often been observed among functionally associated brain areas and is present even under anesthesia (Gusnard et al., 2001; Dosenbach et al., 2007; Fair et al., 2007, 2008; Greicius et al., 2009).
◆ Both ‘data-driven’ (e.g. independent component analysis (ICA)) and ‘hypothesis-based’ (a priori seed-selection) approaches have been used to investigate developmental changes in functional connectivity.
52
Measuring Networks with:
2. rs-fMRI◆ On the other hand, hypothesis-based approach starts with
a priori selected seed (brain region) with which functional connectivity is computed for the rest of the brain regions.
◆ Developmental trajectories from late childhood (8–12 years) through adolescence (13–17 years) to early adulthood (19–24 years) of 5 functionally distinct cingulate-based intrinsic connectivity networks (ICNs) -5 domains of self-regulatory control: i)motor control, ii) attentional/cognitive control, iii) conflict monitoring and error processing, iv) mentalizing and social processing, and v) emotional regulation.
◆ A pattern of diffuse local functional connectivity in children while more focal patterns of functional connectivity were seen in adults, consistent with developmental patterns of activation seen in functional neuroimaging studies that move from diffuse to more specific/focal patterns (Durstonand Casey, 2006; Durston et al., 2006).
53
Pros and Cons to
Traditional methods◆ Pro: Both hypothesis-based (seed-based) and ICA
approaches have proved very useful in studying specific networks or modules in detail.
◆ Con: Misses the bigger picture of how specific networks or modules interact with the remaining brain regions. This becomes especially relevant considering the fact that the coordinated activity within and across modules produces large-scale brain networks that are essential for efficient functioning of the brain
◆ E.g. studying a particular brain region, say, the anterior cingulate cortex (a seed region for the salience network) within the context of only the salience network will likely mask its function in default mode activity.
54
Pros and Cons- Example
Central Executive Network( Frontal Parietal) -key nodes
◆ dlPFC & PPC----memory and attention
Salience Network-key nodes
◆ frontal insular and dACC---detecting mapping internal and external events relevant to context. Connects w subcortical-limbic structures critical to reward and motivation
Cingulo-Opercular-key nodes
◆ dACC, ant insula, antPFC
DMN-key nodes
◆ PFC and PCC; self-referential, autobiographical, memory for scenario planning, moral decision making
55
Graph-based connectivity:
Developmental Studies
◆ There appears to be a developmental trajectory toward increased structural connectivity with development, consistent with white matter maturation.
◆ Graph-theoretic studies based on SCNs have revealed that brain networks in early development (1month) are stable exhibiting economic/optimal small world topology and non-random modular organization and show increased global efficiency and modularity in early development (Fan et al., 2011). Khundrakpam et al. (2013) showed that this stable organization continues in childhood and adolescence (Khundrakpam et al., 2013).
57
Graph-based connectivity:
Developmental Studies
◆ During late childhood, prominent changes in global topological properties, specifically a significant reduction in local efficiency and modularity and increase in global efficiency, suggesting a shift of topological organization toward a more random configuration.
◆ The studies are confounded by:
◆ fMRI studies used selected age ranges and selected brain regions
◆ Comprehensive statistical comparisons were not performed for the global topological properties
◆ Since these graph theoretic studies on development are from different imaging modalities that capture different tissue types and brain structures, different developmental trajectories of the global topological properties might be expected
58
Utilization of Connectivity in
Clinical Disorders
59
60
Heredibility
◆ Disrupted functional connectivity in frontal lobe was associated with common genetic variants implicated in several neurodevelopmental disorders including autism (Scott-Van Zeeland et al., 2010).
◆ A study using longitudinal DTI data of 162 healthy adolescent twins and their siblings showed that the efficiency measures (global and local efficiency) of structural brain networks are highly heritable (Koenis et al., 2015)
◆ The efficiency measures increase during early adolescence relate to increase in intellectual abilities.
61
Conclusions from
Developmental Connectivity
◆ 1. At the cellular level, synaptogenesis and synaptic pruning act as progressive and regressive forces, beginning in primary sensorimotor regions and later in anterior regions such as prefrontal cortex (Huttenlocher, 1990; Huttenlocher and Dabholkar, 1997), thus continuously shaping the formation and evolution of neural circuits.
◆ 2. In parallel to these synaptic level changes, brain structure and function also undergo progressive and regressive events (i.e., WM myelination and GM loss, respectively) at the macroscopic level, starting earliest at primary sensorimotor areas and occurring latest in higher-order association areas.
◆ 3. Intrinsic functional connectivity exhibits a shift from diffuse local functional connectivity in children to more focal patterns of functional connectivity in adults
◆ 4. The dynamic process of synaptogenesis and pruning that rewires connectivity at the neuronal level also operates at systems level helping to refine network connectivity in the developing brain.
62
Network
Communication
63
Communication within
and Between Networks
◆ While the afore studies help to understand the
“hard wiring” for structural networks facilitating
communication in these networks, how is specific
information encoded in these networks to direct
information processing and adaptive responses –
in other words - learning?
64
Is Learning “hard-wired”?
Hierarchical development
promotes more expansive
“learning”
◆ Introduction of stimulation on a system trying
“calibrate” or coordinate events in the ANS fundamentally limits adaptability by the CNS
Ex: Newborn study – competing reflexes
65
66
Functional Hierarchy of Functional
System:
= Historical (Genetic)
+ Neural
= Sensory/Perceptual
= Cognition/Physical Mngmnt
= Communication/Independent Care
= Psychosocial /Occupational Success
67
Definition
◆ Sensory Processing (Integration) Disorders refers to the body’s way of handling and processing inputs from the environment
◆ Jean Ayres, Ph.D. (Occupational Therapy) – Ayres, 1979.
◆ Result of an aberrant developmental process
◆ Estimated to affect 5-16% of children and this can cause long-term impairment of intellectual and social development from disrupted processes attempting to integrated “high-bandwidth” information from multiple sensory modalities – Owen et al. ,2013
◆ White matter problems
◆ Co-morbid with ADHD, ASD, and other pathologies BUT often exists in isolation (Ahn et al., 2004).
68
Communication:
Backtracking from
Experience
◆ Coordination of stimulus “perception”
comes from time locked redundancies of
stimulus events (classical conditioning)
and the extent that responses to events
lead to successful adaptive functioning
(Operant conditioning)
69
What is defining temporal
order in neural firing?
◆ Metabolic periodicity “coding”
70
Biological Rhythms- Cellular
Processes
◆Goodwin (1967) –continuous oscillations of cellular constituents establishes relative stable states of equilibrium despite minor fluctuation in environmental events.
71
◆ The production and depletion of
metabolites serves to regulate synthesis of
cellular proteins and the utilization of these
proteins in a dynamic manner.
◆ Not random but rather occur in patterns of non-
linear oscillations and can be described by a set
of differential equations.
◆ These equations describe cellular volumetric
shifts as function of the utilization curves of
cellular proteins.
◆ The time gradient in these utilization curves
describes the oscillatory mode of general
cellular activity
72
◆Negative feedback control loops play critical role in regulation of number of oscillations per unit time,,i.e. concentration of argenine, CTP, pyruvate, etc that spill out into intercellular space with certain periodicities establish temporal signals for neighboring cellular activity◆The frequency and phase
relations of these signals can establish a “code word” which would have specific effects on cells or axons receiving it.
73
Cellular communication
◆ It has become apparent that the complex
“biochemical” activities which underlie the
structure and function of cells and organisms do
not form a homogeneous pattern in time such
that all processes occur simultaneously at fixed
rates. Rather there is a rhythm to these activities
whereby they are ordered relative to one another
in time, first one and then another activity rising to
a maximum and then falling off again.
