) and Functional Neuroimaging (CfN), University of Pennsylvania, 3400 Spruce Street,
Philadelphia, PA 19104, USA
Accepted 22 November 2005
Available online 19 January 2006
Abstract
In this study, we examine the suitability of a relatively new imaging technique, arterial spin labeled perfusion imaging, for the study of
continuous, gradual changes in neural activity. Unlike BOLD imaging, the perfusion signal is stable over long time-scales, allowing for
accurate assessment of continuous performance. In addition, perfusion fMRI provides an absolute measure of blood
Xow so signal
changes can be interpreted without reference to a baseline. The task we used was the serial response time task, a sequence learning task.
Our results show reliable correlations between performance improvements and decreases in blood
Xow in premotor cortex and the infe-
rior parietal lobe, supporting the model that learning procedures that increase e
Yciency of processing will be reXected in lower metabolic
needs in tissues that support such processes. More generally, our results show that perfusion fMRI may be applied to the study of mental
operations that produce gradual changes in neural activity.
© 2006 Elsevier Inc. All rights reserved.
Keywords: Motor learning; Sequence learning; SRT; Parietal; Arterial spin-labeling; Perfusion imaging; Premotor; Plasticity; fMRI; Neuroimaging;
Brain; Insula
1. Introduction
Some mental operations of interest to the cognitive neu-
roscientist evolve over relatively long time-scales. Examples
of these include changes in emotional state, adoption of a
particular cognitive set during the performance of a task,
or the e
Vects of sleep and alertness. A particularly salient
case is the cognitive process of learning, examples of which
produce enduring changes in performance that accumulate
slowly over minutes to hours. Continuous motor sequence
learning, in which a subject becomes skilled at the execution
of an ordered set of motor movements, is of this kind
(
Nissen & Bullemer, 1987
).
Attempts to study learning, or other slow changes in neu-
ral activity, with BOLD fMRI face the obstacle that BOLD
fMRI data are rather unstable at long time scales, as the sig-
nal tends to drift up and down over time (
Zarahn, Aguirre, &
DEsposito, 1997
). This presents an obvious limitation for
studies that attempt to discern the slow neural changes that
are associated with continuous learning: it is very di
Ycult to
discriminate the changes in imaging signal that are due to
learning from those that are present as drift noise.
In this study, we examine the suitability of a relatively new
imaging technique for the study of continuous, gradual
changes in neural activity. Arterial spin labeled (ASL) perfu-
sion imaging permits the noninvasive quanti
Wcation of
regional brain tissue perfusion using labeled in
Xowing arte-
rial protons as an endogenous tracer (
Alsop & Detre, 1998
).
Label images include a radiofrequency irradiation,
1
aimed
at the carotid and vertebral arteries, that precedes the image
*
Corresponding author.
E-mail address:
aguirreg@mail.med.upenn.edu
(G.K. Aguirre).
1
It should be noted that radioactive isotopes are not used in perfusion
imaging. I.R. Olson et al. / Brain and Cognition 60 (2006) 262271
263
acquisition. Label images are alternated with control
images, and the di
Verence in signal between adjacent image
pairs yields the signal due to perfusion. The perfusion e
Vects
of ASL are independent of the pulse sequence used to obtain
the image after the label. For example, if echoplanar images
are used, then the raw image data also contain BOLD con-
trast, which is attenuated during subtraction of control and
labeled pairs. This allows BOLD and perfusion e
Vects to be
compared within the same data set (
Wong, Buxton, & Frank,
1997
). We have previously shown (
Aguirre, Detre, Zarahn, &
Alsop, 2002
) that the perfusion fMRI signal is stable at long
time scales, indicating that it may be useful for the study of
slow changes in neural activity.
We obtained neuroimaging data from subjects while
they performed a motor serial response time (SRT) task.
During scanning, subjects performed a series of
Wnger
movements in response to visual cues over a 20 min period.
