Neuroimaging

in Rehabilitation

Understanding Neuroimaging

Downloaded from https://academic.oup.com/ptj/article-abstract/87/6/670/2747230 by guest on 22 June 2020

Teresa Jacobson Kimberley, Scott M Lewis
TJ Kimberley, PT, PhD, is Assistant
Professor, Program in Physical Neuroimaging is an emergent method of investigation for studying the human brain
Therapy, MMC 388, University of
Minnesota, 420 Delaware St SE, in healthy and impaired populations. An increasing number of these investigations
Minneapolis, MN 55455 (USA). involve topics important to rehabilitation. Thus, a basic understanding of the more
Address all correspondence to Dr commonly used neuroimaging techniques is important for understanding and inter-
Kimberley at: [email protected]. preting this growing area of research. Included in this article is a description of the
SM Lewis, MD, PhD, is Assistant signal source, the advantages and limitations of each technique, considerations for
Professor, Department of Neurol- study design, and how to interpret cortical imaging data. Particular emphasis is placed
ogy, University of Minnesota Med- on functional magnetic resonance imaging because of its ubiquitous presence in
ical School, Brain Sciences Center rehabilitation research.
VA Medical Center, Minneapolis,
Minn.

[Kimberley TJ, Lewis SM. Under-

standing neuroimaging. Phys
Ther. 2007;87:670 – 683.]

© 2007 American Physical Therapy

Association

For more about this PTJ Special

Series on the role of neuroimag-
ing in rehabilitation, read the
editorial by Richard K Shields on
page 639.

Post a Rapid Response or

find The Bottom Line:
www.ptjournal.org

670 f Physical Therapy Volume 87 Number 6 June 2007

 Understanding Neuroimaging

Historical Background entific underpinnings of diseases, MEG

Mapping brain function has been a disorders, and recovery that effective Magnetoencephalography is a tech-
pursuit of scientists for more than 2 treatments can be developed. Neuro- nique that relies on the naturally oc-
centuries. Phrenologists proposed imaging is a powerful tool that may curring magnetic fields in living or-
the first popular method in the early be harnessed by rehabilitation spe- ganisms. These fields are found in
19th century. They believed that the cialists to improve understanding of association with the electric currents
amount of brain tissue devoted to a the mechanisms by which treat- that are part of normal brain func-
cognitive function determined its in- ments improve function. Therefore, tion. As ionic currents change during
fluence on behavior.1 Although they the purpose of this article is to pro- brain activity, there is an associated
were unable to measure cortical vol- vide a brief overview of the com- change in the detectable magnetic
ume directly, they assumed that in- monly used neuroimaging tech- fields. With MEG, it is believed that

Downloaded from https://academic.oup.com/ptj/article-abstract/87/6/670/2747230 by guest on 22 June 2020

creases in brain size would translate niques, with a particular emphasis the detectable neuronal signals orig-
into measurable bumps on the skull. on the technique most broadly used inate from dendritic activity in the
in the field of rehabilitation, fMRI. pyramidal cells of the cerebral cor-
Although phrenology lacked scien- This article was written for people tex.5 Thus, this technique provides a
tific rigor, it did introduce the con- without an extensive physics or im- direct measure of cellular activity.
cept of brain function localization, in aging background. We describe the
which different aspects of human be- foundation of fMRI, what to consider With the MEG technique, sensors ar-
havior are represented primarily in when one is designing or conducting ranged on the surface of the head
specific locations of the brain. Sub- an fMRI experiment, and how a cli- measure the magnetic fields. Typi-
sequent advances were made by nician may interpret the results of cally, the sensors are fixed in a hel-
studying the effects of damage to the research with fMRI. met design that is placed on the head
human brain, such as the localization of the subject in a seated or supine
of language in patients with stroke Neuroimaging Techniques position. The measurements made
by Paul Broca,2 or by studying the The following investigative tech- by each sensor reflect the synchro-
effects of stimulation on brain areas niques all have advantages and disad- nous activity of thousands of neu-
during neurosurgery, such as the lo- vantages that are based on the rons contained in the brain areas un-
calization of the motor cortex by strengths and limitations of each. We der the sensor, particularly the
Fritsch and Hitzig.3 By the end of the discuss the different techniques in activity that occurs in the sulci of the
20th century, the study of functional terms of the degree of invasiveness, brain.6 This pattern is attributable to
localization took a major step for- tasks that can be performed, spatial the nature of the magnetic field. The
ward with the development of new resolution, and temporal resolution. field follows the “right-hand rule,”
neuroimaging technologies. These Spatial resolution refers to how ac- whereby with the thumb pointed in
noninvasive or minimally invasive curately the measured activity is lo- the direction of the electric current,
methods allow the human brain to calized within the brain, and tempo- the magnetic field wraps around the
be studied in both healthy people ral resolution refers to how closely current in the direction of the fin-
and those affected by disease or in- the measured activity corresponds gers. If the current is perpendicular
jury. In addition, these approaches to the timing of the actual neural to the surface of the head, then the
allow for repeated measurements in activity. magnetic field will be parallel to the
the same subject. Among these tech- surface of the head and will not
niques, positron emission tomogra- It is important to appreciate that thread through the detection coils. If
phy (PET), transcranial magnetic none of the techniques discussed in the current is parallel to the surface,
stimulation (TMS), magnetoencepha- this article allows researchers to in- then the magnetic field will be ori-
lography (MEG), and functional mag- terpret brain activity at a scale that is ented ideally to exit and enter the
netic resonance imaging (fMRI) are available to invasive techniques, head (Fig. 1). This induced field is
used most commonly. which measure responses from sin- about 1 billion times smaller than the
gle neurons.4 However, given the rel- earth’s magnetic field, approxi-
To date, the fields of neuroscience ative advantages of these techniques mately 50⫻10–15 to 500⫻10–15 T.6
and psychology have dominated the over such invasive measures for For a detectable measurement to be
neuroimaging landscape, but a grow- studying brain activity in humans, obtained from this current, the sub-
ing number of researchers are using this type of research will remain a ject being studied and the MEG in-
these techniques to investigate staple for many years. strument must be in a magnetically
rehabilitation-related questions. It is shielded room.
only through elucidation of the sci-