74
Metabolic “Turnover” is
the “rate” limiting factor
One of the major determinants of how rapidly steady states can be reached in the metabolic system, is the turnover rate of substrate molecules by the enzymes of intermediary metabolism. This falls largely in the range of 10-10"* molecules 'sec. (cf. Eigen and Hammes, 1963). The detailed studies of Chance and Hess (1959), Hess and Chance (1961), on changes in the pattern of glucose metabolism in ascites tumour cells following different disturbances show that very extensive changes in metabolic steady state occur in a matter of 1 or 2 min in response to large stimuli. For example, the level of glucose-6-phosphate rises from a very low value (~005 /LtM/g cells) to a new steady state value of about 0-8 /LiM/g cells in about 1 min after the addition of 7-5 mM of glucose to the system.
75
Stahl, 2002
Time course
for cellular
processes to
facilitate
neural firing
pre-
determines
firing rate
76
◆Palva and Palva (2012)
◆have determined that the “UltradianRhythm” (< 0.01 hz) in EEG recordings, BOLD signals, neuronal activity levels, and behavioral time series are likely to image the same fundamental phenomenon; a superstructure of oscillatory ISFs that regulate both the excitatory level of functional networks and the integration between them.
77
◆A complex communication “field” would be set up in terms of frequency, amplitude, and phase relationships among metabolic signals over a population of similar cells [ or cells utilizing similar chemical components]
78
Coupling of higher-frequency oscillations to ISFs. In EEG recordings,
amplitudes of 1–40 Hz oscillations (colored lines) modulate according
to the phase of an ISF (0.01–0.1 Hz) and also mirror changes in
behavioral performance (black line). ISFs inhabit a similar frequency
range as that of the fluctuating BOLD signal, suggesting the latter
may bear a similar phase-relationship to higher frequency oscillations
and behavioral performance. Reproduced with permission from
Monto et al. (2008). ISF, infra-slow fluctuation.
Taken from Meehan et. al, 2012
79
80
◆Kandel & Schwartz (1982) –changes in neural cells underlie the basis of learning and memory:◆The learning/memory of an event
resides in the time course of the chemical processes utilized in facilitating the events juxtaposed on other ongoing cyclical events
81
◆Varela, John and Schwartz (1978)
noted that photic stimulation
differentially phase locked to the
alpha rhythm determined the
perception of two contiguous flashes
as one flash, one light in motion, or
as two successive light flashes. ◆ Background alpha serves as “temporal template” . A
significant correlation between percent alpha or
frequencies which were harmonic or sub-harmonic in alpha as noted in EEG and the accuracy of time related
tasks.
82
Proposed relationship between
oscillatory phase, neuronal excitability,
and stimulus events. (A) A putative
relationship between LFP phase of a
local neuronal population and action
potential firing rate, which here is high
during optimal and low during non-
optimal phases. (B) When
superimposed, traces from single trials
show random ambient phases, which
reset and align following a stimulus
presentation (time zero, arrow). (C)
Response amplitudes to sensory
events are highly variable during the
pre-stimulus period, are enhanced
during optimal phase, and are
suppressed during non-optimal phase.
(D) A stereotypical complex
waveform, when broken into its
component frequencies, reveals low-
frequency phase modulation of
higher-frequency oscillation
amplitude, or phase-amplitude
coupling, in a nested fashion.
Reproduced with permission from
Schroeder et al. (2008). LFP, local field
potential.
Taken from: Neurocognitive
networks: Findings, models, and
theory Timothy P. Meehan, Steven L.
Bressler, 2012.
83
Adaptability is achieved by
temporal coupling in networks
underlying harmonics
84
Meehan et al., 2012◆ Beta and gamma oscillations appear to have different
functional capabilities in cortex. The switch from co-existent gamma and beta2 to global beta1 oscillations implies that the cortex is able to: (1) manifest temporal memory as ongoing beta oscillations; (2) engage in long-distance coordination by phase synchronization of those oscillations across cortical regions; and (3) bind multimodal features by interregional phase synchronization without competition between inputs (Kopell, 2010). Since experimental evidence shows that beta oscillatory phase synchronization is involved in cognitive functions such as visual short-term memory (Tallon-Baudry et al., 2001), sentence comprehension (Weiss et al., 2005), and sensory-motor integration (Lalo et al., 2007), these functions appear to be dependent on the dynamic properties of cortical cell assemblies, which may represent the nodes of neurocognitive networks.
85
Connection Basics –Phase
Couplings
Cross frequency PR uses same model; when 2nd derivative
between two oscillators =0, then in PL
86
Connection Basics –
Developmental, Phase Reset
87
PSD & PLD-Perceptual
Frame
◆ To distinguish events successive in time need 40
ms-auditory & 140 ms for visual stimuli,
(Varela,1995).
◆ Learning from discontinuous sequence of narrow
time windows.
◆ PSD & PLD are on going related to external events
or background emergent process of approx. 40-
80 sec (uncertainty/chaos) PSD, 150- 800ms (
certainty/stability)PLD
88
PSD & PLD-Perceptual
Frame
◆ During phase reset large assemblies of neurons
may be in a simulated “refractory” period
because locked neurons are not available for
reallocation called by a different cluster.
◆ When PLD occurs over long distance it reduces
the size of the cluster of idling neurons-
synchronous high amplitude.
◆ Seen when stimulus lock PR causes Event Related
Desynchronization (Klimesh,2007) ie spatially
distributed/differentiated micro binding= < idling neurons,= less amplitude post stimulus.
89
Connection Basics –
Developmental, Phase Reset
◆ Phase reset is comprised of phase shift duration
(PSD) and phase lock duration (PLD)
◆ PSD is a period of instability, due to minimally
coupled oscillators, on the edge of chaos/collapse.
◆ PSD allows for an energy free release of existing
(cross)frequency couplings within and between
areas, and a call out to available neuronal
resources.
◆ Longer PSDs associated w higher I.Q. ( recruitment of
sufficient neuronal resources for processing); shorter
PSDs - insufficient resources proposed to account for
perseveration in such disorders as autism.
90
Connection Basics –
Developmental, Phase Reset
◆ Phase lock duration (PLD), a time period of
stability-the binding of neuronal units within and
across frequencies for information processing.
◆ Shorter PLD intervals associated with higher I.Q.;
long PLD intervals associated with less efficient
processing.
◆ High correlation between PLD and Coherence
because COH is the consistency of phase
relationships which contributes to stability.
◆ Inverse relationship between PSD & PLD; low R2
91
Autism and EEG Phase Reset: Deficient
GABA Mediated Inhibition in Thalamo-
Cortical Circuits
◆ Thatcher et al., 2009
92
93
94
95
96
Connection Basics 97
Thatcher, 2015
(Thatcher, 2015)
98
A Role of Phase-Resetting in
Coordinating Large Scale
Neural Networks During
Attention and Goal-Directed
Behavior
Benjamin Voloh and Thilo Womelsdorf
REVIEW PUBLISHED: 08 MARCH 2016 DOI:
10.3389/FNSYS.2016.00018
99
◆ Short periods of oscillatory activation are ubiquitous signatures of neural circuits. A broad range of studies documents not only their circuit origins, but also a fundamental role for oscillatory activity in coordinating information transfer during goal directed behavior. Recent studies suggest that resetting the phase of ongoing oscillatory activity to endogenous or exogenous cues facilitates coordinated information transfer within circuits and between distributed brain areas. Phase resets: (1) set a “neural context” in terms of narrow band frequencies that uniquely characterizes the activated circuits; (2) impose coherent low frequency phases to which high frequency activations can synchronize, identifiable as cross-frequency correlations across large anatomical distances; (3) are critical for neural coding models that depend on phase, increasing the informational content of neural representations; and (4) likely originate from the dynamics of canonical E-I circuits that are anatomically ubiquitous. These multiple signatures of phase resets are directly linked to enhanced information transfer and behavioral success. Phase resets re-organize oscillations in diverse task contexts, including sensory perception, attentional stimulus selection, cross-modal integration, Pavlovian conditioning, and spatial navigation.
