+ All Categories
Home > Documents > Complejidad Dia 8

Complejidad Dia 8

Date post: 18-Jan-2016
Category:
Upload: kirti
View: 16 times
Download: 0 times
Share this document with a friend
Description:
Complejidad Dia 8. G eo fi sic a. Biología. MacroEconomía. Psicologia. M eteorolog ía. E colog ía. UBA, Junio 26, 2012. Martes 26: 1era parte More on preprocessing of fmri images 2da Parte Redes , desde Eguiluz a Tagliazuchi . Jueves 28 - PowerPoint PPT Presentation
Popular Tags:
38
1 Complejidad Dia 8 Ecologí a Biolo gía P s i c o l o g i a Meteorolo gía MacroEconomí a Geofisic a UBA, Junio 26, 2012.
Transcript
Page 1: Complejidad Dia 8

1

Complejidad Dia 8

Ecología

Biología

Psico

log

iaMeteorología

MacroEconomía

Geofisica

UBA, Junio 26, 2012.

Page 2: Complejidad Dia 8

Martes 26:

1era parte More on preprocessing of fmri images

2da Parte Redes, desde Eguiluz a Tagliazuchi.

Jueves 28 1era parte anomalous scaling and phase transition2da parte Modeling

“Our brain is a network. A very efficient network to be precise. It is a network of a large number of different brain regions that each have their own task and function, but who are continuously sharing information with each other. As such, they form a complex integrative network in which information is continuously processed and transported between structurally and functionally linked brain regions: the brain network”

Where is the router?

Page 3: Complejidad Dia 8

conventional task-related fMRI

Resting state fMRI

From: Exploring the brain network: A review on resting-state fMRI functional connectivity. Martijn P. van den Heuvel, Hilleke E. Hulshoff Pol. European Neuropsychopharmacology, 20 (2010) 519–534

Page 4: Complejidad Dia 8

modeling the brain as a functional network with connections between regions that are functionally linked

Page 5: Complejidad Dia 8

Graph clustering-coefficient

characteristic path length

connectivity degreecentrality modularity

Graph, clustering-coefficient, characteristic path length, connectivity degree, centrality and modularity.

Page 6: Complejidad Dia 8

Network topologies: regular, random, small-world, scale-free and modular networks.

Page 7: Complejidad Dia 8
Page 8: Complejidad Dia 8

8

the “small-world” phenomenon

• Connectivity is sparse (i.e., 104 / 1011 )• Most connections are local (high clustering

coefficient)• The distance between any two network nodes is still

relatively small: how is possible?

–1011 neurons

–104 synapses per neuron

–On average two neurons are only 2 ~ 3 “synapses” apart

Page 9: Complejidad Dia 8

9

How toHow to extract extract functional brain networks? functional brain networks?

222 ,, txVtxVxV

(I)

(II)

(III)

From Eguiluz et al, Phys. Rev. Letters (2005).

fMRIfMRI

Page 10: Complejidad Dia 8

10

fMRI

My brain’s network (finger tapping)My brain’s network (finger tapping)

Undirected Degree (k)

Nodes spatial locationNodes spatial location

Colors indicate the number of links (or “degree”) of each node. yellow=1, green 2, red=3, blue=4, etc

Indicate Indicate “airports”“airports”

From Eguiluz et al, Phys. Rev. Letters (2005).

Page 11: Complejidad Dia 8

11

fMRI

Group statisticsGroup statistics

From Eguiluz et al, Phys. Rev. Letters (2005).

rc N C L <k> Crand Lrand

0.6 31503 0.14 11.4 13.41

2.0 4.3x10-

4

3.9

0.7 17174 0.13 12.9 6.29 2.1 3.7x10-

4

5.3

0.8 4891 0.15 6. 4.12 2.2 8.9x10-

4

6.0

“Small-world”

C >> Crand

L ~

Lrand Network N C L <k> . Crand Lrand

C. Elegans1

282 0.28 2.65 7.68 . 0.025 2.1

Macaque VC2

32 0.55 1.77 9.85 . 0.318 1.5

Cat Cortex2

65 0.54 1.87 17.48

. 0.273 1.4

fM RI-results

Previous related results

Brain Brain networks networks are small-are small-wordword

Page 12: Complejidad Dia 8

12

fMRI

Brain’s degree distribution (i.e., how many links each node Brain’s degree distribution (i.e., how many links each node have)have)

Scale-free

k- with ~ 2

From Eguiluz et al, Phys. Rev. Letters (2005).

Page 13: Complejidad Dia 8

13

fMRI

Average Degree DistributionAverage Degree Distribution

From Eguiluz et al, Phys. Rev. Letters (2005).

=2

Few but very well connected brain sites

n=22 from 7 subjects

Page 14: Complejidad Dia 8

14

fMRI

Average Links Length Distribution Average Links Length Distribution

From Eguiluz et al, Phys. Rev. Letters (2005).

Voxel length“~ Brain radius”

Probability of finding a link between two nodes separated by a distance x <

k() ~ 1/x2

Page 15: Complejidad Dia 8

15

fMRI

Something that bother us: Degree vs ClusteringSomething that bother us: Degree vs Clustering

From Eguiluz et al, Phys. Rev. Letters (2005).

