Brain and Language 71, 391–399 (2000)
doi:10.1006/brln.1999.2272, available online at http://www.idealibrary.com on

Using fMRI to Study Recovery from Acquired Dysphasia
Gemma A. Calvert,* Michael J. Brammer,† Robin G. Morris,†
Steve C. R. Williams,† Nigel King,‡ and Paul M. Matthews*
*Centre for Functional Magnetic Resonance Imaging of the Brain (FMRIB), University of
Oxford, John Radcliffe Hospital, Headington, Oxford OX3 9DU; †Institute of Psychiatry,
De Crespigny Park, Denmark Hill, London SE5 8AF; and ‡Rivermead Rehabilitation
Centre, Abingdon Road, Oxford OX1 4XD United Kingdom
We have used functional magnetic resonance imaging (fMRI) to characterize
brain activations associated with two distinct language tasks performed by a 28year-old woman after partial recovery from dysphasia due to a left frontal hemispheric ischemic stroke. MRI showed that her ischemic lesion extended posteriorly
from the left inferior frontal to the perisylvian cortex. fMRI scans of both language
tasks revealed substantial differences in activation pattern relative to controls. The
nature of this difference was task-specific. During performance of a verbal semantic
decision task, the patient, in contrast to controls, activated a network of brain areas
that excluded the inferior frontal gyrus (in either hemisphere). A second task involving rhyme judgment was designed to place a heavier cognitive load on language
production processes and activated the left inferior frontal gyrus (Broca’s area)
strongly in normal controls. During this task, the most prominent frontal activation
in the patient occurred in the right homologue of Broca’s area. Subsequent analysis
of this data by methods able to deal with responses of changing amplitude revealed
additional, less sustained recruitment by the patient of cortex adjacent to the infarct
in the region inferior to Broca’s area during rhyming. These results suggest that in
addition to changes in cognitive strategy, recovery from dysphasia could be mediated by both the preservation of neuronal networks in and around the infarct and
the use of homologous regions in the contralateral hemisphere.  2000 Academic Press
Key Words: fMRI; stroke; dysphasia; language; rehabilitation.

We gratefully acknowledge MRC programme grant support for the FMRIB Centre and for
stroke studies (PMM). G.A.C. is supported by a Research Fellowship from the Medical Research Council. The authors thank Chris Andrew for technical support and Dr. Judith Allinson
at the Rivermead Rehabilitation Centre for her enthusiastic clinical assistance.
Address correspondence and reprint request to Gemma A. Calvert, FMRIB Centre, University of Oxford, John Radcliffe Hospital, Headington, Oxford, OX3 9DU United Kingdom.
Fax: 01865 222717. E-mail: gemma@fmrib.ox.ac.uk.
391
0093-934X/00 $35.00
Copyright  2000 by Academic Press
All rights of reproduction in any form reserved.

392

CALVERT ET AL.

INTRODUCTION

The mechanisms of recovery from dysphasia following infarction of left
hemispheric language zones remain elusive. Modern functional neuroimaging techniques offer great potential for elucidating these mechanisms but
have had only limited success to date in resolving many of the outstanding
issues raised in the clinical and anatomical literature. While some imaging
studies have linked recovery to the recruitment of homologous sites in the
nondominant hemisphere (Weiller et al., 1995; Ohyama et al., 1996), others
suggest instead that improvement is achieved by activation of residual left
hemispheric language areas (Heiss et al., 1993, 1997). Functional neuroimaging data in healthy subjects have shown that linguistic tasks differ in the
extent to which they activate the network of language centers distributed
through the frontal lobe and along the perisylvian region in temporal and
parietal cortex (Frackowiak, 1994). In view of these findings, an incomplete
picture of function in these regions in patients with dysphasia might result
from reliance on a single experimental paradigm.
Functional magnetic resonance imaging (fMRI) is emerging as a powerful
tool for mapping patterns of brain activation associated with language processing. Potential advantages of fMRI over positron emission tomography
(PET) or single photon emission computed tomography (SPECT) for functional imaging of patients include the ability to map activations sensitively
in single individuals, lack of restrictions on the number or frequency of studies that can be performed, and a relatively higher spatial and temporal resolution. Here we report a case study illustrating both the potential advantages
and problems associated with the clinical application of fMRI for defining
possible mechanisms of recovery from dysphasia.
CASE REPORT

