Biomed Imaging Interv J 2006; 2(3):e27
© 2006 Biomedical Imaging and
Effects of single-trial averaging on spatial extent of brain activation detected by fMRI are subject and task dependent
Shin‑Yi Fang1,2, MS,
Jiun‑Jie Wang1,3, PhD,
Yuan‑Yu Hsu4, MD,
Yung‑Liang Wan1,3, MD,
Yau‑Yau Wai1,3, MD,
Ho‑Ling Liu1,3, PhD, DABR
1 MRI Center, Chang Gung Memorial Hospital,
2 Department of Psychology, University of Connecticut, Connecticut,
3 Department of Medical Imaging and Radiological Sciences, Chang
Gung University, Taoyuan, Taiwan;
4 Department of Medical Imaging, Buddhist Tzu Chi General Hospital,
Aim: The effects of single-trial averaging on the
spatial extent of event-related fMRI activation may vary between subjects and
tasks. The purpose of this study was to evaluate this variability using a
visual task and a word generation task.
Patients, materials, and methods: Five Chinese
right-handed male volunteers participated in the experiment. Experiments were
conducted using a 1.5 T clinical MRI scanner with a T2*-weighted single-shot
gradient-echo EPI sequence. Each task contained 150 trials that were separated
into 5 runs. For each voxel, time courses averaged across different numbers of
randomly selected trials, were obtained. They were applied for determining the
voxels with significant activations, using a students� t-test (p<0.001,
Results: Consistent with previous findings, the
number of the activated voxels increased monotonically with the number of
trials combined. The ascending rate and the maximum number of the activated
voxels were different, however, between tasks and among subjects.
Conclusions: The effects of single-trial averaging
were found to vary significantly between tasks and subjects. Therefore, we
strongly advise to carefully consider such variability when using the spatial
extent of activation as a measure in a group or a task comparison. � 2006 Biomedical
Imaging and Intervention Journal. All rights reserved.
Keywords: Spatial extent; event-related fMRI; trial
The spatial extent of the activation is a critical measure
in many fMRI studies [1-3], and the differences may be interpreted as
hypoactive or dysfunctional. In addition, the size of the spatial extent can be
used to reduce the detection of the false activation . The true regions of activation
tend to occur over contiguous voxels, where noise showed much less tendency to
form clusters . Few studies have, however, directly addressed those issues
related to the spatial extent of the BOLD response.
One previous study found that an exponential relationship existed
between the number of trials and the spatial extent. Over 100 trials were averaged
in this event-related study, using a visual stimulation task . Saad et al.
also reported the spatial extent increased monotonically with trial averaging
in a block-designed visual task .
Recently, event-related fMRI has been widely applied in the
exploration of human cognitive process . The shape of the hemodynamic response
function, however, varies between areas in the visual and the motor cortices
even in the same subject . Therefore, the effect on the spatial extents in averaged
trials would remain unclear in different brain areas using different stimuli, such
as, cognitive tasks. The optimised number of trials for the steady-state
response in a cognitive experiment using event-related fMRI design requires
further investigation. �
In this study, a cognitive task using word generation 
was designed to examine the effects of single-trial averaging. The result was
compared with that from a visual task. Different effects between both tasks
will be discussed and an optimised value of averaging trials will be considered
for the future applications. It should be noted that the spatial extents
detected by fMRI can be interpreted as involvement of neural activity only if
the steady-state condition in single-trial averaging is reached, i.e.,
independent on contrast-to-noise ratio (CNR). Therefore, the current study is
crucial for the experiments studying graded activations, as well as clinical
fMRI where CNR of the hemodynamic responses in patients can be quite different.
Materials and Methods
The experiments were conducted at the MRI Centre, Chang Gung
Memorial Hospital, LinKou, Taiwan. Five male volunteers participated in this
study. They were aged 20-26 years, with a mean age of 22.4 years. They were
native Chinese speakers. The participants had no history of neurological or
psychiatric disorders and reported to be right handed, which was confirmed with
the Handiness inventory . A written, informed consent was obtained in all
Neural activation was monitored by a 1.5 T Magnetom Vision
MRI scanner (Siemens, Erlangen, Germany) with a single-shot gradient-echo EPI
sequence (TR / TE /FA = 1000 ms / 60 ms / 90�, Slice Thickness = 8 mm, in-plane resolution = 3.3 mm). The stimuli were shown through a goggle display system (Resonance Technology Inc., CA, USA). Corrective lenses were used when necessary. Prior to the MR imaging,
each subject was visually familiarised with the procedures and the experimental
conditions, to minimise anxiety and enhance task performance. The subject,
before being transferred to the scanner, was fitted with a plastic earmuff and
a tightly fitting, thermally moulded, plastic facial mask that extended from
the hairline to the chin.
