fMRI data analysis at CCBI Vladimir Cherkassky
Dec 18, 2015
fMRI data analysis at CCBI
Vladimir Cherkassky
Detection of Activated Voxels
• Always performed on a voxel-to-voxel basis
• T-test comparing mean values during two experimental conditions (FIASCO)
• Covariance of the time course with the paradigm function (SPM, Voxcor.id, etc.)
• Threshold estimation and selection
FIASCO data preprocessing
• Baseline correction
• de-ghosting
• mean correction
• motion correction
• outlier detection and removal
• Trend correction
• T-map computation
Processing steps (single subject)
• Fiasco processing (active voxel detection)
• Co-registration of functional and anatomical data.
• Subject-specific ROI definition
• Volume and distribution of activation
• Functional connectivity
• Morphing into standard space
Activated voxels detected at t=6
Probability map thresholding
• Correction for multiple comparisons
• Scanner-specific data properties
• Spatial correlation among voxels
• Additional considerations
• Comparing groups:– “High activators” and “low activators”– Normalization (selecting most activated voxels)– FDR (false discovery rate) method
Co-registration of functional and anatomical data
Brain areas
Regions Of Inerest (ROI)
Activation at t=6 and ROIs defined on the basis of subject-specific anatomical landmarks
Number of activated voxels per ROI
Processing steps (group analysis)
• ANOVA analysis of volume of activation
• ANOVA analysis of location of activation (centroids in standard space)
• ANOVA analysis of functional connectivity
• Standard space averaging (“hit” maps)
• Additional analyses:– Factor analysis, MDS, ...
Average volume of activation per ROI for the two groups of subjects:
autistic participants and matched control participants.
Sentence comprehension task.
Highlighted areas show statistically significant group difference.
Normalization thresholds used for between-group analysis.
Difference in the distribution of activation in the main language areas for the two groups
Averaged group activation
Overall pattern of activation:similar pattern for three groups of subjects
Functional connectivity
• FC is measured as a correlation between averaged time courses (tc) of activated voxels for the two brain areas (ROIs)
• Synchronization can be induced by the connection (direct or indirect) between areas or some common input.
• We interpret FC as a measure of interaction between brain areas working together on the same task.
Time course of activated voxels
• Useful signal is ~2%
• As a result, single voxel tc is extremely noisy
• Averaged tc of all activated voxels within an ROI (min number of voxels we use is 3)
• Most of the ROIs consist of more than one slice
• Correction for the slice acquisition sequence is necessary for proper connectivity estimation
• Images to include in time course calculation
Averaged tc for the two areas involved in thelanguage task.
Note the scale (% signal change from fixation) andthe images included in theanalysis.
LB
-3
-2
-1
0
1
2
3
4
5
6
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59
fourier
linear
last
LDLPFC
-3
-2
-1
0
1
2
3
4
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59
fourier
linear
last
LT
-3
-2
-1
0
1
2
3
4
5
6
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59
fourier
linear
last
Averaged tc with differentmethods of interpolation(correction for slice acquisition time)
Though the differences lookminor, linear interpolationintroduces considerably largerlevel of correlation betweenthe time courses.
SYNAM2 cond 5 LT (normalized to15 voxels) 10 items averaged across subjects
-2
-1
0
1
2
3
0 15 30 45 60 75 90 105 120 135 150
2945 14 2953 15
2964 14 2999 15
3002 16 3003 13
avg last
Between subjects variability of the averaged tc (event-related
study)
Functional connectivity within thelanguage network for autistic andcontrol groups.
Note the systematically higherconnectivity level for the controlgroup.
Pairs of areas with statisticallysignificant differences in functional connectivity for thetwo groups of participants
Overall connectivity pattern:
1. Note the high levelof similarity for the twogroups (r=0.69 for allmeasured connections;r=0.95 for the connectionswith significant differences)
2. Note the systematicallyhigher connectivity for thecontrol group
Conclusions
• Our approach provides accurate measures of activation volume and location, as well as functional connectivity between the brain areas.
• These measures can be used for testing the effects of experimental manipulations for single subject, group, and between-group analyses.