10
0
10
1
Example: Language
Development
◆ 1. Reproducing sounds (sound over time only)
◆ 2. Reproducing words (complex sounds over time
only)
◆ 3. Reproducing phrases (series of complex sounds
without context)
◆ 4. Utilizing phrases in the right “contexts” – now
communication
◆ 5. Applying phrases to generalize across contexts
– higher order – adolescent language- creating a “new language”
10
2
Building Network Clusters
◆ Emergence from
perceptual/emotional networks to
memory networks which provide
frames of references which enable
response networks as part of the
decision making process (all of this
happens in milliseconds –
precognition?).
10
3
Cognition (Schemas)
= Historical (Genetic)
+ Neural
= Sensory/Perceptual
= Cognition/Physical Mngmnt
= Communication/Independent Care
= Psychosocial /Occupational Success
10
4
Meehan et al., 2012
◆Neurocognitive networks may have
a nested structure, whereby nodes
contain levels of complexity at
progressively more microscopic
scales, and processes at all these
lower levels may contribute to
function at the macroscopic scale.
10
5
Connection Basics
Thatcher, 2015
10
6
Recognizing sensory/temporal
redundancies is the root of
perception and learning
◆ The redundancy of two events in time allows us to
“associate” the events and allows us to
incorporate this association into or world of
schemas or knowledge. When the associated events are responses themselves and these
responses “compete” the system must either
accommodate them adaptively or if non-
adaptive response cannot be realized – we
experience distress
10
7
Sensory- Perceptual
Schemas
◆ The system first develops network “schemas” by
the type of experiences we have in the real world
impinging on the neural system over and over
again so these neural systems “learn” to have
expectancies of how the world should work.
10
8
Neuroception(FROM STEPHEN PORGES)
10
9
https://youtu.be/wzP9X5Eclm8
Brain Games 1 Watch This (this is mind blowing Perception)
11
0
Piagetian Concepts
◆ Assimilation is a process that manages how we take in new information and incorporate that new information into our existing knowledge.
◆ Piaget used the term schema to refer to a category of knowledge that you currently hold that helps you understand the world you live in and provides some basic guidance for future events. A schema describes how we organize information. We store information as a particular schema until it is needed.
◆ The process of accommodation involves altering one's existing schemas, or ideas, as a result of new information or new experiences. New schemas may also be developed during this process.
11
1
11
2
Higher Ordered
Developmental Shift
◆ Networks the define and coordinate “perceptual”
organization and thereby define cognitive “sets”
give way to enable using learned concepts and
abstractions to “over-rule” network neuro-caption
as a decision making process about how the
world works and how to respond in such a world.
◆ This facilitates alternative response thinking –
executive functioning – problem solve
11
3
Piagetian Development11
4
11
5
Level I –
Perceptual/Emotional
Level II – Memory Matching -
Assimilation
Level III- Decision Making
Response/Accommodation
11
6
Day 1 Final Conclusions
◆ There are definitively structure-function
relationships that evolve into networks optimizing
adaptability and “survivability”
◆ While methods describing the development of
gray matter and white matter in the human brain
have been useful, the details of “temporal
coding” and directionality within and across
networks is still in infancy.
◆ The ability to refine optimized interventions is
confounded by dynamic nature across age and across individuals including factors of metabolic
variance due to dietary/chemical intrusions
11
7
Day 1: Final Conclusions (2)
◆ Behavior adaptation and response is a result of our histories and experiences represented in nested networks
◆ ALL of our constructs are a function of the brain
◆ We define our reality by experiencing things in context- we have expectations
◆ We “feel” things that also based on these expectations
◆ Common Experiences result in common behaviors and ways we come to expect how people will respond and react to a given situation –acceptable “social” behavior
11
8
Problem for EEG
Neurofeedback Therapy
Protocols:◆ The structure-function landscape keeps changing in course of
development when training a specific structure-function relationship in a child, we may be assuming certain maturational “states” of these networks that simply do not exist
◆ If genetics plays a significant role in developmental connections at a given age and the “efficiency” of these networks, what are the “limits” in an individual to which we can expect to alter these intrinsic networks and their efficiency
◆ An important limitation of the available fMRI studies is that hemodynamic signals only provide an indirect measure of neuronal activity. In the contrast, electroencephalography (EEG) directly measures electrophysiological activity of the brain. Little is known about the brain-wide organization of such spontaneous neuronal population signals at resting state. It is not entirely clear if or how the network structure built upon slowly fluctuating hemodynamic signals is represented in terms of fast, dynamic and spontaneous neuronal activity.
Day 2: Review
and challenges
with regard to
training
Single Cell to Networks
Single Channel Training:
The Basis
◆ Single channel training has been the basis for the
majority of research showing efficacy of NFB to
date.
◆ Over two hundred fifty studies.
◆ Efficacy level is very high.
◆ Can we justify more complex approaches?
◆ Does our knowledge base support such
approaches?
◆ Do we have evidence of comparable efficacy?
◆ Do we have evidence of enhanced efficacy?
Is Training Networks or ROIs
A Reasonable Approach?
◆ Can we identify Networks at the right level?
◆ Does EEG have the spatial resolution we need?
◆ Can we isolate specific network structures related to specific
functions?
◆ Do we know how networks interface?
◆ Do we understand how their spatial and temporal function?
◆ Do we understand up and downstream influences?
◆ Can we predict the impact of a protocol?
Micro Structural Considerations
The Present Structural
Theory Sporns & van den Heuval,
2013
Limitations of Graph TheorySporns & van den Heuval, 2013
“it should be noted that current graph-based
analyses of communication cannot fully predict
dynamic (i.e., time-varying) patterns of
communication. Factors influencing the dynamics of
neuronal time series such as local firing rates of
neurons and/or level of activity, external inputs or
task demands, coherent phase relationships, or
synaptic efficacy are generally not incorporated into
current graph analyses.”
Limitations of Anatomy
“knowledge of anatomical connectivity in the
human brain is still embryonic”
(Sporns et al., 2005)
“At NCN 2010, Sporns compared the current state of
the human connectome to a map of the world circa
1570 (Sporns, 2010)”
Meehan & Bressler (2012)
Can We Use DTI Fiber
Pathways
▪ We don’t know how the networks map to the
fiber pathways of the brain.
▪ There is no standard definition of an edge or
node.
▪ Anatomical connectivity constrains functional
connectivity, but cannot account for it fully.
Meehan & Bressler (2012)
The Hub System-Fiber Path vs Synaptic
MapThe DMN Cortical Network
8 Anatomical Subregions
◆ Posterior Cingulate
◆ Precuneus
◆ Cuneus
◆ Paracentral Lobule
◆ Isthmus of the Cingulate
◆ Superior Temporal Sulcus
◆ Inferior Parietal Cortex
◆ Superior Parietal Cortex
Highest elevated fiber counts &
densities(node degree and
strength)
“To fully understand the basis for such cases
will require a precise synaptic connectivity
map of the human brain; however, such
a map does not yet exist, and fiber-tract
imaging (DTI and DSI) may never be able
to produce one.” (Freeman, 2005)
ANATOMICAL AND FUNCITIONAL
CONNECTIVITY RELATIONSHIP: STILL
UNCLEAR
Honey et al. (2009) found only a partial agreement
between anatomical and functional connectivity, with significant functional connections occurring
between areas where no underlying direct
anatomical connection was detected.
