Dynamic Cerebral Contrast-enhanced Perfusion MRI - Biomedical Signal Analysis: Contemporary Methods and Applications - page 301

ing a 2-D gradient echo echoplanar imaging (EPI) sequence employ-
ing 10 transversal slices with a matrix size of 128

× 128 pixels, pixel

size 1 . 88 1 . 88 mm, and a slice thickness of 3 . 0mm (TR = 1 . 5 sec, TE

=0 . 54 sec, FA = 90 ◦ ). The dynamic study consisted of 38 scans with
an interval of 1 . 5 sec, between each scan. The perfusion sequence and an
antecubital vein bolus injection (injection flow 3 ml/sec) of gadopente-
tate dimeglumine (0 . 15 mmol/kg body weight, Magnevist TM ,Schering,
Berlin, Germany) were started simultaneously in order to obtain several
(more than six) scans before cerebral first pass of the contrast-agent.
The registration of the images was performed based on the automatic
image alignment (AIR) algorithm [288].

×

Data analysis

In an initial step, a radiologist excluded by manual contour tracing the
extracerebral parts of the given data sets. Manual presegmentation was
used for simplicity, as this study was designed to examine only a few
MRI data sets in order to demonstrate the applicability of the perfusion
analysis method.
For each voxel, the raw gray-level time series S ( τ ) ,τ

}
was transformed into a pixel time course (PTC) of relative signal reduc-
tion x ( τ )by

∈{

1 ,..., 38

x ( τ )= S ( τ )
S 0

α

, (11.1)

where S 0 is the precontrast gray level and α> 0 a is distortion ex-
ponent. The effect of the native signal intensity was eliminated prior
to contrast-agent application. If time-concentration curves are not com-
puted according to the above equation (i.e., avoiding division of the raw
time series data by the pre-contrast gray level before clustering), implicit
use is made of additional tissue-specific MR imaging properties that do
not directly relate to perfusion characteristics alone.
In the study, S 0 was computed as the average gray level at scan times

τ , excluding the first two scans. There exists an exponential

relationship between the relative signal reduction x ( τ )andthelocal
contrast-agent tissue concentration c ( τ ) [223], [181], [83], [137]:

∈{ 3 , 4 , 5 }

α ln S ( τ )
S 0

,

c ( τ )= − ln x ( τ )= − (11.2)