Subjects participated in both event-related fMRI
experiments, a visual stimulus and a word generation task, respectively, on two
separate days. This was done to reduce subject discomfort in a prolonged
scanning session. The time span between both experiments was set at one week.
The experiment consisted of five separate runs of total 150 trials. Each run was
made of 460 s, starting with 10-s dummy scans and followed by 30 trials (0.5 s
stimulus + 14.5 s fixation). The visual task was a black and white circular
checkerboard flashing at 8Hz. Seven contiguous slices were acquired parallel to
the calcarine sulcus. In the word generation task a different, single Chinese character
was displayed each time, and the subject was asked to silently associate it with
a semantically related word. All characters used were sampled from the Mandarin
Promotion Council (Ministry of Education, Taiwan), with a frequency ranging
between 15 and 25 per 10 million occurrences. Seven contiguous slices were
acquired parallel to the AC-PC line.
The data analyses for each subject were performed within an
anatomically defined region of interest (ROI) within the middle slice. ROIs
were chosen, based on the anatomical images, in the gray matter along the
calcarine fissure in the primary visual cortex in visual tasks or at BA10, 44,
45, and 47 in word generation tasks . All analyses were conducted in Matlab (The Math Works, Inc., Natick, MA, USA). The time-serial data were
normalised to the averaged signal intensity of the 1st, 2nd, and 15th time
points within each single trial. The single trial populations were sampled randomly.
A series of averaged signals, varying in the number of trials combined, were then
computed for each subject. Twenty averaged signals were computed for each possible
number of trials combined. For each averaged signal, a Pearson product-moment
correlation coefficient was calculated between the mean time course and a gamma
variate function. Activated voxels were determined using a student�s t-test (t
>3.8, p < 0.001, uncorrected). The number of voxels that exceeded the
threshold was determined for each of the twenty averaged signals, at each number
of combined trials for each subject.
Figure 1 shows, from all five subjects, the number of the
activated voxels plotted against the number of trials combined. Consistent with
previous studies, the volume of activation increased with the number of trials
combined in the visual task as well as the word generation task, in all
subjects in this experiment. The number of activated voxels at x trials
combined, Vx , was then fitted with an exponential function:
[View this figure]
| Figure 1 The number of activated
voxels plotted against the number of trials combined
from all subjects.
where Vmax represented the number of total
expected activated voxels and ξ, the ascending rate, equalled to the
number of trials combined when Vx is equal to 63% of Vmax.
Table 1 showed the Vmax and ξ from all subjects in both
experiments, respectively. The ascending rate and the number of total expected
activated voxels varied significantly between tasks and among subjects.
[View this table]
| Table 1 The number of total expected
activated voxels, Vmax, and the ascending rate, ?,
from all subjects in both visual stimulation and work
Discussion and Conclusion
Our findings are consistent with the observations of Huettel
and McCarthy . The spatial extent of the fMRI activation in a visual task is
influenced by the number of trials combined. Similar results were observed in a
word generation task, which indicated that the spatial extent increased with
the number of trials combined in an event-related fMRI experiment; despite the
different tasks performed or brain areas studied.
Figure 1 shows that the number of the activated voxels did
not reach the plateau even after averaging 150 trials. Table 1 shows that the
maximum number of the detected active voxels is, however, close to Vmax
in each subject. This suggests that most of the activated voxels were
identified and is consistent with the report of Huettel and MaCarthy .
[View this figure]
| Figure 2 An example of the slide
used in the a) visual task; and b) word generation
Our experiment showed that the effects of single-trial
averaging on the spatial extent of an event-related fMRI study may vary between
tasks and among subjects. The spatial extent can be affected by the amplitude
of the hemodynamic response  or the voxel-wise noise. Our results indicate
that such effects should be taken into consideration in experiments using
different tasks and involving different brain areas or in a comparative study
on subjects. Similar findings were reported by Duann et al. , who
demonstrated that the hemodynamic response varied substantially across trials
as well as sessions, subjects, and brain areas. Purdon et al.  concluded
that there exists a wide range of noise variances in the BOLD signal within and
between subjects, which could affect the determination of the spatial extent.
The wide range of noise variances might have led to the
different ξ between tasks and subjects in our experiment. An optimised
number of trials have to be carefully considered in future applications, to
reduce the noise. The variances in Vmax in this study can be
explained by the between-subjects reproducibility effect as was reported by the
previous studies [16, 17]. From the current study and the within-subject
reproducibility of the spatial extent [16, 18-21], the use of the spatial
extent of the activation as an index to compare the results from different
groups and/or tasks could introduce errors. Furthermore, the use of spatial
extent as a measure of activation could be a source of variance in itself .