A 28-year-old right-handed woman suffered a left hemispheric ischemic
stroke due to a spontaneous left carotid dissection. On admission 1 h after
the onset of symptoms she was mute, with a severely depressed level of
consciousness (GCS 8). An MRI scan performed at day 3 showed an abnormal hyperintense signal on the T2-weighted images extending posteriorly
from the inferior frontal gyrus through superior frontal regions to the angular
gyrus.
Over the course of the subsequent 2 weeks the patient recovered normal
awareness and was able to follow simple commands despite severe expressive dysphasia. During the next 5 months she received intensive rehabilitation therapy. Despite substantial recovery of language function, the patient
continued to display marked word finding and sentence construction difficulties and speech was effortful for all but commonplace phrases. Although
comprehension was considerably less impaired than production, as indicated
by her continued enjoyment of reading in addition to neuropsychological

STUDYING ACQUIRED DYSPHASIA WITH fMRI

393

assessment, the patient maintained a persistent moderate semantic impairment at the single word level (Psycholinguistic Assessments of Language
Processing in Dysphasia (PALPA) 48: 36/40) and demonstrated difficulty
understanding plausibly reversible sentences. Neuropsychological testing
demonstrated a full-scale IQ of 88 (Verbal ⫽ 82; Performance ⫽ 101) with
an estimated premorbid IQ in the ‘‘above average’’ range. Educational attainment included 3 A levels and Registered General Nursing status.
METHODS
To investigate brain activation associated with the patient’s recovery of language comprehension and production 5 months after the stroke, the patient and three female controls matched
for age, handedness, and educational background underwent fMRI scanning during the performance of two distinct linguistic tasks. The tasks were selected on the basis that they had
previously been shown to activate large areas of the language network (Bullmore et al., 1996;
Calvert, unpublished findings) and were also presumed to differ in the extent to which they
taxed language comprehension and production processes, respectively (Shaywitz et al., 1995).
Both paradigms employed a standard boxcar design with 30 s of an experimental stimulus
alternating with 30 s of a control stimulus for a total scanning time of 5 min per experiment.
Responses were recorded via a pushbutton box held in the left hand and accuracy scores noted
after scanning. The visible stimuli were presented via a back-projection screen located at the
end of the scanner bed and were visible by means of a mirror strategically angled above the
subject’s head in the scanner. Verbal stimuli were presented in blue against a white background. Prior to scanning, each subject was shown examples of the stimuli and completed
practice trials outside the scanner. Task difficulty was matched between controls and the patient
by calibrating trial rate to performance scores outside the scanner. For Task 1, the patient and
controls were both able to complete the task with the same trial intervals; for Task 2, the
patient required fewer trials to perform to the same level as controls.
Task 1 was a verbal semantic decision task which involved judging whether a presented
word belonged to the category of living or nonliving things compared with a resting control
condition. English nouns were visually presented on a screen every 3 s and subjects were
asked to indicate by pressing a button when the item they saw belonged to the category of
living things.
Task 2 was a rhyme judgment task in which pairs of pseudohomophones (nonwords that
obey lexical rules, e.g., ‘‘flait’’ and ‘‘klate’’) were presented on a screen during the experimental condition and subjects were asked to indicate by pressing a button if the two words on
each trial rhymed. In the control condition, subjects were presented with either ‘‘XO’’ or
‘‘XX’’ and were asked to press the button for ‘‘XX’’ only. This contrast was designed to
control for the motor component of the task and low level visual analysis.
Gradient echo echoplanar imaging (EPI) MRI data were acquired using a 1.5T GE Signa
scanner with Advanced NMR operation console with a quadrature birdcage head coil. One
hundred T2-weighted images depicting bold oxygen level-dependent (BOLD) contrast (Ogawa
et al., 1990) were acquired with an in-plane resolution of 3 mm (TE 40 ms, TR 3000 ms) at
each of 14 near-axial noncontiguous 7-mm-thick slices (with 0.7-mm interslice gap) positioned
parallel to the anterior–posterior commissural line. For each subject a multislice inversion
recovery echoplanar image was also acquired in the same plane for spatial normalization.
Correction for slight subject movement during image acquisition was performed using standard
algorithms (Friston et al., 1996). Activations were determined using time series regression
with a significance threshold determined at p ⬍ .004 from randomization tests (Bullmore et al.,
1996). Activation maps for both the patient and controls were then transformed into standard
stereotactic space of Talairach and Tournoux (Brammer et al., 1997) and overlaid onto a