Micrograph of the cell body
with synaptic inputs.
As many as 30-50K inputs per
cell in the mammalian brains
are seen.
Cerebellar cells may have as
many as 100K inputs.
Parcellation Limitations of
Brodmann
Different
methods
provide us with
different
configurations
Parcellation Issues-
Anatomical or Functional◆ Many “imaging” modalities in identifying nodes use different measures—
metabolic (PET), blood (MRI,fMRI,rs-fcMRI), spectral power (LFP, EcoG, EEG, MEG)
◆ Time domain of MRI is several seconds, EEG –milliseconds; hence different resolution due to time dampening (does MRI translate to qEEG?)
◆ Correlation of heightened activity ( under task) of a node identified by one modality not fully understood how relates to node identified by different modality. Various advantages- model constructions from each ( Zalesky et. a. 2010)
◆ Network not simply defined by co-activated nodes, must also identify edge
◆ Node could be defined not only by co-activity but by increased correlation with other node while activation remains constant
◆ Use of different values for parameters results in different nodes and connectivity patterns, determines to which network(s) a nodal areas is associated
◆ Measure of single neural unit spiking identifies that 30% of neurons
133
Limbic & Brainstem Study
Deficits
◆ Most anatomical parcellation studies have
focused on the cerebral cortex. Less attention has
been paid to sub- cortical structures such as the
basal ganglia and the thalamus, which have only
been demarcated at a coarse level using sMRI.
Brainstem systems mediating motivation,
autonomic function and arousal have been
poorly studied because they are notoriously
difficult to identify using in vivo techniques.
Nonetheless, it is important to identify these
structures because they significantly influence cortical signaling and thus affect cognitive
function.
ROI’s Have Weak
Identities
More recently, fMRI activation studies have been
used to more precisely demarcate nodes of specific
functional circuits associated with such dedicated
networks. However, regions of interest (ROIs)
identified in this manner tend to vary considerably
with task demands, patient groups used, and the
specific control or baseline conditions used to
identify them. As a result, uncovering the nodes of
neurocognitive networks in a principled and reliable
manner has turned out to be elusive.
Menon, 2011
Harmonic Interface?
There may
be a whole
level of
network
interface
that is not
fiber
based.
Functional Networks- Hubs
From Structure To Function
◆ Functional nodes were first determined by lesion studies, then fMRI- Localizationist Perspective. (discrete regions showing elevated metabolic activity (in PET), blood flow (in fMRI), spectral power (in LFP, ECoG, or MEG), or elevated firing rate (in single-unit or multi-unit recording) in correlation with performance of a cognitive function.
• Structural networks provide a complex architecture that promotes the dynamic interactions between nodes that give rise to functional networks- Network Perspective.(changes in spectral coherence between distributed neuronal assemblies may occur without changes in spectral power.)
A meso level hub can be a macro level node.
Freeman, 2014
Sources & Sinks
Studies examining the total sum of afferent and efferent connections of hub regions in such data sets have suggested that some cortical hub regions maintain an unequal balance of incoming and outgoing projections. This imbalance suggests a potential role for these cortical hub regions as neural communication ‘sources’ and ‘sinks’.
Source hubs include central brain ROIs of attentional networks
◆ dorsal prefrontal, posterior parietal, visual and insular cortex
Sink hubs (driven/receptor) include
◆ posterior cingulate, precuneus, and medial frontal
Driven vs Driving Hubs Directionality: Incoming vs
outgoing projections
◆ Posterior cingulate, precuneus, and medial
prefrontal cortex are driven hubs or sinks.
◆ Dorsal prefrontal, posterior parietal, visual and
insular hubs as driving hubs or sources.
Some areas act as controllers such as pre frontal and parietal
control areas which channel the flow of sensory and motor
actvities.
van den Hueval & Sporns
Should we be driving either category in particular with NFB?
Seed RegionsAlternative Approach To ICN
Delineation◆ Based on functional interdependence analysis
◆ First, an fMRI seed region associated with a
cognitive function is identified (active on task).
◆ Then a map is constructed of brain voxels showing
significant functional connectivity with the seed
region.
Structural Hubs & Node
ComponentsExamples In Brodmann Terms◆ Precuneus : BA-7,19,31,39
◆ ACC & BA-25,24,32,33,10
◆ PCC: BA-23,29,30,31
◆ Insular cortex BA-13
◆ Frontal cortex :BA-6,8,9,10,11
◆ Temp cortex:
◆ Lat- BA 38,20,21,22,37,40,41, 42
◆ Inf-BA 20,21, 22
◆ Sup-BA 13,21,22,38,39,41,42
◆ Transverse- BA 41, 42
◆ Middle-BA 20,21,22,37,40,6
◆ Lateral parietal cortex -BA 5, 7
14
2
Genardi
In Summary
Area Of Direct
Conduction T3-T4
60,000,00 Neurons
10-14k Connections
Each
Is our spatial
resolution
sufficiently
parsimonious ?
MRI Location Functions
Fiber Pathways & Metastable Networks
Timescale Conundrums
Cognition, and the neurocognitive networks that support it,
evolve continuously in real time on the timescale of milliseconds.
They constitute “a rapid sequence of spatial distributions of
electric potential, known as ‘microstates’ (Lehmann et al., 1998)
or ‘frames’ (Freeman,2006), which remain stable for around 100
ms and then transition abruptly to the next state (Freeman,
2006; Lehmann and Michel, 2011; Lehmann et al., 2009;
Lehmann et al., 1998) (Freeman, 2005).
SLOW TIMESCALE CONSTRAIN FAST TIME SCALES
Where neurocognitive network interactions undergo the
instantaneous transformations that manifest cognitive function
under constraint. (The constraint is provided by longer and
slower networks.)
Large Scale Networks
Network Paradigms
◆ Various authors from different research
paradigms, using different analytic and
conceptual approaches, arrive at different
network identities and configurations.
◆ These reflect different levels and methods of
analysis.
◆ They have considerable overlap and converge
over time.
Menon, 2015
Large Scale Connectivity
Short path = physical fiber (not graph)
Six Major Principles of
Large Scale Studies
First, large-scale functional organization is characterized by a non-random, small-world, modular global brain architecture with strategic hub regions that regulate communication among different functional systems (Sporns, 2011a, 2011b).
Second, strong interhemispheric connectivity between homotopic regions, with a gradient of decreasing left–right connectivity from sensory to association and heteromodal cortices, is a prominent feature of large-scale functional brain organization (Ryali, Chen, Supekar, & Menon, 2012; Stark et al., 2008).
Third, the human brain is intrinsically organized into coherent functional networks (Bressler & Menon, 2010), with brain areas that are commonly engaged during cognitive tasks forming brain networks that can be readily identified using intrinsic functional connectivity (Damoiseaux et al., 2006).
Principles Continued◆ Fourth, functional brain organization is characterized by task-
and context-dependent activated and deactivated brain systems, pointing to bottlenecks in parallel processing and temporally restricted access to neural resources (Fox et al., 2005; Greicius et al., 2003; Greicius & Menon, 2004; Honey, Kotter, Breakspear, & Sporns, 2007).
◆ Fifth, the most widely deactivated regions form a coherent large-scale network, termed the default-mode network, which is a tightly function- ally and structurally connected system important for self- referential information processing and monitoring of the internal mental landscape (Greicius et al., 2003; Greicius, Srivastava, Reiss, & Menon, 2004; Qin & Northoff, 2011).
