Modeling and Nonlinear Analysis in fMRI via Statistical Learning - Advance Image Processing in Magnetic Resonance Imaging - page 566

a single-shot, gradient-echo, echo-planar sequence. The sequence parameters
were TR = 1500 msec, TE = 50 msec, flip angle = 80 ° , NEX = 1, providing

an in-plane voxel size of 3.125 3.125 mm 2 . Nine axial slices were collected

and were 4 mm thick (skip 0 mm). Corresponding T1-weighted structural
images of the same thickness were collected in the same session in the same
slice locations (TR

×

192 matrix, 2NEX).
Two experimental conditions were a finger-tapping task with left and right
hands, respectively. During each run, the control (motionless and relaxed) and
experimental conditions were alternated (11 scans active, 6 rest) and repeated
so that each experimental condition would occur 4 times for a total of 136
volumes (3.125

= 500, TE = 11, FOV = 200 mm, 256 ×

4 mm 3 ). Four runs of data were collected in this
experiment. For this data set, in order to examine the activation detection in
the supplementary motor area (SMA), we focused on the task of right hand vs.
the rest condition, with results shown in Figure 18.9. The window size used is
3

× 3.125 ×

× 3 × 3 × 136, covering a brain region whose physical size is almost isotropic

12 mm 3 ). Although activations in the primary motor area are
consistently detected by both the ST-SVR method and the traditional t -test (on
presmoothed data with empirically chosen FWHM

(9.4 × 9.4 ×

×
4.71 mm), ST-SVR on two-run data leads to activations in the SMA (Figure 18.9a),
which are not shown by t -test on the same two-run data (Figure 18.9b). For the
t -test on the two-run data in Figure 18.9b, if we intentionally further decrease
the t threshold and try to detect the activations in SMA, spurious activations
over regions other than the primary motor area and SMA, as well as false
activations outside of the cerebellum, would appear. However, if the t -test is per-
formed on four runs of data, the activations in SMA are able to show up as in

= 3.68 mm × 3.68 mm

ST-SVR (2 Run) t -test (2 Run) t -test (4 Run)

SMA

5.3 < t < 10.24
p < 10 −

3.0 < t < 7.17
p < 0.002

3.6 < t < 9.45
p < 0.001

6

(a) (b) (c)

FIGURE 18.9 (See color insert following page 306 ). Multirun result comparison for a motor

experiment of real fMRI data. Top: coronal slices; bottom: axial slices (color activation maps).