394

CALVERT ET AL.

structural template or, in the case of the patient, on her own structural scan following suitable
spatial transformation. This allowed identification of the activations with respect to her infarct
as well as determination of Brodmann areas (BA) and thus comparison with other imaging
studies. In the case of the rhyme judgment task, data from the individual activation maps are
shown for clearer representation of the extent of individual variability. However, in this study
as well, Brodmann areas corresponding to activations were determined by group averaging
and mapping both control and patient data into standard stereotactic space.
Most commonly used fMRI analysis methods assume that the level of response (seen as a
change in image intensity) remains constant throughout the experiment. Failing to identify
when this is not the case will lead to a loss of sensitivity in detecting activations. Although this
phenomenon may occur in healthy subjects, for example during habituation, it is potentially a
much more serious problem among patients where pathological changes may affect the capacity to sustain cognitive performance. To address this problem, the data were subsequently
examined by wavelet analysis. This is a technique commonly employed when there is a requirement to look at different frequency components of a signal while preserving information
about where those frequency components might be concentrated in time (information which
is lost in traditional Fourier-based analysis methods). We have developed methods of fMRI
time-series analysis which identify clusters of significant wavelet coefficients by KolmogorovSmirnov tests, with correction for multiple comparisons (Brammer, 1998). These methods
were applied in the current study to identify brain regions in which there were responses at
the frequency of alternation of the experimental and control conditions which may not be
maintained for the duration of the experiment.

RESULTS

Analysis of the response data collected during both fMRI scans revealed
no significant difference in accuracy scores between the patient and the control group (close to 100% accuracy in both cases).
Task 1: Discrimination between Visually Presented Names of Living and
Nonliving Objects
For the controls as a group, the most prominent areas of activation included left occipital and fusiform gyri (extrastriate visual cortex; BA 18/19),
the left inferior frontal gyrus (BA 44/45), and left middle frontal gyrus (BA
44/9) (Fig. 1A). Smaller clusters of activated pixels were also detected in
the left superior temporal gyrus (BA 22), the right inferior frontal gyrus
(ventral to those observed in the left hemisphere), the anterior cingulate gyrus
(BA9/32), and the left angular gyrus (BA 39). By comparison, the patient
showed far more extensive bilateral activation of occipital and fusiform gyri
(BA 17; 18/19), a small cluster of activated pixels in the left middle frontal
gyrus (BA 9), and no activation of the left inferior frontal gyrus (either
around or adjacent to the infarct) (Fig. 1B). Small clusters of activation were
also detected in the left superior temporal gyrus, bordering on the occipitotemporal junction (BA 22/37) and, more extensively, in the precuneus (BA
31/7). Activation in the right inferior frontal gyrus (BA 44/45) was confined
to 2 pixels, comparable to that observed in the controls.