◆ Sixth, core prefrontal–parietal control systems can be dissoci-ated into distinct brain networks with distinct roles in cognition. Notably, the salience network, anchored in the insula and anterior cingulate cortex, is a system that plays an important role in attentional capture of biologically and cognitively rele- vant events while the lateral frontoparietalcentral executive network, anchored in the dorsolateral prefrontal cortex and supramarginal gyrus, is important for the working memory and higher-order cognitive processes (Menon & Uddin, 2010; Seeley et al., 2007; Sridharan, Levitin, & Menon, 2008).
15
2
Menon,2011
Leech- 14
ICNs
Laird Interconnectivity
Networks
15
3
Genardi
Laird Networks Identified
Task Based Co-operation of
Primary Networks & Hubs
In task-based functional
imaging, coactivations of
the anterior insula, anterior
cingulate cortex, and the
dorsolateral and the
ventrolateral prefrontal
cortices, as well as the
supramarginal gyrus,
intraparietal sulcus, and
superior parietal lobules of
the lateral parietal cortex
are common across a wide
range of cognitive tasks.
Network SelectionEach Node/Hub Has
Complex Network Interface
Do we have the norms to correctly adjust each network node?
Do we need to adjust micro level nodes?
Metastable Network
Competition“findings appear to confirm that distributed networks
underlie
cognitively demanding or executive capabilities, and
that these networks arise dynamically with task
demands.”
“transitions between metastable states are facilitated by
competition. Different states of neurocognitive networks
are conceived to be in continuous competition.”
(Rabinovich et al., 2008)
Convergence Zones
Van den Heuval, M., Sporns, O. (2013)
“the connective core hypothesis’suggests that
interconnected hub regions that are topologically
central offer an important substrate for cognitive
integration, not only for broadcasting and dynamic
coupling of neural signals but also by offering an
‘arena for dynamic cooperation and competition’
among otherwise segregated information “
Convergence Zone Issues
“hub regions across areas of the cortex in which
multiple functional domains overlap, forming
‘confluence zones’ or ‘convergence zones’ of neural
interactions” (van den Hueval & Sporns)
Network Exclusivity:
Multimodality
If we correctly adjust a network will it
adversely influence one of its alternate
functions?
Resting State Networks (ICN): 6 out of 7 Use Temporal Lobe Nodes
Can I separate just one of the networks out to train?
The 01-02 Anxiety SurpriseUnanticipated Consequences of
Training◆ For example, BA area 19 is not just “a” visual region
but actually a mosaic of different regions belonging to the extrastriate visual cortex (Orbanet al., 2004)
◆ This region has inputs to limbic regions as well as the amygdala.(Wang et al 2012)
◆ Soutar trained this area and documented that very robust reductions occurred in anxiety across clients and clinics as measured with a a normed anxiety instrument.
ICNs- Large Scale NetworksCo-ordinating Specific Functional
Nodes◆ AI- Anterior Insula Network- Self Awareness, Switching
◆ ACC- Anterior Cingulate Network- Attention, Reward Anticipation, Impulse Control
◆ DLPFC- Dorsolateral Prefrontal Network- Working Memory, Planning, Abstract Reasoning
◆ CEN- Central Executive Network- Memory & Attention
◆ PPC Posterior Parietal Cortex Network- Emotional and self-Referential Processing
◆ MTL-Medial Temporal Lobe Network- Episodic Memory, language
◆ DMN- Default Mode Network
◆ FPN-Fronto-Parietal- Rule based problem solving, Sensory contents of attention. Select spatil & category. Select behavioral relevant info-control network.
◆ PCC –Posterior Cingulate Gyrus- Lnguage, Memory, ST Memory. Attention Release (Peterson & Posner)
EEG
Correlate
Networks
Chen et al 2012
Derived from ICA,
sLORETA and
Coherence Analysis
combination.
Homologous networks
highly dependent on
corpus callosum with
increased FP
intrahemispheric
activity in the EO state.
Match MRI findings.
3 Major ICNsFunctional Hubs
◆ Central Executive- Memory & Attention DLPFC, PPC
◆ Default Mode- Autobiographical, self-monitoring and social cognitive
functions AI, DACC, PFC, PPC, MTL
◆ Salience- Modulates autonomic reactivity to salient stimuli AI,ACC,
DACC
◆ Note different nodes/hubs based on different imaging anaysis.
Subcortical Connections
CEN & SN have subcortical connectivity in the
◆ anterior thalamus (antTHAL),
◆ dorsal caudate nucleus (dCN),
◆ dorsomedial thalamus (dmTHAL),
◆ hypothalamus (HT),
◆ periaqueductal gray(PAG),
◆ putamen (Put),
◆ sublenticular extended amygdala (SLEA),
◆ SuN/VTA
◆ temporal pole (TP).
16
6
Genardi
Subcortical InputsMemory and Emotion
If we choose a network for a Brodmann function will it
adversely affect a related but unacknowledged
subcortical function?
Modular vs Large Scale
ParadigmsBressler & Mennon, 2010
◆ Modular perspective expected functions to be
isolated to specific ROIs
◆ New Data indicates cojoint function of brain areas
is more accurate portrayal
◆ Cognitive function is emergent and constrained by
core structural and functional aspects of networks.
◆ A process of excitatory vs Inhibitory subnetworks-
nodes
Process vs Dynamic
Networks
◆ Processing-type networks are considered more
modular and static.
◆ Control-type networks are hypothesized to be dynamic and flexible, with an ability to adapt to a
wide variety of tasks.
Information Processing: Intro into
what we need to do to learn
Global Workspace
With hub nodes and their connections attracting and disseminating a large number of all neural communication paths, brain hubs and their connections, as a system, have been hypothesized as a convergent structure for integration of information, together forming a putative anatomical substrate for a functional ‘global workspace’.
Such a workspace is hypothesized as a cognitive architecture in which segregated functional systems can share and integrate information by means of neuronal inter- actions, with an important role for pathways that link central regions and constitute a global workspace.
Salience & Switching
Field Theory Perspective
The critical inference to be drawn from the
multichannel EEG data is that phase transitions occur
over large fractions of each cerebral hemisphere, by
which the oscillations of populations of neurons in the
beta and gamma ranges are repeatedly reinitialized, resynchronized within very few milliseconds [Freeman,
2005], and then restabilized with a new AM
(amplitude modulation) pattern for 3-5 cycles of the
center frequency of the shared carrier wave, while
the intensity of pattern transmission rises to a
maximum [Freeman,2004, 2005].
Switching Internal To
External
◆ AI switches between CEN for external processing
and DMN for internal processing
◆ Leech et al. ( 2013) found that the ventral PCC
has high connectivity with the DMN when
attention is focused internally; whereas the dorsal
PCC shows increased functional connection with
the DMN and concomitant anti-correlation with
the DAT for external attention task
DMN & SwitchingEC vs EO Maps
Eyes closed main cortical generator in the posterior cingulate. alpha oscillations generated to all other regions.
Eyes open: anterior cingulate activates with delta and beta dominating-as a sender of mainly theta-alpha oscillations to the dorsolateral pre-frontal cortices?
17
5
Phase Shift and Phase
LockSwitching dynamics referred to as "Phase Shift" and "Phase Locking"
also called "Phase Reset" are a macro-to-meso level transition metric
in which phase shift recruits available neurons
Phase lock is the binding or synchrony of groups of neurons that
simultaneously mediate different functions in different brain regions.
Short distances are exponentially related to long phase lock
durations and short phase shift duration
Long distances are exponentially related to long phase shift durations
and short phase lock dura-tion. (local networks are likely more
devoted to a small set of calculations whereas long distance
connections synchronize large scale networks (rich club) that draw on local resources.