To conclude, ξ, the ascending rate of the detected
activation volume and Vmax, the number of total expected activated
voxels vary across different tasks and subjects when increasing number of
trials are combined. This may be due to the complex contrast and noise
characteristics from the various effects in the hemodynamic changes in
different task performances, brain areas, and other subject-dependent physiological
parameters. Special precautions should be taken in fMRI studies, when using
detected spatial extent as an index for quantitative comparisons.
Meador KJ, Allison JD, Loring DW, et al. Topography of somatosensory processing: cerebral lateralization and focused attention. J Int Neuropsychol Soc 2002;8(3):349-59.
Wang Y, Sereno JA, Jongman A, et al. fMRI evidence for cortical modification during learning of Mandarin lexical tone. J Cogn Neurosci 2003;15(7):1019-27.
Vannini P, Almkvist O, Franck A, et al. Task demand modulations of visuospatial processing measured with functional magnetic resonance imaging. Neuroimage 2004;21(1):58-68.
Friston KJ, Worsley KJ, Frackowiak RSJ, et al. Assessing the significance of focal activations using their spatial extent. Hum Brain Mapp 1994;1:214-20.
Roland PE, Levin B, Kawashima R, et al. Three-Dimensional analysis of clustered voxels in 15O-Butanol brain activation images. Hum Brain Mapp 1993;1:3-19.
Huettel SA, McCarthy G. The effects of single-trial averaging upon the spatial extent of fMRI activation. Neuroreport 2001;12(11):2411-6.
Saad ZS, Ropella KM, DeYoe EA, et al. The spatial extent of the BOLD response. Neuroimage 2003;19(1):132-44.
Buckner RL, Bandettini PA, O'Craven KM, et al. Detection of cortical activation during averaged single trials of a cognitive task using functional magnetic resonance imaging. Proc Natl Acad Sci U S A 1996;93(25):14878-83.
Miezin FM, Maccotta L, Ollinger JM, et al. Characterizing the hemodynamic response: effects of presentation rate, sampling procedure, and the possibility of ordering brain activity based on relative timing. Neuroimage 2000;11(6 Pt 1):735-59.
Tan LH, Spinks JA, Gao JH, et al. Brain activation in the processing of Chinese characters and words: a functional MRI study. Hum Brain Mapp 2000;10(1):16-27.
Snyder PJ, Harris LJ. Handedness, sex, and familial sinistrality effects on spatial tasks. Cortex 1993;29(1):115-34.
Tan LH, Liu HL, Perfetti CA, et al. The neural system underlying Chinese logograph reading. Neuroimage 2001;13(5):836-46.
Carpenter PA, Just MA, Keller TA, et al. Time course of fMRI-activation in language and spatial networks during sentence comprehension. Neuroimage 1999;10(2):216-24.
Duann JR, Jung TP, Kuo WJ, et al. Single-trial variability in event-related BOLD signals. Neuroimage 2002;15(4):823-35.
Purden PL, Solo V, Brown E, et al. fMRI noise variability across subjects and trials: Insights for noise estimation methods. Neuroimage 1998;7:s617.
Machielsen WC, Rombouts SA, Barkhof F, et al. FMRI of visual encoding: reproducibility of activation. Hum Brain Mapp 2000;9(3):156-64.
Vandenbroucke MW, Goekoop R, Duschek EJ, et al. Interindividual differences of medial temporal lobe activation during encoding in an elderly population studied by fMRI. Neuroimage 2004;21(1):173-80.
Noll DC, Genovese CR, Nystrom LE, et al. Estimating test-retest reliability in functional MR imaging. II: Application to motor and cognitive activation studies. Magn Reson Med 1997;38(3):508-17.
Rutten GJ, Ramsey NF, van Rijen PC, et al. Reproducibility of fMRI-determined language lateralization in individual subjects. Brain Lang 2002;80(3):421-37.
Chee MW, Lee HL, Soon CS, et al. Reproducibility of the word frequency effect: comparison of signal change and voxel counting. Neuroimage 2003;18(2):468-82.
Neumann J, Lohmann G, Zysset S, et al. Within-subject variability of BOLD response dynamics. Neuroimage 2003;19(3):784-96.
Cohen MS, DuBois RM. Stability, repeatability, and the expression of signal magnitude in functional magnetic resonance imaging. J Magn Reson Imaging 1999;10(1):33-40.
Received 19 January 2006; received in revised form 12 April 2006; accepted 20 April 2006
Correspondence: MRI Center, Chang Gung Memorial
Hospital, 5 Fuhsing Street, Kweishan, Taoyuan, 333, Taiwan.
Tel: +886-3-3281200 ext. 8407; Fax: +886-3-3275798; E-mail:
Please cite as: Fang SY, Wang JJ, Hsu YY, Wan YL, Wai YY, Liu HL, Effects of single-trial averaging on spatial extent of brain activation detected by fMRI are subject and task dependent, Biomed Imaging Interv J 2006; 2(3):e27
This article has been viewed 2690 times.