STUDYING ACQUIRED DYSPHASIA WITH fMRI

395

Task 2: Discrimination between Rhyming and Nonrhyming
Pseudohomophones
Regions consistently activated by controls included left occipital and fusiform gyri (BA 17/18/19), the superior temporal gyri (left greater than right)
(BA 22), and both middle frontal (BA 44/9) and inferior frontal (Broca’s
area) gyri (BA 44/45) in the left hemisphere. A smaller cluster of activated
pixels was found in the right inferior frontal cortex, although activation in
this region of the brain was lateralized strongly to the left, as can be observed
by inspection of the individual scans (Fig. 2A). The patient showed much
less activation overall but one of the two largest clusters of activation was
in the right inferior frontal cortex. Activation was also observed in the left
middle frontal gyrus (BA 6/9) and anterior cingulate (9/32). More intriguingly, there was a small cluster of pixels along the posterior border of the
infarct in the region of the left inferior frontal gyrus. Subsequent wavelet
analysis of the data revealed that for the patient, extensive activation was
detectable in the left inferior frontal gyrus, ventral to the infarct (Fig. 2B).
As this was not detectable by standard time-series regression (Bullmore et
al. 1996) it is likely that the responses showed considerable changes in amplitude during the experiment.
DISCUSSION

There are at least two important questions that can be addressed by functional neuroimaging studies of patients with acquired dysphasia. The first is
what are the neurophysiological changes that lead to the behavioral deficits?
The second is what mechanisms are responsible for recovery of language?
Our patient demonstrated a pattern of brain activation that was broadly consistent with her neuropsychological deficits according to current neural models of language processing (Frackowiak, 1994; Mazoyer et al., 1993).
Marked reduction in activation of Broca’s area and activation of undamaged
areas of posterior temporal cortex during both tasks is consistent with the
patient’s better performance on tasks involving language comprehension
compared with those involving language production. This basic pattern of
activation is in keeping with the location of her infarct, which extends more
prominently into the inferior frontal gyrus than posterior temporal and parietal sites. However, there were several less predictable observations that may
shed further light on the mechanisms underlying this patient’s recovery of
function, particularly with respect to articulation.
The most striking difference between the patient and the control group in
response to the semantic classification task was observed in extrastriate visual cortex. Activation of this area was much more extensive in the patient.
This is unlikely to reflect the transfer of certain linguistic processes to these
regions. It seems much more plausible to suppose that the patient may be
making more extensive use of visual processing during task performance,

FIG. 1. The figure shows generic brain activation maps for the patient (A) and the controls (B) for the visual semantic classification task. Activations
are superimposed on the morphology of the patient (A) and the control subjects (B) transformed into the standard stereotactic space of Talairach and
Tournoux. In the case of the control group, the morphology shown is the voxelwise average of the three subjects. Images are shown in radiological
convention, right side of the brain on the left side of the image and vice versa.

396
CALVERT ET AL.

FIG. 2. Row A depicts the brain slices which in each individual showed the largest extent of activation in Broca’s area (extending through BA
44/45) during performance of the rhyme judgment task. Activation in the right homologue of Broca’s area for controls was confined to one or two
voxels. The inset shows activations in the area surrounding the patient’s infarct detected by wavelet analysis (see text).

STUDYING ACQUIRED DYSPHASIA WITH fMRI

397

398

CALVERT ET AL.