(Ermentrout and Kopell, 1994; Ko and Ermentrout, 2007; Thatcher et al, 2014)
17
6
Genardi
Salience Afferents &
Efferents
DMN
LORETA Version DMN
DMN
Its key nodes have been variously linked to: ◆ Episodic memory retrieval (Sestieri, Corbetta, Romani, &
Shulman, 2011; Vannini et al., 2011),
◆ Autobiographical memory (Dastjerdi et al., 2011; Spreng et al., 2009), and
◆ Internal speech (Binder, Desai, Graves, & Conant, 2009),
◆ Specific nodes in the medial prefrontal cortex have been differentially associated with self- related and social cognitive processes (Amodio & Frith, 2006; Spreng et al., 2009),
◆ Value-based decision making (Rangel, Camerer, & Montague, 2008), and
◆ Emotion regulation (Etkin, Egner, & Kalisch, 2011). Genardi
FPN-Central Executive Network-CENComposed of approximately 18 independent subregions involving thalamus, occipital, temporal and parietal regions;
Subdivisions: Into control regions in fronto-temporal and thalamus- and-sensory processing areas of the occipital and temporal cortex.
Switches objects based on Interhemispheric Competition
Frontoparietal is an important network (FPN)( Central Executive ( CEN). It is essential in cognitive processing. Anchored in dlPFC & PPC, has strong intrinsic functional coupling.
Select objects of attentional interst based on PPC coding of feature, object quality and category type.
Rule based problem solving, Sensory contents of attention. Select spatial & category. Select behavioral relevant info-control network. The central executive network is critical for actively maintaining and manipulating information in the working memory, for rule-based problem solving, and for decision- making in the context of goal-directed behavior (Koechlin & Summerfield, 2007; Miller & Cohen, 2001; Muller & Knight, 2006; Petrides, 2005)
Maintains and manipulates WM info, involved in rule based problem solving, decisions related to goal attainment. Dysregularities of FPN found in most psych disorders.
It has important sub-functions . The FPN is involved in the top down control of the dorsal attention network, DAT, which is devoted to visual attention and eye movements. DAT includes
◆ intraparietal sulcus
◆ parts of the superior parietal lobes
◆ the medial temporal lobe complex
◆ frontal eye fields
Scolari et al, 2015
Paradigm Alternative
Example FPCN
Another important network is the frontoparietal control network ( FPCN) which is involved in executive control & becomes more active duringTask Positive activity. During Internal Attention as well. It includes
• parts of the frontal pole
• dorsolateral prefrontal cortex
• the rostral anterior cingulate cortex
• supplemental motor area
• the anterior insula
• and parts of the inferior parietal lobe.
18
3
Subcortical Affective
Contributions
Psychopathology & Salience
Autism
Depression
Anxiety
Schizophrenia
Genardi
Emotion & Memory--PCC
According to functional studies the posterior cingulate gyrus/cortex PCC has 14 independent subregions
◆ Not as directly involved in motor initiation as the anterior cingulate gyrus is, still active when learning a complex motor skill.
◆ More frequently activated during language tasks (e.g., lexico-semantic processing) than its anterior segment, but its role in emotion is obvious (e.g., fear conditioning) as well as its participation in different types of memory (e.g., topographic memory, episodic memory, etc.).
◆ The brain areas involved in emotion, mainly the limbic system, including the MTL & cingulate gyrus, are the very same areas involved in memory. Indicating there is a close association between emotion/motivation and memory: only information that is significant from the emotional/motivational point of view is memorized. Emotionally neutral information is usually ignored.
Genardi
MTL Network
◆ Includes:
◆ the hippocampal formation
◆ extra-hippocampal cortical regions
◆ entorhinal cortex (ERC),
◆ perirhinal cortex (PRC), and
◆ parahippocampal cortex (PHC)
◆ Part of DMN
◆ Connects to Orbital Frontal and Posterior Cingulate
◆ Participtes in several cognitive domains, including episodic memory, perception, social cognition, and semantic cognition.
Seed Based Identity Das et al, 2016
MTL Subcortical Inputs
Chen at al,
2015
Cognitive Control
Network Control Paradigm Gu
et al 2015Specific nodes “drive” the
networks into modes or states
of function.
Processes:
• Link multiple sources of
information to solve problems
• Selective retrieval of
information from memory
• Inhibition of inappropriate
behavioural responses
Outcomes: Transitory changes in
patterns of cooperation and
competition between distributed
neural systems, including regions
in attention, default mode,
frontoparietal and cingulo-
opercular networks.
Control Systems
Profiles
◆ Average Controllability Areas-Hubs that require little energy to shift state.
◆ Modal Controllability Areas-Nodes with lower degree of connectivity require more energy to shift to harder to achieve cognitive states, EspExecutive and Attention states.
◆ Boundary Controllability Areas-Manage coupling and decoupling of networks.
Modular structure has been
reported in structural, functional
and dynamic brain networks.
Key Networks By Control
State
Category B is correlated with IQ: Higher B = Higher IQ
Implications of Paradigm
“moving any diseased state to a healthy state is
difficult, even with a complex combinations of drugs,
brain stimulation and cognitive therapies.”
Implication:
Loss of mitochondrial efficiency would lower brains
ability to reach higher energy state and
inflammation and lesions could do the same.
Implications 2
“Moreover, the default mode is the state to which
the brain relaxes back after the task has been
performed, readying the brain to move to new task
states, when the cycle will repeat.”
Implication:
Stermans Aviation Research predicted this resulting
in his theory of required Synchronizaiton between
tasks.
Gyorgy Buzsaki
◆ Network hierarchy is determined by
computational solution- ie no real top and
bottom solution.
◆ Function is highly distributive
◆ Synaptic weighting emerges from auto-
associative attractor networks.
◆ All circuits can sustain autonomous self-organizing
activity.
Self-Organization of
Critical States Issue
◆ The brain operates near the edge of chaos
according to Power Law observations of EEG.
◆ Self organization requires a high degree of error, error detection and error correction.
◆ Cortical networks must be free to engage in
activity outside a normative range in order to self
correct.
Pettersen et al, 2014
Systems Theory & The Brain
We are dealing with a system of micro, meso and macro network structures of profound complexity that operate more like a nested system that generates shifting attractor states defining a field of “network state space.”
Freeman, 2005
A Demo of A Network –
Decision Making
NFB Training StrategiesConsiderations & Conundrums
Goals of Assessment &
Training NFBCan We Do This?
◆ To identify key hubs (or nodes)
◆ To stimulate key hubs (or nodes)
◆ To exercise associated networks to improve function.
◆ Stimulate dendritic growth.
◆ Enhance network growth.
◆ Enhance network connectivity with associated
networks (1 electrode local, 2 electrode distant).
◆ Encourage capillary density & increased perfusion.
◆ Enhance delivery of metabolic resources.
◆ Increase network activity-beta/gamma sympathetic
◆ Decrease activity-alpha parasympathetic
◆ Increase Limbic Input- theta affective-memory
Genardi
Targeting Theories
Select Regions of Interest based on symptoms.
◆ ROIs are defined based on BA research.
◆ Select network paradigm (Hagmann, Laird,
Alternative DTI, etc).
◆ Select networks based on general symptoms.
◆ Train network hubs or nodes.
◆ No maximum limit to number of target variables
trained (one vendor).
