1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

23
1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett

Transcript of 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

Page 1: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

1st Level AnalysisContrasts and Inferences

Nico Bunzeck

Katya Woollett

Page 2: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

What do we use fMRI for?

Functional specialisation:– Identification of regionally specific effects

that can be attributed to changing stimuli or task conditions

Functional integration:– Identification of interactions among

specialised cortical areas and how these interactions depend upon context

Page 3: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

Planning the experiment

we should have a clear…

• Hypothesis / Question • Design that help me to answer my question• How I am going to build my SPM model• How I am going to analyse my data• What contrasts and inferences are made is

dependent on choice of experimental design

Page 4: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

Design overview

Cognitive subtraction

Conjunctions

Factorial

Page 5: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

Definition: the difference between two tasks can be formulated as a separable cognitive or sensorimotor component

Then, regionally specific differences in haemodynamic response, evoked by the two tasks, identify the corresponding functionally specialised area

Many subtraction designs rely on the assumption of pure insertion

Cognitive subtraction

Page 6: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

Pure insertion

Pure insertion: A new cognitive (A) component can be purely inserted without affecting the expression of the previous ones (e.g., B).

If this were not the case the difference between tasks that did, and did not, include component B would depend on the presence of component A.

Pure insertion discounts both functional and psychological interactions and therefore represents a very restricted precondition for cognitive subtraction

Page 7: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

1. The experimental task and baseline/ control task must be identical in every way except for the process of interest

Assumptions of cognitive subtraction

Activation task involving

process of interest

Baseline/control task

identical to A except for process

of interest

Region(s) involved in the

cognitive/ sensorimotor

process of interest

A B

Page 8: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

For example…

Violates assumption 1: task A and B identical but the process of interestMany processes in addition to presence of biological motion in A including visual motion and eye movements

Question: brain regions are responsible for biological motion?

Activation task

Point light display movie

Baseline / control task

Point light display static image

Regions involved in biological motion?

Task A Task B

MT / V5STS Frontal eye fields Cerebellum Parietal cortex

Page 9: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

For example…

A better baseline to answer this question.

Question: areas for biological motion?

Activation taskPoint light

display moves

Baseline / control task

Random dots moves

Regions involved in biological motion?

Task A Task B

MT / V5STS Frontal eye fields Cerebellum Parietal cortex

Page 10: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

2. There must be no implicit processing of the component of interest in the baseline/control task

Assumptions of cognitive subtraction

Activation task involving

process of interest

Baseline/control task

identical to A except for process of

interest

Region(s) involved in the

cognitive/ sensorimotor

process of interest

AB

Page 11: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

For example…

Activation task

read words aloud

Baseline / control task

look at words

Regions involved in semantic

processing?

Task A Task B

Violates assumption 2: Note implicit processing of the component of interest in the baseline task

Question: is inferotemporal cortex involvedin the semantic processing?

Page 12: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

1. The experimental task and baseline/ control task must be identical in every way except for the process of interest

2. There must be no implicit processing of the component of interest in the baseline task

Summary

Page 13: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

• Cognitive conjunctions combine a series of subtractions with the aim of isolating a process that is common to two (or more) task pairs

• The assumption of pure insertion can be avoided by extracting the presence of a main effect in the absence of an interaction

• Conjunctions have the advantage of testing the effect independently of the task context, thereby controlling for influences of the effect on the context.

Conjunctions

Page 14: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

For example…

-A1

Activation taskRead words

B1

Baseline task

Say “Yes”

A2

Activation task

Name picture

B2

Baseline task

Say “Yes”

-

Badge

Phonological

retrieval

Form processing

Lexical orthography

Sublexical orthography

Form and colour processing

Object structure

Page 15: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

• The only restriction is that differences between the task pairs both contain the component of interest

• The analysis results in any commonality in activation differences between the task pairs

• The resulting region should be uniquely associated with the process of interest, not any interactions specific to each subtraction

Using Conjunctions- Summary

Page 16: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

Factorial Design

In factorial designs there are two or more factors

The main effects of each factor identify brain areas that respond to that particular factor of interest

The interaction between factors identifies brain areas where the effect of one factor varies depending on the presence or absence of the other factor

This allows to measure the effect of one factor on the expression of the other factor

Page 17: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

Factorial design 2x2F

acto

r B

Factor A

B1

A1

B2

A2

1 2

3 4

Sugiura et al, 2005

Page 18: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

Factorial design 2x2

MAIN EFFECTS

- Difference between columns

- Difference between rows

INTERACTIONS

- Difference between columns depending on status of rows (or vice versa)

Page 19: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

Main effects

Main effect of factor A1

(1+3)-(2+4)

Fac

tor

B

Factor A

B1

A1

B2

A2

1 2

3 4

BOLD signal in voxel Y

Factor A1

Factor A2

B1 B2

Page 20: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

Main effects

Main effect of factor B1

(1+2)-(3+4)

BOLD signal in voxel Y

Factor A1

Factor A2

B1 B2

Fac

tor

B

Factor A

B1

A1

B2

A2

1 2

3 4

Page 21: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

Interactions …

Interaction between the factors (1-2)-(3-4)

BOLD signal in voxel Y

B1 B2

Factor A1

Factor A2

Fac

tor

B

Factor A

B1

A1

B2

A2

1 2

3 4

Page 22: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

Crossover interaction

Interaction between the factors A1 B1 and A2 B2:

(1-2)-(3-4)

Factor A1

Factor A2

B1 B2

Fac

tor

B

Factor A

B1

A1

B2

A2

1 2

3 4

BOLD signal in voxel Y

Page 23: 1st Level Analysis Contrasts and Inferences Nico Bunzeck Katya Woollett.

How do I do it in SPM

Now we have clear what comparisons we want to make… to answer our question