fMRI reveals large-scale network
activation in minimally
conscious patients
N.D. Schiff, MD; D. Rodriguez-Moreno, MS; A. Kamal, MD; K.H.S. Kim, MD, PhD; J.T. Giacino, PhD;
F. Plum, MD; and J. Hirsch, PhD
Abstract—Background: The minimally conscious state (MCS) resulting from severe brain damage refers to a subset of
patients who demonstrate unequivocal, but intermittent, behavioral evidence of awareness of self or their environment.
Although clinical examination may suggest residual cognitive function, neurobiological correlates of putative cognition in
MCS have not been demonstrated. Objective: To test the hypothesis that MCS patients retain active cerebral networks
that underlie cognitive function even though command following and communication abilities are inconsistent. Methods:
fMRI was employed to investigate cortical responses to passive language and tactile stimulation in two male adults with
severe brain injuries leading to MCS and in seven healthy volunteers. Results: In the case of the patient language-related
tasks, auditory stimulation with personalized narratives elicited cortical activity in the superior and middle temporal
gyrus. The healthy volunteers imaged during comparable passive language stimulation demonstrated responses similar to
the patients’ responses. However, when the narratives were presented as a time-reversed signal, and therefore without
linguistic content, the MCS patients demonstrated markedly reduced responses as compared with volunteer subjects,
suggesting reduced engagement for “linguistically” meaningless stimuli. Conclusions: The first fMRI maps of cortical
activity associated with language processing and tactile stimulation of patients in the minimally conscious state (MCS) are
presented. These findings of active cortical networks that serve language functions suggest that some MCS patients may
retain widely distributed cortical systems with potential for cognitive and sensory function despite their inability to follow
simple instructions or communicate reliably.
NEUROLOGY 2005;64:514–523
The minimally conscious state (MCS) refers to a sub-
category of patients with severe brain damage who
demonstrate unequivocal, but intermittent, behav-
ioral evidence of awareness of self or their environ-
ment.
1
The episodic nature of the interactions of
patients in MCS presents diagnostic challenges for
physicians and increases emotional burdens for fam-
ilies and caregivers. Herein we report the first fMRI
maps of brain activity of two MCS patients in re-
sponse to language and sensory stimulation para-
digms. Recent observations of fragments of cerebral
activity in persistent vegetative state (PVS) patients
provide evidence that at least some partially func-
tional cerebral regions can remain in catastrophi-
cally injured brains.
2-4
Such remaining cerebral
activity in PVS, however, appears isolated to small
regions of the forebrain that retain limited anatomic
and functional connection. Whereas patients in PVS
remain behaviorally unconscious, inconsistent evi-
dence of higher integrative brain function demon-
strated in some MCS patients invites further
investigation of potential residual cerebral function.
Moreover, comparisons of lesion volume and location
in patients remaining in a vegetative state after a trau-
matic brain injury with other patients remaining se-
verely disabled suggest that wide differences in
structural injury patterns are present in patients with
behavioral evidence of consciousness.
5
Evidence for conscious awareness is indirectly in-
ferred from behavior, and investigations of underly-
ing brain function utilizing fMRI provide a unique
opportunity to characterize the neurobiological corre-
lates of MCS. Emergence from MCS is contingent in
part on demonstration of a reliable communication
system.
1
Methods for interrogation of cortical sen-
sory and language systems are therefore of consider-
able potential importance, as evidence of residual
functional substrates that could support communica-
tion would motivate and guide further efforts at re-
habilitation. Based on previous findings in PVS
patients, we hypothesized that MCS patients would
fail to activate the complete cerebral network identi-
From the Department of Neurology and Neuroscience (Drs. Schiff, Kamal, and Plum) and Graduate School of Medical Sciences (D. Rodriguez-Moreno), Weill
College of Medicine, Cornell University, and Departments of Radiology and Psychology, fMRI Research Center (Dr. Hirsch), Center for Neurobiology and
Behavior, Columbia University, New York, Center for Functional and Molecular Imaging, Division of Clinical Pharmacology (Dr. Kim), Georgetown
University Medical Center, Washington, DC, and JFK Johnson Rehabilitation Institute (Dr. Giacino), Edison, NJ.