Recall that clustering estimates the proportion of nodes forming “triangles”.

Clustering relatively independent of connectivity

Assortative

Page 16: Complejidad Dia 8

16

fMRI (Directed links)

A node tends to be either an in-hub or an out-hub

few “airports”

in-hub vs und. out-hub vs und.

From Cecchi et al, BME (2007).

Page 17: Complejidad Dia 8

17

fMRI (Directed links)

From Cecchi et al, BME (2007).

Page 18: Complejidad Dia 8

18

random

lattice

brain

(Directed links)

Assortative?

From Cecchi et al, BME (2007).

Page 19: Complejidad Dia 8

19

Finger tapping vs. Finger tapping vs. MusicMusic

From Eguiluz et al, Phys. Rev. Letters (2005).

•Different tasksDifferent tasks•Different networksDifferent networks•Similar scalingSimilar scaling

Networks are scale free across different tasksNetworks are scale free across different tasks

And during “resting state” And during “resting state” =>=>

Page 20: Complejidad Dia 8

20

Summary until now:

The large scale brain network extracted from correlations seems to be scale-free and small word

But what about dynamics?

Page 21: Complejidad Dia 8

21

Even in resting state, each positively correlated clique have a negatively correlated contrapart

Areas coloured redish have significant positive correlation with seed regions and are significantly anticorrelated with regions coloured blueish

(Fox et al , PNAS, 102, 2005)

Page 22: Complejidad Dia 8

22

Each positively correlated clique have a negatively

correlated contrapart

Healthy Controls

Chronic Pain Patients

Chialvo et al. 2007, “Beyond feeling: chronic pain hurts the brain disrupting the default-mode network dynamics”

Page 23: Complejidad Dia 8

23

Chialvo et al., “Beyond feeling: chronic pain hurts the brain disrupting the default-mode network dynamics” J.Neuroscience (2008)

~ 1

Page 24: Complejidad Dia 8

24

Snapshots Snapshots of spins of spins states in a states in a model model system system (Ising)(Ising)

What is special about being critical? What is special about being critical? Recall the Ferromagnetic-paramagnetic Phase-TransitionRecall the Ferromagnetic-paramagnetic Phase-Transition

Snapshots of spins states in the Ising model. Long range

correlations emerges at the critical point

Subcritical

SuperCritical

Critical

T<TC

T>TC

T~TC

TC

Critical Pointorder

disorder

Page 25: Complejidad Dia 8

25

Critical Ising networks ~ brain networks

Ising

Brain

EJ<i,j> Si Sj – B k Sk

Only local positive interactions Chialvo DR, Balenzuela P, Fraiman D. The brain:

What is critical about it? (arXiv.org/ cond-mat/0804.0032)

Fraiman D, Balenzuela P, Foss J. Chialvo DR, Ising like dynamics in large-scale brain networks. (arXiv.org/ cond-mat/0811.3721)

Positive correlated networks

SubCritical

Critical

SuperCritical

Page 26: Complejidad Dia 8

26

Critical Ising networks ~ brain networks

Brains

Ising

Page 27: Complejidad Dia 8

27

Critical Ising networks ~ brain networks

Negative correlations with fat tails similar to the brain data appear in the Ising data, despite the absence of negative “structural” interactions (i.e. no “inhibitory” connectivity).

Ising

Brain

Negative correlated networksSubCritical

SuperCriticalCritical

Page 28: Complejidad Dia 8

28

Assortativity

Page 29: Complejidad Dia 8

29

Critical Ising networks ~ brain networks

Page 30: Complejidad Dia 8

Resting-state networks. (functionally linked resting-state networks during rest identified using different methods (e.g. seed, ICA or clustering)

Page 31: Complejidad Dia 8

31

Easy problem # 2:Easy problem # 2:

Define a (reasonable) heuristic order Define a (reasonable) heuristic order parameter for the large scale brain parameter for the large scale brain dynamics seen in the fMRI experimentsdynamics seen in the fMRI experiments

Price: A year postdoct salary in ChicagoPrice: A year postdoct salary in Chicago(renewable)(renewable)

20072007

Laspia 2007 Laspia 2007

Page 32: Complejidad Dia 8

32

Related results

Page 33: Complejidad Dia 8

33

Page 34: Complejidad Dia 8

34

1/x2 replicated independently with fMRI

Average Links Length Distribution agrees with recent results (in Average Links Length Distribution agrees with recent results (in resting condition)resting condition)

Functional connectivity vs. Functional connectivity vs. anatomical distance.anatomical distance.

( Symmetric ( Symmetric interhemispheric)interhemispheric)

From Salvador et al, From Salvador et al, (Cerebral Cortex, 2005.)(Cerebral Cortex, 2005.)

PC()~1/x2

interhemisphericinterhemispheric

intrahemisphericintrahemispheric

Page 35: Complejidad Dia 8

C/Crandom = 2.08

L/Lrandom = 1.09

Page 36: Complejidad Dia 8

C

Lthreshold

1 2

3 4

EEG

GraphSynchronization I

Page 37: Complejidad Dia 8

37

Alzheimer patients

Path length is related to cognitive score

Control subjects

cognitive score

Clustering Path Length

Page 38: Complejidad Dia 8

38

Clauset, Newman & MooreAlgorithm*


Recommended