fMRI: Biological Basis and Experiment Design Lecture 17: Baselines, SNR & CNR Visualizing data Trial triggered average Baselines SNR and CNR –Block vs. ER 1 light year = 5,913,000,000,000 miles?
Localizers scans define ROI
Sample data – selected voxels, impulse response Subject 3 Subject 4 Subject 2 Subject 1 voxel time (seconds... scans concatenated) Scan 1Scan 2Scan 3Scan 4
Sample data – ROI average Subject 3 Subject 4 Subject 2 Subject 1 voxel time (seconds... scans concatenated) Scan 1Scan 2Scan 3Scan 4
Sample data – response average (60 trials) Subject 3 Subject 4 Subject 2 Subject 1 time (seconds)
Sample data – response average (30 trials) Subject 3 Subject 4 Subject 2 Subject 1 time (seconds)
Sample data – response average (15 trials) Subject 3 Subject 4 Subject 2 Subject 1 time (seconds)
Effect of averaging Subject 3 Subject 4 Subject 2 Subject 1 Error bars are S.E.M. ( M )
Analyzing each scan separately Subject 3 Subject 4 Subject 2 Subject 1
Baselines and contrasts 20s block cycles10s block cycles
Trade-off between CNR and block length 20s block cycles, 40 repetitions 10s block cycles, 80 repetitions More contrast, fewer samples Less contrast, more samples
Signal-to-noise ratio Thermal noise –white –independent of voxel size –independent of tissue type –independent of voxel location –trivial with 3 x 3 x 3mm voxels –a problem with 1.5 x 1.5 x 1.5 mm voxels (1/8 th volume) –for the sake of argument: SNR thermal,3mm = 160 SNR thermal,1.5mm = 20
SNR thermal : block and ER design
Contrast-to-noise ratio: block and (slow) ER design
Response estimate: single voxel (SNR = 120)
Sample data – block experiment Scan 2 Scan 1 time (seconds... 2 scans concatenated) voxel ROI averagepredicted from HIRF
Block response estimated with different HIRFs HIRFHRF