Supported by a National Cancer Institute Cancer Center Support Grant (J.H.), the Charles A. Dana Fund (N.D.S., F.P.), National Institute of Neurological
Disorders and Stroke R21 NS43451 (N.D.S., J.T.G., F.P., J.H.), and the Cornell–New York Presbyterian-NIH–Supported General Clinical Research Center
(N.D.S., F.P.).
Received May 21, 2004. Accepted in final form October 15, 2004.
Address correspondence and reprint requests to Dr. J. Hirsch, Functional MRI Research Center, Neurological Institute, Columbia University, 710 W. 168 St.,
Box 108, New York, NY 10032; e-mail: jh2155@columbia.edu
514 Copyright ? 2005 by AAN Enterprises, Inc.

fied in normal subjects in response to similar passive
listening paradigms. However, we rule out this hy-
pothesis for these two patients in favor of the alter-
native hypothesis that the networks remain largely
intact but show significant differences in their level
of responsiveness.
Methods. Control subjects and patients. Seven healthy volun-
teers without history of neurologic disorders or chronic disease
were recruited according to institutional informed consent proce-
dures and performed language-related tasks similar to those per-
formed by the MCS patients. All subjects were right handed as
assessed by the Edinburgh Handedness Inventory (average later-
ality quotient 76.43
24.76)
6
with mean age of 30
4.06 years
(table 1). One volunteer who was age and gender matched to the
patients also participated in a tactile study performed similarly to
the patients. Structural MRI revealed no abnormalities in any of
the healthy subjects.
Two right-handed male MCS patients, ages 21 (Patient 1) and
33 (Patient 2), were recruited for this study. Legally authorized
surrogates for both patients were contacted by medical personnel
not directly involved in the current studies. Informed consent was
obtained according to institutional guidelines on two occasions,
allowing for a period of evaluation and opportunity for additional
information. The patients had no history of neurologic disorder
prior to the episodes that led to their severe brain injuries. The
durations of the MCS at the time of the study were 18 (Patient 1)
and 24 (Patient 2) months. Serial bedside examinations conducted
by multiple examiners over a period of months were consistent
with the diagnosis of MCS.
Patient 1 experienced a spontaneous intracranial hemorrhage
in the left temporoparietal region with brainstem compression
injury. Vegetative state was diagnosed at 3 months and pro-
gressed over the course of first year to MCS. Neurologic evalua-
tion at the time of the study indicated right hemiparesis, intact
oculocephalic and optokinetic responses, visual tracking, and sac-
cades to stimuli. A large area of encephalomalacia over the left
temporoparietal region was seen in structural MRI. Resting flu-
orodeoxyglucose (FDG) PET demonstrated 38.6% of normal re-
gional cerebral metabolic rate. The highest-level behavioral
responses observed for this patient were one-step command fol-
lowing, inconsistent identification of objects via eye gaze, and
intelligible single-word verbalizations.
Patient 2 received blunt head trauma to the right frontal re-
gion that led to bilateral subdural hematomas and associated
brainstem compression injury. Neurologic evaluation at the study
time revealed released oculocephalic responses with intact visual
tracking and both saccades to stimuli and to command, marked
increased motor tone bilaterally, and frontal release signs. Right
frontal lobe encephalomalacia and a small right-sided paramedian
thalamic infarction were seen on structural MRI. Resting FDG-
PET demonstrated 40.6% of normal regional cerebral metabolic
rate. The highest-level behavior observed in this patient was his
ability to inconsistently follow complex commands including go/
no-go and countermanding tasks and occasional verbalization.
Image acquisition. A 1.5-T General Electric MR scanner (Mil-
waukee, WI) was used to obtain T2*-weighted images with a gra-
dient echo pulse sequence that was sensitive to MR signal changes
induced by alteration in the proportion of deoxyhemoglobin in the
local vasculature accompanying neuronal activation (repetition
time  4,000 seconds, echo time  60 seconds, flip angle  60°).