Using Symptom- Function
Mapping◆ In neurofeedback we choose training targets/strategies by
compiling convergent data of symptom expression/dysfunction which can be mapped to a specific ROI/network
◆ Variety of simple ( at times laborious) to sophisticated and streamlined mapping of training goal to training target. Inputs to treatment plan include analysis of raw EEG dynamics, qEEG analysis, neuropsychological testing, variety of behavioral, cognitive and emotional assessments.
◆ Excellent examples Neuroguide Sx Checklist w integrated protocol recommendations, Loreta Progress Report, CNC with qEEGPro analysis, NewMind checklists & cognitive assessments w integrated protocol recommendations, BrainDxanalysis and report, Brainmaster live sLoreta analysis and data summary table.
Genardi
Training Strategy Options
◆ Train an ROI or group of ROIs (network) known to support a specific function; train associated parameters to a defined value- Remediation
◆ Train an essential/ area(s) subserving multiple functions/networks. Optimization (Generalized)
Rich Club
DMN
PCC
◆ Train dynamic relationship among key Large Scale networks
DMN, CEN, SN, DAT
◆ Train along some key attribute
Primary, secondary, tertiary cortex
Sources and sinks
Controllability
Genardi
Metabolic Aspects of
TrainingBressler & Menon, 2010
“the elevated excitability of neurons within an area leads to elevated metabolic activity, which in turn causes an increase in local blood oxygen availability. The elevated excitability could also cause increased interactions between neurons within the area. Interactions between different populations can produce oscillatory activity and can have important functional consequences if, for example, the interactions lead to increased sensitivity of neurons within the area to the inputs that they receive. “
Stimulate dendritic growth.Enhance network growth.Encourage capillary density & increased perfusion.Enhance delivery of metabolic resources.
Physiological Accuracy
Issues & LORETA
◆ THERE ARE MICROSTRUCTURAL SUBREGIONS IN BRODMANN AREAS-THEY ARE NOT HOMGENEOUS IN FUNCTION AND SERVE A MULTIPICITY OF EXTENDED NETWORK FUNCTIONS.
◆ FMRI SPATIAL RESOLUTION ONLY RECOGNIZES MACROANATOMY (BA/fMRI findings only match macro).
◆ STRUCTURAL FUNCTIONAL CORRELATIONS BASED SOLEY ON MACROANATOMY ARE QUESITONABLE.
◆ MICROANATOMY IS HIGHLY VARIABLE ACROSS BRAINS
◆ “Talairach and Tournoux’s (1988, 1993) atlas is of limited value as well.”(Geyer et al, 2011)
Select ROI: Brodmann LimitationsSelection & Cell Columns
Limitations of Spatial Resolution
◆ ROIs are dependent on the generation of current from large Dipole Layers for detection.
◆ ROIs may contain millions of cells and network components. Which one and how do we select it and train it with Limited Resolution?
“in vitro transmitter receptor
autoradiography (Zilles et al., 2002) have
revealed functionally relevant subregions
within many of the areas considered by
Brodmann to be homogeneous.”
Anatomical Accuracy &
10-20 System
Homan et al (1987) reports,
based on cadaver studies, that there is a 10% variance
between the 10-20 system and
actual gross anatomical
locations (Homan et al, 1987).
Which Symptoms Relate To DMN: Laird
#18 ?
◆ Self-reflective thoughts
and judgments that
depend on inferred
social and emotional
content. dMPFC
◆ Episodic memory HF
◆ Fantasy, daydreams,
envisioning the future,
past ruminations, moral
judgments, inferring
thoughts of others PCC.
Selectivity Challenge
Network Sequencing
Effective Connectivity PatternsNetworks have dynamic, not
static, patterns of information
flow.
Some network nodes may
need to be more active while
others need to be less active.
These patterns of activity may
need to unfold in a specific
temporal sequence
Can we identify the correct or
most efficient pattern of
activity?
Which Frequencies at what
time sequence is optimal for
training?
Upstream vs Downstream
IssuesWhere Do We Start Training?
Spectral Compensation During
Training
The Brain Changes In
Iterative Sequences of Error
Correction“Implicit to the commonly held notion
of plasticity is the concept that there is a definable
starting point after which one may be able
to record and measure change. In fact, there
is no such beginning point because any event
falls upon a moving target, i.e., a brain undergoing
constant change triggered by previous
events or resulting from intrinsic remodeling
activity.” (Pascual-Leone)
Theta Amplitude
Compensates for Alpha &
Beta Asymmetry Training.
Training Symmetry vs Z
Score
Amplitude AsymmetryZ Score
Implications For Training
Approaches
Using Z Score To Train
Networks?
Global Constraint Theory:
Z = 0◆ More target variables contrained result in a better
outcome?
◆ Reduce all outliers equally so they shift from the tails of the distribution to within 2 standard deviations or less.
◆ Key Strategy: minimize efforts to train targets with values closer to the norm.
◆ Evidence suggests that more normed targets move outside the arbitrary thresholds and require that movement for maximum plasticity required to reorganize.
◆ Each individual may require a different definition of threshold boundaries for maximal function.
Averaged vs Dynamic Movement
Actual EEG
Variance is
greater than
mean values
and often
outside normal
range.
Alternative Z score
Strategy
◆ Train midrange targets and capture extreme
outliers as they progressively move into range.
◆ Z momentum, ZMO or Z mean
◆ Train extreme outliers first and then progressively
incorporate midrange targets.
◆ Use Min and max or Z bars to evaluate targets.
◆ Should a hub be weighted differently than a node
with respect to deviance?
The sLORETA Z Score
Method
Select all locations and values in all neurometric
dimensions and constrain to within 2 SD using upper
and lower thresholds.
◆ Select key ROIs and train all values
◆ Select key ROIs and train some values.
◆ Select values based on approximation to mean.
◆ When average of all means remain within 2sd
provide reinforcement.
Strengths & Weaknesses
◆ Captures all measured deviations.
◆ May have more specific local effect.
◆ Some values may need room to deviate outside
the norm.
◆ Deviance may be dynamic requiring moving
targets rather than a static issue.
◆ Training to a Static Norm while on task.
A Model for QEEG and sLORETA
Correlates to Predicting and
Enhancing Human
Performance: A Multivariate
ApproachDAVID S. CANTOR, PH.D., DICK GENARDI, PH.D.,
BARBARA MINTON, PH.D., ROBERT CHABOT, PH.D.
QEEG and LORETA Correlates
to Neuropsychological
Performance
Subjects:
• 128 cases were drawn from a clinical population from the private practice of first author
• Age range = 5 – 77 (M = 18.45, SD=16.5)
• Scores from approximately 27 test instruments have been implemented into this database with varying N subjects in each of the cells
Measures: EEG
• Data collected either on Cadwell Easy II or Deymed TruScan-32– Minimum of 15 minutes of eye closed data
collected- 0-70 Hz; 256 time points/sec; 10-20 System minus Fpz, Oz
– Minimum of 60 seconds of artifact free data processed off-line
– All data translated for use with NYU database
– A total of 1614 bipolar and monopolar, and various multivariate measures were derived with age regressed Z-score transformations
Methods:
• Step-wise regression analyses for 1300 of the qEEg variable set were used to establish which variables were highest in predicting various performance scores
• Regression fitting was used to establish equations including intercept values that can be used to predict such scores.