suggesting a change in cognitive strategy rather than cortical plasticity. As
this case illustrates, therefore, mechanisms underlying functional recovery
from dysphasia need not be confined to the unmasking of functions in homologous right hemisphere sites or performance of residual tissue but might also
include the adoption of different processing strategies.
Had our investigation of this patient been confined to this one task, we
may have concluded that she lacked function both in the left inferior frontal
gyrus and its right hemisphere homologue. However, to probe function in
these areas further we included an additional language task (rhyme judgment) designed to place greater cognitive demand on these regions and by
doing so were able to elicit activation of the right inferior frontal gyrus in
the patient. Indeed, further examination of this data using wavelet analysis
revealed additional activations even in the peri-infarct region inferior to Broca’s area. The observation that wavelet analysis was necessary to define these
activations suggests that the responses showed an abnormally large variation
in response amplitude across the set of trials.
The findings of these two experiments indicate that in this patient, although there may be some preservation of tissue in left frontal language
areas, recruitment of nondominant hemispheric regions and novel processing
strategies also appear to play a significant role in recovery of function. As
this single case study shows, fMRI has the capacity to yield useful data on
the mechanisms underlying both functional impairment and recovery, but the
success of such studies may depend critically on matching task performance,
understanding the potential range of alternative processing strategies for various tasks, and application of sensitive tools for data analysis. Although functional activation studies can provide insight into the cerebral metabolic correlates of language processing both in normal subjects and dysphasic patients,
they cannot reveal the computations being performed in metabolically activated areas. Future studies might address this difficulty by combining transient functional interference using transcranial magnetic stimulation with
fMRI. In view of many fundamental problems in understanding dysphasia,
there is clearly room for well-designed case studies to contribute to the compilation of a database of information in this area.
REFERENCES
Brammer, M. J. 1998. Multidimensional wavelet analysis of functional magnetic resonance
images. In Human Brain Mapping. Vol. 6, 378–382.
Brammer, M. J., Bullmore, E. T., Simmons, A., Williams, S. C., Grasby, P. M., Howard,
R. J., Woodruff, P. W., & Rabe, H. S. 1997. Generic brain activation mapping in functional magnetic resonance imaging: A nonparametric approach. Magnetic Resonance Imaging, 15, 763–770.
Bullmore, E., Brammer, M., Williams, S. C., Rabe, H. S., Janot, N., David, A., Mellers, J.,
Howard, R., & Sham, P. 1996. Statistical methods of estimation and inference for functional MR image analysis. Magnetic Resonance Medicine, 35, 261–277.

STUDYING ACQUIRED DYSPHASIA WITH fMRI

399

Bullmore, E. T., Rabe, H. S., Morris, R. G., Williams, S. C., Gregory, L., Gray, J. A., &
Brammer, M. J. 1996. Functional magnetic resonance image analysis of a large-scale
neurocognitive network. Neuroimage, 4, 16–33.
Frackowiak, R. S. 1994. Functional mapping of verbal memory and language. Trends in Neuroscience, 17, 109–115.
Friston, K. J., Williams, S., Howard, R., Frackowiak, R. S., & Turner, R. 1996. Movementrelated effects in fMRI time-series. Magnetic Resonance Medicine, 35, 346–355.
Heiss, W. D., Emunds, H. G., & Herholz, K. 1993. Cerebral glucose metabolism as a predictor
of rehabilitation after ischaemic stroke. Stroke, 24, 1784–1788.
Heiss, W. D., Karbe, H., Weber, L. G., Herholz, K., Kessler, J., Pietrzyk, U., & Pawlik, G.
1997. Speech-induced cerebral metabolic activation reflects recovery from aphasia. Journal of the Neurological Sciences, 145, 213–217.
Mazoyer, B. M., Tzourio, N., Frak, V., Syrota, A., Murayam, N., & Levreir, O. 1993. The
cortical representation of speech. Journal of Cognitive Neuroscience, 5, 467–479.
Ogawa, S., Lee, T. M., Kay, A. R. & Tank, D. W. 1990. Brain magnetic resonance imaging
with contrast dependent on blood oxygenation. Proceedings of the National Academy of
Science USA, 87, 9868–9872.
Ohyama, M., Senda, M., Kitamura, S., Ishii, K., Mishina, M., & Terashi, A. 1996. Role of
the nondominant hemisphere and undamaged area during word repetition in poststroke
aphasics. A PET activation study. Stroke, 27, 897–903.
Shaywitz, B. A., Shaywitz, S. E., Pugh, K. R., Constable, R. T., Skudlarski, P., Fulbright,
R. K., Bronen, R. A., Fletcher, J. M., Shankweiler, D. P., & Katz, L. 1995. Sex differences
in the functional organization of the brain for language. Nature, 373, 607–609.
Weiller, C., Isensee, C., Rijntjes, M., Huber, W., Muller, S., Bier, D., Dutschka, K., Woods,
R. P., North, J., & Diener, H. C. 1995. Recovery from Wernicke’s aphasia: A positron
emission tomographic study. Annals of Neurology, 37, 723–732.