The in-plane resolution was 1.5  1.5 mm, and slice thickness was
4.5 mm. Twenty-one contiguous axial slices of the brain, which
covered the entire cortex, were taken parallel to the anterior–
posterior commissural plane. Thirty-six images were acquired for
each run: a baseline (resting) period of 14 images (56 seconds), a
stimulation period of 10 images (40 seconds), and a baseline (re-
covery) period of 12 images (48 seconds). The initial three images
of each run were not included in the analysis to allow the mag-
netic signal to reach a steady state. Conventional high-resolution
(T1-weighted) images were also acquired along the same plane
locations as the T2*-weighted images and served as anatomic
references.
Tasks. Three passive stimulation tasks were performed: light
touch of both hands, auditory narratives of familiar events pre-
sented by a familiar person, and the same auditory passages with-
out language-related content. Each condition was performed twice
for a total of six runs per subject. During the first two runs,
patients “listened” to a narrative prepared by a familiar relative
through headphones (Gradient Muff Headset; Resonance Technol-
ogy, Northridge, CA). A similar task was employed for the healthy
volunteers, where the narrative consisted of four 20-second En-
glish paragraphs chosen for neutral emotional content without
personally meaningful content.
7
The neutral content was em-
ployed to minimize cross-subject variability and was intended to
influence a response restricted to the essential language areas.
Similar methods to study passive language function are employed
for neurosurgical planning prior to tumor resection
8
and for as-
sessment of language function in unresponsive infants and chil-
dren.
9
In the subsequent scans, the narratives were played in time
reverse (backward) so that they were recognizable as speech, but
content, propositional, and prosodic information were absent. Re-
versed speech is acoustically matched to normal speech in terms
of duration, amplitude, and power spectrum but lacks the tempo-
ral attributes contained in the phase relationships in the forward
signal. The final task consisted of two runs of passive tactile
stimulation of the two hands simultaneously by gently rubbing
the subject’s palm and fingers with a coarse-textured plastic sur-
face. These methods are routinely employed to localize sensory
and motor-related cortex for patients in preparation for neurosur-
gical procedures
8
and were selected for this study because of the
inflexible hand positions of the patients.
Image processing and data analysis. Prior to statistical anal-
ysis, images were computationally aligned using the Woods algo-
rithm,
10
and a two-dimensional Gaussian filter (approximately 3
voxels, at half-height) was applied. The last image of the resting
baseline and the first two images of the recovery baseline corre-
sponding to transitions between epochs were also not part of the
statistical analysis. Each analyzed epoch consisted of 10 images
(40 seconds). Activation of each voxel was determined by a multi-
stage statistical process that compared mean amplitudes of sig-
nals acquired during stimulation and baseline periods (voxel 
voxel t maps) and identified voxels with signal changes between
baseline and activity epochs on two identical runs. This conjunc-
tion yielded an empirically determined false-positive rate of p 
0.0008 for each condition.
11
Because of the descriptive goal of this
study, patient images were analyzed as “fixed effects,” meaning
that the observations are valid for the subjects studied, but the
investigation is without the benefit of a sufficient patient sample
size to statistically generalize beyond the single subjects who par-
ticipated in the study. However, the small sample size is assumed
to be justifiable in these extraordinary and high-impact cases
where the experimental design relies on the internal consistency
of multiple measurements and comparison with matched healthy
volunteers.
Assignments of anatomic labels and coordinates for active re-
gions were based on correspondence between the patient’s brain
anatomy and the human brain atlas.
12
These assignments were
made without registration of the acquired brains to a normalized
brain owing to atypical morphology secondary to brain injury. The
stages of assignment included identification of the brain slice
passing through the anterior–posterior commissural line, assign-
Table 1 Healthy volunteers: Biographic information
Subject Age, y Gender Handedness
Laterality
quotient
A 24 M Right 100
B 29 F Right 100
C 32 M Right 100
D 29 M Right 50
E 28 F Right 56
F 37 M Right 47
G 32 F Right 82
Average 30.1 76.43
SD 4.06 24.76
February (1 of 2) 2005 NEUROLOGY 64 515