June 2007 Volume 87 Number 6 Physical Therapy f 671

Understanding Neuroimaging

The MEG technique allows for a

wide variety of behavioral tests en-
compassing visual, auditory, somato-
sensory, and motor studies.5–7 Base-
line studies with no task also can be
performed to assess resting brain ac-
tivity. The data that are obtained
must be processed before areas of
task-related brain activity can be de-
termined. One consideration of MEG
analysis is the “inverse problem.” Al-

Downloaded from https://academic.oup.com/ptj/article-abstract/87/6/670/2747230 by guest on 22 June 2020

though it is possible to calculate un- Figure 1.
ambiguously the fields outside the Illustration of the relationship between the magnetic field and the scalp surface. Two
sources of electric current are indicated by the arrows, one on the brain surface and one
head given the currents within the
in the brain sulcus. The associated magnetic field is indicated by the dashed line. The
head, the reverse problem has no magnetic field associated with the electric current in the brain sulcus is oriented best to
unique solution. To restrict the re- be detected on the surface of the head.
sult to one solution representing the
most likely brain activity pattern,
one must add information in the
form of assumptions of simplicity or as a hydrogen analog) commonly are which brain activity during task per-
data from other studies, such as used. These isotopes have a short formance is compared with baseline
other neuroimaging or neurophysiol- half-life, ranging from 2 minutes to 2 activity, the temporal resolution is
ogy studies. hours (because of the short half-life typically 1 to 2 minutes and is limited
of the isotopes, a device to produce by both the technique and the me-
The MEG technique provides excel- the isotopes, a cyclotron, must be tabolism of the molecule. An advan-
lent temporal resolution, on the or- on-site or nearby, depending on the tage of PET studies is the variety of
der of 1 millisecond, thereby allow- isotope). With PET, one of these iso- data that can be obtained with the
ing determination of the temporal topes is incorporated into a biologi- various labeled molecules that are
sequence of brain activity.6 In addi- cally relevant molecule and injected available, such as 11C-labeled raclo-
tion, it potentially can provide a spa- into or inhaled by the subject. It then pride to study the dopamine D2 re-
tial resolution of 5 mm.6 The pro- concentrates in brain areas accord- ceptor, 15O-labeled water to study
cessed MEG data can be presented in ing to the molecule’s chemistry and blood flow, and 2-fluoro-2-deoxy-D-
several ways, including magnetic the brain’s metabolic and blood flow glucose to study cerebral metabolic
field contour maps of the surface of demands. These areas are detected activity. Because this technique ex-
the head or areas of brain activity when a positron is emitted from the poses the subject to radiation, a dis-
displayed on an image of the brain molecule. This positron is annihi- advantage is that repeated scanning
obtained by magnetic resonance im- lated upon interaction with a nearby is not possible. This disadvantage
aging (MRI). This technique can be (1–2 mm) electron, producing 2 limits task repetition and the perfor-
used to answer many motor, cogni- gamma rays almost perfectly 180 de- mance of multiple tasks in the same
tive, and sensory questions, particu- grees apart. The PET scanner detects session. As with MEG, the data can
larly when the timing of neural activ- the 2 oppositely directed gamma be presented in several ways, includ-
ity is of interest. This technique has rays when these rays hit 2 photo- ing maps of the distribution of the
potential in the field of rehabilita- detectors coincidently and col- labeled molecule or maps of task-
tion, as demonstrated by studies of linearly. These types of detection related changes in brain activity.
stroke8 and aphasia.9 form the basis of the PET image. These maps also can be displayed on
a brain image obtained by MRI.
PET Because a wide range of molecules
Positron emission tomography is a can be used, PET can provide mea- TMS
technique in which a radioactively surements of blood flow, blood vol- Transcranial magnetic stimulation is
labeled molecule is used (for a re- ume, brain metabolism (especially unique among the techniques used
view, see Cherry and Phelps10). The glucose), and neuroreceptor or neu- to investigate function in the human
label is a positron-emitting isotope. rotransmitter chemistry. The spatial brain (for a review, see Pascual-
In PET studies, isotopes of carbon, resolution of PET is approximately 4 Leone et al11). Unlike methods of
nitrogen, oxygen, and fluorine (used to 5 mm.10 In functional studies, in neuroimaging that provide measures