Unbeknownst to the subject, there was a repeated pattern
to the movements. Under these circumstances, subjects gen-
erally demonstrate a gradual and continuous decline in
response time (e.g.,
Nissen & Bullemer, 1987
). Prior imaging
studies of subjects performing SRT tasks have demon-
strated a reduction of neural activity in response to motor
execution after training as compared to the start of train-
ing. Because the change in performance is slow and contin-
uous, we assumed for this study that the neural correlate of
performance improvement during SRT training is a grad-
ual reduction in regional activity. Our goal in the current
study was to use perfusion fMRI in an attempt to detect
these neural changes. A positive result would demonstrate
the ability of perfusion fMRI to measure dynamic changes
in CBF over long time scales. Any e
Vects found in the per-
fusion data can be contrasted with e
Vects present in the
simultaneously acquired BOLD data in each subject.
Finally, because BOLD imaging is better suited to the
detection of transient neural activity, we examined if the
BOLD data could be used to detect trial-wise di
Verences
between correct and incorrect responses.
To foreshadow our results, insu
Ycient power was pres-
ent to detect learning e
Vects at a map-wise level in the per-
fusion data. However, within regions of interest de
Wned
with a lowered map-wise threshold, separate statistical tests
revealed signi
Wcant correlations of CBF changes with reac-
tion time measures of learning. These e
Vects could not be
detected within the BOLD data. However, there were mea-
surable, event-related di
Verences in the BOLD signal
between correct and incorrect responses. Thus, this study
serves as a demonstration of the potential of perfusion
fMRI data to simultaneously examine rapid changes in
neural activity with BOLD contrast and slow changes in
neural activity with perfusion contrast.
2. Methods
2.1. Participants
Ten participants (ages 1940, average age D 25, 5 males)
were recruited from the University of Pennsylvania and
received payment for participation. All subjects had normal
or corrected-to-normal visual acuity, were right handed,
and were free of any history of neurological or psychiatric
disease. Written consent was given according to an Institu-
tional Review Board approval from University of Pennsyl-
vania.
2.2. Experimental procedure: SRT task
The classic serial-response (SRT) task was adopted as
the learning task during perfusion fMRI scans. Subjects
were required to use four
Wngers (left middle, left index,
right middle, right index) to press keys as quickly and as
accurately as possible in response to the presentation of
visual cues that consisted of four white outline squares
(2.65° £ 2.65°, separated by 0.05°) that were arrayed hori-
zontally on a medium gray (RGB 127) background (see
Fig. 1
). Each square would illuminate with a color that cor-
responded to those present on the response box: red, green,
blue, or yellow. The spatial organization of the colors
remained consistent throughout the experiment. The color
target appeared for 500 ms and was followed by a 300 ms
Fig. 1. Schematic drawing of the task (stimuli are not to scale). Four horizontally arrayed boxes lit up sequentially to indicate the required keypress. 264
I.R. Olson et al. / Brain and Cognition 60 (2006) 262271
blank response window. This was followed by a 200 ms ITI
after which the next color target was shown.
Prior to scanning, each subject participated in 11 prac-
tice trials and task-related questions were answered. The
practice trials used a di
Verent sequence than that presented
during scanning. The scanning protocol consisted of three,
25.4 min long, functional scans and one, 6 min anatomical
scan at the end of the scan session. Each functional scan
had three ordered parts: (1) a 2.5 min long baseline
Wxation
condition; (2) a 20.9 min long SRT task condition; and (3) a
Wnal, 2 min long Wxation condition.
The two baseline
Wxation conditions consisted of a gray
screen with a white, central
Wxation cross. There was no task
requirement. The SRT task condition consisted of two parts.
The
Wrst part, the sequence learning session, consisted of an 11-
item
Wxed-order sequence that was repeated 86 times for a
total time of 15.8min. Subjects were not told that the
sequence would repeat. The repeated sequence was computer-
generated and designed to meet these criteria: a color target
could not repeat itself (e.g., green, red, red, blue) and each
color was used at least one time per 11-item sequence. This
sequence was repeated without break until 86 repetitions had
occurred. The second part of the SRT task was the transfer
session, consisting of 28 randomly generated, 11-item
sequences (5.1 min). All surface features of the transfer session
were identical to that used in the sequence learning (except the
stimulus order) and the subject was given no indication that
the task had changed. These sequences were subject to the
same criteria as the repeated sequences. The purpose of the
transfer phase was to ensure that any gains in RT were
sequence-speci
Wc, and not due to general task, or procedural
learning. The three functional scans were identical except for
the particular order of color cues in the SRT task.
2.3. Behavioral data analysis
All analyses were performed on data that were collapsed
across the th