Analysis

Question 1

Why is preprocessing needed?

Answer 1

• Head moves during scan and this must be accounted for by motion correction
• When analysing group data it is of advantage to spatially smooth the data (overlap between participants)
• Different brains are different sizes and shapes and this must be somehow accounted for with spatial normalisation
• The data are not free of noise and this can be treated with temporal filtering

Question 2

Temporal filtering – linear drift

Answer 2

• Sometimes the signal “drifts” over a long period and this drift can often be treated as noise and removed – also note that if all of one condition was at the beginning of scan run and all of other condition was at the end then it would be harder to model this drift as noise

Question 3

Physiological Noise

Answer 3

• Respiration
• Cardiac cycle
• Solutions: gating, avoiding paradigms at those frequencies

Question 4

Canonical Hemodynamic Response Function (HRF)

Answer 4

models the transfer of brain activity into BOLD signal
- Characterised by delaying peak activity and “smearing” in time
describes how the BOLD signal evolves over time in response to changes in brain activity

Question 5

Blocked Design

Answer 5

fixed sequential order of presentations, extended time intervals
–> the spacing in time of stimuli is an essential aspect of fMRI design and must take into account the HRF

Question 6

Event related design

Answer 6

randomized order of presentation, relatively short time intervals, lower detection power

• Slow ER design
• rapid interbalanced ER design
• rapid Jittered ER design
• Mixed design

Question 7

General Linear Model

Answer 7

• The GLM is an overarching tool that can do anything that the simpler tests do
• You can examine any combination of contrasts with one GLM rather than multiple correlations
• Allows much greater flexibility for combining data within subjects and between subjects
• Makes it much easier to counterbalance orders and discard bad sections of data
• Allows you to model things that may account for variability in the data even though they aren’t interesting in and of themselves

Question 8

Controlling for the Type 1 error

Answer 8

• False discovery rate (FDR) – controls proportion of Type 1 errors in relation to true positives
• Family Wise Error Rate (FWER) – controls probability of making a Type 1 Error at all
• NO Bonferroni (assumes that voxels are independent of another)