Image Processing Reference
In-Depth Information
640
620
600
1
21
41
61
81 101 121 141 161 181 201 221 241 261 281 301 321
640
620
600
1
21
41
61
81 101 121 141 161 181 201 221 241 261 281 301 321
640
620
600
1
21
41
61
81 101 121 141 161 181 201 221 241 261 281 301 321
640
620
600
1
21
41
61
81 101 121 141 161 181 201 221 241 261 281 301 321
FIGURE 15.4
Temporal filtering. (a) Unfiltered time course from a region in the visual
cortex during an event-related fMRI study. (b), (c), and (d) show filtered versions of the
same time course obtained after (b) high-pass filtering (HPF, filter cutoff = 5 cycles/run),
(c) HPF and moderate low-pass filtering (LPF1, filter cutoff = 75 cycles/run), and (d)
HPF and heavy low-pass filtering (LPF2, filter cutoff = 25 cycles/run).
15.3
STATISTICAL LOCALIZATION
OF BRAIN ACTIVATION
The most important step in the analysis of the functional time series is the
detection of task-related BOLD activation and the creation of activation maps.
In the first fMRI studies, a simple method based on image subtraction was
used to create descriptive images of the task-dependent brain areas [1-3]. Accord-
ing to the “pure insertion” hypothesis, voxels with a high gray level in the “dif-
ference” image, formed by subtracting the “control” from the “task” condition
images, reflect the areas with a task-induced differential activation. With this
method, the value of the intensity threshold between activated and nonactivated
voxels is arbitrarily chosen. Furthermore, image subtraction is very sensitive to
movement-related effects and to other unexpected signal changes. More reliable
Search WWH ::
Custom Search