• Group average QEEG maps were derived to characterize some of general QEEG profile correlated for these measures
• Source localization methods are being used to identify key neuro-anatomic structures associated with deviations in the QEEG spectrum that are then correlated with performance
Results -1:
Results – 2:
Case Example – FSIQ = 49
Case Sample – FSIQ = 49
Results -3: Sample Equations
• FSIQ = 99.28 – 6.25(MCoLatD) – 3(MReO2L) + 5.46
(MAsF7F8C) + 7.0(MCoT5T6D) – 6.40 (MCo0102B) –
3.52 (BAsHeadD) – 4.22 (MCoMedT)
• VisMem = 88.1 + 7.87(MMFC3S) – 5.95(MCoPostT) –
11.0 (MAsF7F8S) + 6.35(MAsPostA) – 7.48(BAsFTA) -
5.50(MCoT5T6C) – 5.21(BReC4CzA) +
5.94(MIcFp2F4D) – 4.86(MMFRPosB)
Further Analyses (1)
◆ FSIQ = 99.28 – 6.25(MCoLatD) – 3(MReO2L) + 5.46
(MAsF7F8C) + 7.0(MCoT5T6D) – 6.40 (MCo0102B) –
3.52 (BAsHeadD) – 4.22 (MCoMedT)
◆ Case 1 – Mitochondrial Disorder
◆ Clinical Findings = IQ = 10-20
◆ Equation = 14.98
◆ Case – Spinal Muscular Dystrophy
◆ EQ: IQ – 81
RESTING STATE CORTICAL
RHYTHMS IN ATHLETES: A HIGH-
RESOLUTION EEG STUDYCLAUDIO DEL PERCIO, PH.D.,
Abstract◆ The present electroencephalographic (EEG) study tested the
working hypothesis that the amplitude of resting state cortical EEG rhythms (especially alpha, 8-12 Hz) was higher in elite athletes compared with amateur athletes and non-athletes, as a reflection of the efficiency of underlying back-ground neural synchronization mechanisms. Eyes-closed resting state EEG data were recorded in 16 elite karate athletes, 20 amateur karate athletes, and 25 non-athletes. The EEG rhythms of interest were delta (2-4 Hz), theta (4-8 Hz), alpha 1 (8-10.5 Hz), alpha 2 (10.513 Hz), beta 1 (13-20 Hz), and beta 2 (20-30 Hz). EEG cortical sources were estimated by low resolution brain electromagnetic tomography (LORETA). Statistical results showed that the amplitude of parietal and occipital alpha 1 sources was significantly higher in the elite karate athletes than in the non-athletes and karate amateur athletes. Similar results were observed in parietal and occipital delta sources as well as in occipital theta sources. Finally, a control confirmatory experiment showed that the amplitude of parietal and occipital delta and alpha 1 sources was stronger in 8 elite rhythmic gymnasts compared with 14 non-athletes. These results support the hypothesis that cortical neural synchronization at the basis of eyes-closed resting state EEG rhythms is enhanced in elite athletes than in control subjects.
Conclusions
◆ Neurometric Multivariate Approaches are to date
the best way to define Normal Profiles from
Disorders suggesting that there is range of
homeostatic mechanism that likely “ready” an
individual to be adaptive
◆ Similar Multivariate Approaches can be used to
specify Psychological Constructs and
Performance Capabilities
◆ The future of Neurofeedback Protocols will
depend on the refinement and ability to use multivariate equations that are the training
protocols
References24
1
◆ Bassette, D.S., Dyer-Lindenberg, A., Archard, S., Duke, T., Bullmore, E. (2006). Adaptive reconfiguration of fractal small-world human brain functional networks. PNAS, 103(51): 19518-19523.
◆ Bressler, S., Menon, V. (2010). Large-scale brain networks in cognition: emerging methods and principles. Trends in Cognitive Sciences, 14, 277–290
◆ Buckner, R. L., Andrews-Hanna, J. R., Schacter, D. L. The brain’s default network: Anatomy, function, and relevance to disease. (2008). Annals of the New York Academy of Sciences,1124, 1-38.
◆ Chen, L., Ros, T., Gruzelier, J. (2012). Dynamic Changes of ICA-Derived EEG Functional Connectivity in the Resting State. Human Brain Mapping 00:000–000, 1-17.
◆ Chu, C.J., Kramer, M.A., Pathmanathan, J., Bianchi, M.T.,Westover, M.B., Wizon,L., Cash, S.S. Emergence of stable functional networks in long-term human electroencephalography. The Journal of Neuroscience, 32(8), 2703–2713.
◆ Gu, S., Pasqualetti, F., Cieslak, M., Telesford, G., YU, A,. Kahn, A., Medaglia, J., Vettel, J., Miller, M., Grafton, S., Bassett, D. (2015). Nature Communications, 6(414), 1-10.
◆ Hagmann, P., Cammoun, L., Gigandet, X., Meuli, R., Honey, C. J., Wedeen, V. J., Sporns, O. (2008). Mapping the structural core of human cerebral cortex. PLoS Biology, 6(7), 1479-1493.
◆ Honey C.J., Kotter R., Breakspear M., Sporns O. (2007). Network structure of cerebral cortex shapes functional connectivity on multiple time scales. Proc Natl Acad Sci USA, 104, 10240–10245.
◆ Honey, C., Sporns, O., Cammoun, L., Gigandet, X., Thiran, J., Meuli, R., Hagmannn, P. (2009). Predicting human resting-state functional connectivity from structural connectivity. PNAS, 106(6), 2035-2040.
◆ Laird, A.R., Fox, P.M., Eickhoff , S.B., Turner, J.A., Ray, K.L., McKay, D.R., Glahn, D.C., Beckmann, C.F., Smith, S.M., Fox, P.T. (2012). Behavioral interpretations of intrinsic connectivity networks. Journal of Cognitive Neuroscience, 1–16. In Press.
◆ Leech, R., Sharp, D. (2014). The role of the posterior cingulate cortex in cognition and disease. Brain, 137(1), 12–32. (listed as 2013)
◆ McIntosh, A. R., Korostil, M. (2008). Interpretation of neuroimaging data based on network concepts. Brain Imaging and Behavior 2, 264-269.
◆ Meehan, Timothy P, & Bressler, Steven L. (2012). Neurocognitive networks: Findings, models, and theory. Neurosci Biobehav Rev., 36(10), 2232-47
◆ Menon V. (2015) Large-Scale Functional Brain Organization. In: Arthur W. Toga, editor. Brain Mapping: An Encyclopedic Reference, 2 , 449-459. Academic Press: Elsevier.
◆ Menon V. (2015) Salience Network. In: Arthur W. Toga, editor. Brain Mapping: An Encyclopedic Reference, 2, 597-611. Academic Press: Elsevier.
◆ Miller, P. (2016). Dynamical systems, attractors, and neural circuits. F1000Research, 5, 992, 1-18.
◆ Niedermeyer, E., Lopes da Silva, F.H. (2005). Electroencephalography: Basic principles, clinical applications, and related fields. Lippincott Williams & Wilkins: New York.
◆ Nunez, P. L., Srinivasan, R.(2006). Electric fields of the brain: The neurophysics of EEG. (2ng Ed.) New York: Oxford University Press.
◆ Rabinovich, M.I., Afraimovich, V.S., Blick, C., Varona, P. Information flow dynamics in the brain. (2011). Physics of Life Reviews, 9: 51–73.
◆ Scolari, M., Seidl-Rathkopf, K., Kastner, S. (2015). Functions of the human frontoparietalattention network: Evidence from neuroimaging. Current Opinion in Behavioral Sciences, 1:32–39.
◆ Van den Heuvel, M., Kahn, R., Goñi, J., Sporns, O. (2012). High-cost, high-capacity backbone for global brain communication. PNAS, 109(28), 1372–11377.
◆ Van den Heuval, M., Sporns, O. (2013). Network hubs in the human brain. Trends in Cognitive Sciences, 17(12), 683-689.
◆ Zaleski, A., Fornito, A., Bullmore, E. (2010). Network-based statistic: identifying differences in brain networks. Neuroimage, 53(4):1197-207
top related