Linking Cognitive Function and Brain Through Neural Context

1
Linking Cognitive Function and Brain Through
Neural Context

• Anthony Randal McIntosh

Department of Psychology University of Toronto
2
Overview

• Localized vs. Distributed Function
• Neural plasticity
• Dynamic processes
• Neural Context
• Theory examples

3
Kleist, 1934
4
Why localize function?

• Enables more precision in clinical evaluation
• Clues as to functional organization
• What are the critical sites?
• Constraints for cognitive/behavioural theory

5
Localization of Cortical Function

• Primarily confirmed through focal lesion and/or
  stimulation
• Additional confirmation through physiological
  measures
• Not always a nice match!
• A region can participate in a function, but not
  be critical for its expression.

6
Localization of Cortical Function

• Problems
• Plasticity The brain changes in response to any
  perturbation (short- or long-term)
• Deficits after damage may be related to tissue
  loss per se, or the response of intact tissue
• Cannot inform as to mechanisms (not explanatory)

7
Cortical plasticity

• Best examples of cortical plasticity come from
  the study of the damaged brain.
• Degeneration tracing in neuroanatomy
• Diaschisis (Von Monakov)
• Spreading cortical depression, distal effects
• Peripheral limb damage
• Merzenich, Ramachandran
• Central damage
• Nudo, Schallert, Whishaw
• Experience-dependent
• Recanzone, Weinberger, Gonzalez-Lima

8
Response to damagePeripheral dennervation
Hickmott Merzenich, J Neurophys, 2002
9
Response to damagePatient M.L.s Frontal Lobe
Lesion
Levine, et al, Brain, 1998
10
Altered Hippocampal Activity During Retrieval in
ML
rCBF response
ENC
ENC
ENC
RET
RET
RET
TBI
M.L.
Ctls
11
Problems

• Brain response to perturbation is rapid and
  changes with time
• Persistant deficits may come about through
  abnormal operations of intact tissue
• Damage may alter the systems supporting intact
  function

12
Historical positionsDynamic function

• We suggest the material basis of the higher
  nervous processes is the brain as a whole, but
  that the brain is a highly differentiated system
  whose parts are responsible for different aspects
  of the unifed whole.Luria (1962) - Higher
  cortical functions in man
• Bethe, Lashley, Hebb, Lorente de No, Mountcastle,
  Edelman, Mesulam, Bressler

13
Localization of Cortical Function

• Primarily confirmed through focal lesion and/or
  stimulation
• Additional confirmation through physiological
  measures
• Not always a nice match!
• A region can participate in a function, but not
  be critical for its expression.

14
Mapping of the brain to cognition
15
Cabeza Nyberg Imaging Cognition II 275 PET
and fMRI Studies (J Cognit Neuro, 2000)
16
Neural Context

• The behavioural relevance of activity changes in
  one brain area depends on the activity in other
  areas.
• The important factor is not that a particular
  event occurred at a particular site, but rather
  under what neural context did that event occur --
  in other words what was the rest of the brain
  doing? (McIntosh, 1999, Memory)

17
Neural Context

• Results from anatomy
• Anatomical determinism
• Depends on dominant afferent and efferent
  influence
• Allows information from specialized areas to be
  combined in different ways depending on the
  present demands
• Differentiate and integrate (Tononi Edelman,
  Science, 1998)
• Aggregate functions

18
Anatomical determinism
19
Aggregate Functions

• Brain areas are sparsely connected
• Enables flexibility in system response
• All parts of the brain possess the rudimentary
  characteristics necessary for cognitive function
  (e.g., response plasticity)
• Brain areas combine through their interactions
  such that their aggregate property is the
  cognitive process
• Population coding

McIntosh, 2000, Neural Networks
20
Neural Context
21
Examples of Neural Context

• Visual cortex several examples (see review
  Worgotter Eysel, 2000, TINS)
• Responses dependent on adjacent cells, background
  noise, stimulus context, salience, feedback
• Invertebrates (see review by Kristan Shaw,
  1997, Current Biol)
• Dissimilar behaviors can be mediated by the same
  networks through spatiotemporal variations in
  population activity