672 f Physical Therapy Volume 87 Number 6 June 2007

 Understanding Neuroimaging

of naturally occurring brain activity, tion is given immediately preceding For an understanding of how this
the TMS technique uses a magnetic or during task performance, thereby process occurs, knowledge of some
field to induce an electric current in interfering with the normal activity physics concepts is required. One
underlying brain tissue, thereby stim- of the brain area under the coil.12 water molecule contains 1 oxygen
ulating the neurons. This technique atom and 2 hydrogen atoms (H2O).
also can be used to map brain func- The principal drawback of this tech- The hydrogen atom has 1 proton
tion, elucidate brain areas involved nique is the direct stimulation of the that spins and can be thought of as a
in task performance by manipulat- brain, which carries the unlikely but small magnet that produces its own
ing brain activity, or potentially still notable risk of a seizure. There- magnetic signal. Under normal cir-
provide treatment as a therapeutic fore, proper investigator training and cumstances, protons are randomly
intervention. subject screening are required.13 In pointing in different directions, so

Downloaded from https://academic.oup.com/ptj/article-abstract/87/6/670/2747230 by guest on 22 June 2020

addition to the use of TMS as a brain- that there is no net magnetic field
With TMS, a coil of wire (in either a mapping device, repetitive TMS is (Fig. 2A, 1).22 The magnetic field in
circular or a figure eight configura- under investigation as a treatment an MRI machine is always “on,” so
tion) is placed above the area of in- modality for a variety of neurological that when a subject enters the mag-
terest on the head of the subject. An problems, such as depression, paral- net bore, all of the protons in the
electric current is passed through ysis, bradykinesia, writer’s cramp, body align with the external mag-
this wire, creating a magnetic field stroke, and more.14 –19 netic field, creating a net internal
that passes painlessly (in most cases) magnetic field (Fig. 2A, 2). During an
and unimpeded through the scalp MRI MRI, a second external magnetic
and skull into the brain below. This As the name implies, MRI uses strong field is applied via a radio-frequency
field induces an electric current, magnetic fields to create images of (RF) pulse, causing the protons to
thereby stimulating the neurons in biological tissues, taking advantage wobble around their axis like a top.
the brain. The spatial resolution is of properties that are intrinsic to tis- This wobbling creates a rotating
dependent on the type of coil, distri- sues of the brain and, as such, does magnetic field that changes with
bution of cerebrospinal fluid, and not expose the subject to radiation. time and generates an electric cur-
other factors and varies in the centi- MRI can provide information about rent in the receiver (Fig. 2A, 3). This
meter to millimeter range.4 The tem- several different atoms, such as hy- signal has both horizontal and verti-
poral resolution of the technique drogen, phosphorus, carbon, and so- cal components (Fig. 2A, 4) and is
must be thought of differently, as dium atoms. However, in both ana- used to create an image. When the
activity is being induced rather than tomical imaging and functional RF pulse ends, the protons gradu-
measured. imaging, the atom of interest is hy- ally align back to their original
drogen, specifically, the hydrogen at- orientation.
A stimulus can be given in under 1 tached to water.
millisecond, approximating the time- Magnetic resonance imaging com-
scale of neural activity. The stimula- The static magnetic field created by monly is directed at measuring 2 pro-
tion can be given as a single pulse, as an MRI scanner is expressed in the cesses of relaxation of the proton
paired pulses to a single brain area or unit Tesla. For comparison, the that are characterized by time con-
2 different brain areas, or as repeti- earth’s magnetic field is approxi- stants: T1 and T2 (Fig. 2B). These
tive pulses.11 This flexibility in pulses mately 0.00005 T.20 The magnetic processes take place as the wobbling
allows the study of single brain areas, fields of scanners used for structural protons relax back to their original
multiple brain areas and their inter- MRI are typically 1.5 to 3 T. To cre- state after the RF pulse ends. A T1-
actions, connectivity between areas, ate images, the scanner uses a series weighted scan measures the “right-
inhibitory and excitatory circuits, of changing magnetic field gradients ing” of the tipped proton as it re-
and the behavioral effects of modu- and oscillating electromagnetic aligns with the original magnetic
lating cortical excitability.4,11 Output fields, known as a pulse sequence, field. The rate of this relaxation is
measures can be electromyographic which is adjusted for the properties influenced by nonexcited molecules
responses in the muscles of interest, of hydrogen nuclei.21 The density in the surrounding tissue and is used
observable movement, patient re- and environment of hydrogen nuclei for differentiating grey matter from
ports (such as subtle visual percep- in different types of tissues allow white matter.23 In a T2-weighted
tions, ie, phosphenes, with occipital pulse sequences to differentiate scan, the focus is on the “falling out”
cortex stimulation), and a disruption among tissue types, such as liga- or dephasing of synchrony of the ro-
of task performance. The latter oc- ments, tumors, and grey or white tating protons.23 Dephasing occurs
curs when low-frequency stimula- matter in the brain. quickly and results largely from the

