RESTING STATE NETWORKS
         Beautiful Noise?
Overview of Talk

    Conclusion

    Methodological
    Recipe and Caveats

    Background

    Future directions
Conclusion
Resting state networks are:

   Highly robust across cultures, laboratories,
   methods

   Localized to gray matter and predicted by
   tractography

   Found in fMRI, EEG, MEG, and intracranial
   recordings

   Implicated in information processing, multi-
   tasking, memory, learning, development and
   emotion

   Easy to obtain, easy to analyze, and damn
   interesting

   Human connectomics - don’t get left behind!
A Methodological Disclaimer

WARNING:
shameless analysis
of low-frequency
fluctuations



                     Source: Smith chapter in Functional MRI: An Introduction to Methods
simple recipe for rsfMRI
5-7 minutes of rest

Fixation, or no-fixation.

Place at beginning of scan

Duplicate for maximum effect

Extra ROI timeseries/foci

  ICA vs a priori seed-r

Regression with whole brain and/or task
Life at < .1 HZ
Acquisition
99 problems but acquisition
         ain’t one
Pre-processing
Band-pass filter:
remove constant offsets
and linear trends,
retain <0.08 Hz

Regression of nuisance
variables; motion,
global signal, average
lateral ventricle, deep
cerebral white matter.
just breath
Analysis: ICA vs seed-based
Data-driven vs a priori
Other Analysis
Task-induced de-activation

Rest-Stimulus interaction

Post-scan rest

Correlation of resting connectivity with cognition, personality,
or psychiatric inventories.
Task induced De-activation
Rest-Stimulus Interaction
Post-Rest
Rest-Inventory Correlation
Why Rest?
Theory and Background
+
A little history
     The “default mode” of brain function originally proposed by
     Raichle and Fox (2001)

     Attracted initial controversy; “why we should resist the baseline”

     Has since exploded, evolved into the functional and human
     connectome projects.
  “Confusion springs from a failure to distinguish between psychological,
physiological and anatomical accounts.” The Rt. Hon. Lord Brain (Brain 1969)
            From “a brief history of the DMN” (Raichle, 2007)
Human Connectome?
Towards a Discovery Science
  Of Human Brain Function


N = 1,093

24 Centers
Resting States are Predicted by
          Structure




                           Predicting human resting-state
                           functional connectivity from structural
                           connectivity
                           1.   C. J. Honeya, O. Spornsa,1, L. Cammounb, X. Gigandetb, J. P.
 Margulies et al. (2009)        Thiranb, R. Meulicand P. Hagmannb,c
What is Anti-Correlation?
Problems with Anti-r
To be safe


If you want to look at anti-r networks:

  Record respiration

  Avoid global signal regression

  Include a task likely to engage CEN and SAL networks.
A Default Mode?

Mariana’s Trench Argument: Don’t conflate structure with
content!

Rumination, introspection, social cognition

Free-energy

Consciousness?
Picking apart DMN
Although exact function remains unclear, DMN connectivity at rest and
during task-processing now implicated in:

  Schizophrenia

  OCD

  Autism

  ADHD

  Levels of consciousness

  Task-irrelevant thoughts
Measuring the Default


The Resting State Questionnaire (RSQ)

Experience Sampling (Schooler et al)

Differential engagement by task

Correlation with individual differences
Future Directions
Real Conclusion
There is simply too much multi-modal, multi-cultural evidence to
dismiss slow-wave processing as a feature of mammalian brains

Simple stories; DMN as ‘consciousness’, ‘pure’ anti-correlation,
1:1 mapping between structure and function = FAIL

If you are spending the time and energy doing fMRI, you might
as well spend 5-7 minutes at rest.
THANK YOU



$       +

Resting-state fMRI: Beautiful Noise?

  • 1.
    RESTING STATE NETWORKS Beautiful Noise?
  • 2.
    Overview of Talk Conclusion Methodological Recipe and Caveats Background Future directions
  • 5.
    Conclusion Resting state networksare: Highly robust across cultures, laboratories, methods Localized to gray matter and predicted by tractography Found in fMRI, EEG, MEG, and intracranial recordings Implicated in information processing, multi- tasking, memory, learning, development and emotion Easy to obtain, easy to analyze, and damn interesting Human connectomics - don’t get left behind!
  • 6.
    A Methodological Disclaimer WARNING: shamelessanalysis of low-frequency fluctuations Source: Smith chapter in Functional MRI: An Introduction to Methods
  • 8.
    simple recipe forrsfMRI 5-7 minutes of rest Fixation, or no-fixation. Place at beginning of scan Duplicate for maximum effect Extra ROI timeseries/foci ICA vs a priori seed-r Regression with whole brain and/or task
  • 9.
    Life at <.1 HZ
  • 10.
  • 11.
    99 problems butacquisition ain’t one
  • 12.
    Pre-processing Band-pass filter: remove constantoffsets and linear trends, retain <0.08 Hz Regression of nuisance variables; motion, global signal, average lateral ventricle, deep cerebral white matter.
  • 13.
  • 14.
    Analysis: ICA vsseed-based Data-driven vs a priori
  • 15.
    Other Analysis Task-induced de-activation Rest-Stimulusinteraction Post-scan rest Correlation of resting connectivity with cognition, personality, or psychiatric inventories.
  • 16.
  • 17.
  • 18.
  • 19.
  • 20.
  • 21.
  • 22.
    A little history The “default mode” of brain function originally proposed by Raichle and Fox (2001) Attracted initial controversy; “why we should resist the baseline” Has since exploded, evolved into the functional and human connectome projects. “Confusion springs from a failure to distinguish between psychological, physiological and anatomical accounts.” The Rt. Hon. Lord Brain (Brain 1969) From “a brief history of the DMN” (Raichle, 2007)
  • 23.
  • 24.
    Towards a DiscoveryScience Of Human Brain Function N = 1,093 24 Centers
  • 25.
    Resting States arePredicted by Structure Predicting human resting-state functional connectivity from structural connectivity 1. C. J. Honeya, O. Spornsa,1, L. Cammounb, X. Gigandetb, J. P. Margulies et al. (2009) Thiranb, R. Meulicand P. Hagmannb,c
  • 26.
  • 27.
  • 31.
    To be safe Ifyou want to look at anti-r networks: Record respiration Avoid global signal regression Include a task likely to engage CEN and SAL networks.
  • 32.
    A Default Mode? Mariana’sTrench Argument: Don’t conflate structure with content! Rumination, introspection, social cognition Free-energy Consciousness?
  • 33.
    Picking apart DMN Althoughexact function remains unclear, DMN connectivity at rest and during task-processing now implicated in: Schizophrenia OCD Autism ADHD Levels of consciousness Task-irrelevant thoughts
  • 34.
    Measuring the Default TheResting State Questionnaire (RSQ) Experience Sampling (Schooler et al) Differential engagement by task Correlation with individual differences
  • 36.
  • 38.
    Real Conclusion There issimply too much multi-modal, multi-cultural evidence to dismiss slow-wave processing as a feature of mammalian brains Simple stories; DMN as ‘consciousness’, ‘pure’ anti-correlation, 1:1 mapping between structure and function = FAIL If you are spending the time and energy doing fMRI, you might as well spend 5-7 minutes at rest.
  • 39.