22
Examples of Neural ContextMultifunctional
Networks
J Neuroscience, 2002
23
Examples of Neural ContextEquivalence of
function in frontal and parietal cortices
Chafee Goldman-Rakic, 1998, 2000, J Neurophys
24
Methodological Note

• Functional connectivity (Gerstein, et al., 1978)
• temporal correlation or covariance among
  measured neural elements
• requires no assumptions about mediation of
  influences
• interregional correlations, multivariate analyses
  (Partial Least Squares)
• Effective connectivity (Aertsen, et al., 1989)
• influence (effect) that one neural element has on
  another
• requires some assumptions about mediate of
  influences
• path analysis, multiple regression

25
Partial Least Squares

• Project voxel value within task onto every other
  voxel in the image within a task yielding matrix
  X
• ZjTYj, where Zj is a vector of voxel values from
  condition j and Y is a matrix of m voxels for the
  rest of the image, matrix X is thus a jm
  rectangular matrix of within-task covariances
• Perform a singular value decomposition (SVD) on X
  to define the latent variables (LV)
• SVD(X) U,S,V where
• X USVT
• U is the j by m orthonormal matrix containing
  voxel weights (singular or eigen image)
• i.e. what is the pattern of functional
  connectivity?
• S is a diagonal matrix of j singular (eigen)
  values.
• VT is the transpose of matrix V, a j by j
  orthonormal matrix of scan weights.
• i.e., does the pattern differ across tasks?

26
Structural Equation Modeling (McIntosh
Gonzalez-Lima, 1991, McIntosh et al, 1994, Buchel
Friston, 1997)
27
Dorsal vs. Ventral Cortical Processing Streams
28
Differential Sensory ConditioningLeft Prefrontal
Interactions Awareness
McIntosh, Rajah Lobaugh, Science, 1999
29
Neural Context Same area engaged in different
functional networks

• Differential Sensory Learning with Reversal
• Visual target
• Two tones, T1 low, T2 high frequency
• Phase 1, T1 CS, T2 CS-
• Phase 2, T1 CS-, T2 CS

30
Differential Conditioning with Reversal
31
Brain-behavior Relations

• One pattern differentiated conditioned
  facilitation between groups
• Second pattern related to CS/CS- differentiaion
  in Aware group

32
Hippocampal Functional Connectivity
33
Hippocampal Functional ConnectivityAware group
network
34
Hippocampal Functional Connectivity
35
Hippocampal Functional ConnectivityUnaware group
network
36
Reversal Study Implications

• Same region may be part of different functional
  networks that support different behaviours
• Neural Context

37
Reversal Study Implications

• Same region may be part of different functional
  networks that support different behaviours
• Neural Context
• Awareness not synonymous with involvement of
  hippocampus, or any particular region, but with
  the configuration of a given functional network
• Aggregate function

38
Conclusions

• Regions may be critical for a particular
  operation, but the operation itself arises from
  combined actions of many regions
• Greater effort needed to merge distributed
  assessment of brain function with study of the
  damaged brain (e.g., E.A. Maguire et al, Brain,
  2001)

39
Effects of bilateral hippocampal damage
Maguire, et al, 2001, Brain
40
Conclusions

• Regions may be critical for a particular
  operation, but the operation itself arises from
  combined actions of many regions
• Greater effort needed to merge distributed
  assessment of brain function with study of the
  damaged brain (e.g., E.A. Maguire et al, Brain,
  2001)
• Brain regions can participate in more than one
  function
• Contributions of a region is determined through
  neural context what is the rest of the brain
  doing?

41
Take home message

• Classic approaches to studying brain organization
  are no longer adequate
• Do not take into account functional plasticity
• Neural Context Partly resulting from plasticity,
  brain regions can participate in more than one
  cognitive operation
• Detailed studies of reorganization in damaged and
  diseased brains will lead to a new appreciation
  of
• Neural dynamics (why does damage in X produce a
  deficit?)
• Neurocognitive mapping