Outline

1
Outline
Using Large-Scale Neural Models to Help Determine
the Neural Substrates of Functional and Effective
ConnectivityBarry HorwitzBrain Imaging
Modeling SectionNational Institute on Deafness
Other Communication DisordersNational
Institutes of Health

• Some general remarks on modeling and functional
  neuroimaging
• Integrating neuroscientic data across multiple
  spatiotemporal scales
• Using biologically realistic neural models to
  understand the neural basis of functional and
  effective connectivity

2
Modeling and Neuroimaging

• Simulation models
• Permits one to make explicit assumptions about
  the neural basis of a task.
• Data fitting models
• The fitted parameters are not necessarily
  interpretable in mechanistic neural terms.

Models that cross spatiotemporal scales
3
Delayed Matched-to-Sample Tasks
Shape
Tonal pattern
4
Large-Scale, Neurobiologically Realistic Auditory
Mode DMS Task
(Husain et al., Neuroimage, 21 1701-1720, 2004)
PFC
Ai
FS
Aii
D2
D1
R
STG/STS
MGN
Model vs. Experiment Signal Change (Tonal
Contours-Tones)
5
Auditory Continuity Illusion Examples
(1) Phonemic Restoration We are able to hear
conversation obscured by extraneous noise such as
some one coughing.
(2) Fragmented sweeps
6
Normal
Stimulus 1
Delay
Stimulus 2
Response
Next Trial
100 200 100
400 Timesteps
Husain et al., JOCN, In Press
7
Predictions of Model fMRI/PET Results
Husain et al., JOCN, In Press
8
Sources of Variability
Functional Connectivity Measure of Covariance
Parameters creating variability
Subject-to-Subject
Trial-to-Trial
Block-to-Block
Each region is represented by a 50/50 ratio of
specific neurons and non-specific neurons with
randomized cross-connections on a trial-by-trial
basis. Specific Task dedicated, perceiving
actual inputs. Non-Specific Represents non-task
related neurons noise inputs.
Alter value of the level of attention accounts
for issues such as habituation, learning,
enthusiasm, boredom with task, etc Test a range
of attention levels per subject.
Interregional anatomical connectivity is altered
for each new subject. Subj1 V1h?V4c 92, D1?D2
85, FS?FR 96, Subj2 V1h?V4c 83, D1?D2
98, FS?FR 94, 13 Connections in all are
altered/subject.
Task and Control Task
Performed task and control trials to test task
dependence. Task Structured inputs, must be
remembered during delay. Control Scrambled
inputs, short-term memory processing unnecessary.
9
Horizontal selective units
Excitatory E?E Inhibitory E?I
Horizontal selective units
FS
Corner selective units
IT
LGN
D1
FR
Vertical selective units
Vertical selective units
D2
PFC
Attention
V1/V2
V4
10
Functional Connectivity Between IT and Other
Brain Areas
11
Functional Connectivity vs. Anatomic Connectivity
for Auditory Model
12
Putting It All Together Neural Modeling
13
Collaborators
Malle Tagamets Fatima Husain Brent Warner Julie
Fitzer Theresa Long