June 2007 Volume 87 Number 6 Physical Therapy f 673

 Downloaded from https://academic.oup.com/ptj/article-abstract/87/6/670/2747230 by guest on 22 June 2020

June 2007
Number 6
Volume 87
Understanding Neuroimaging

Physical Therapy
f
674
 Understanding Neuroimaging

loss of energy to spinning nuclei measure of this activity. Since the in the microvasculature of the brain.
nearby (and is influenced by the late 1800s it has been known that The signal in BOLD contrast images
quality of the magnet used). These blood flow increases with neural ac- is used to reveal where and with
time constants vary in different tis- tivity (for a historical review, see what intensity brain activity occurs
sue environments (such as grey mat- Friedland and Iadecola24). In 1936, it during a behavioral task, such as fin-
ter and blood vessels) and form the was discovered that the hemoglobin ger tapping or viewing a flashing
basis for differentiating healthy tis- molecule has magnetic properties checkerboard. This technique is
sues from diseased tissues. that differ depending on whether it based on the same principles as tra-
is bound to oxygen. Oxygenated he- ditional MRI, but the very fast imag-
Beyond the standard MRI technique moglobin (hemoglobin with oxygen ing sequence used in BOLD fMRI
that is useful for detecting pathology attached) has no magnetic proper- (called “echoplanar imaging”) is sen-

Downloaded from https://academic.oup.com/ptj/article-abstract/87/6/670/2747230 by guest on 22 June 2020

or trauma, different MRI acquisition ties, whereas deoxygenated hemo- sitive to blood-based properties. In-
procedures can be used to generate globin (hemoglobin without oxygen creased neural activity within a par-
functional maps. These functional attached) does.25 An increase in ticular brain region results in an
maps describe brain activity occur- brain activity alters the ratio be- increase in blood flow and leads to
ring in awake and performing hu- tween the 2 forms of hemoglobin. In decreased concentrations of deoxy-
mans and can be acquired on the the early 1990s, it was determined genated hemoglobin in nearby ves-
basis of regional changes in tissue that this phenomenon could be used sels. This situation is somewhat
perfusion, cerebral blood volume, or to investigate brain function with counterintuitive, but the decrease is
the ratio of oxygenated hemoglobin what is known as the BOLD signal attributable to an excessive increase
to deoxygenated hemoglobin sec- (T2*-weighted image).26 in the supply of oxygenated blood
ondary to neural activity. The former compared with what is required for
2 factors have not been used widely As explained above, fMRI is an indi- the increased neural activity. The rel-
in functional imaging studies; there- rect measure of neural activity, but ative decrease in deoxygenated he-
fore, the focus of this article is on the recently the linkage between BOLD moglobin concentrations results in
latter, which is referred to as blood contrast and neural activity was ex- higher intensities on fMRI scans.
oxygen level– dependent (BOLD) plored in both animals27,28 and hu-
contrast. This is the source of the mans.29 Significant correlations were The fast acquisition time of fMRI al-
signal change in fMRI. T2 weighting, found between the BOLD signal and lows whole brain images to be col-
with an additional factor added to both the local field potential27 and lected in about 3 seconds, meaning
account for the nonhomogeneity of the neuronal firing rate.29 Although that hundreds of fMRI volumes (ie,
the magnetic field (and thus referred more work is needed to clarify this one set of slices through the brain)
to as “T2* weighting”), is the basis of relationship, these studies confirmed can be collected in a given experi-
fMRI.23 that the BOLD signal reflects neural ment. However, because of this fast
activity. acquisition time and the need for an
BOLD Contrast Method adequate signal, fMRI does not have
The BOLD contrast method is based The BOLD contrast in a brain region the fine degree of spatial resolution
on the hemodynamic response to is dependent on the balance of oxy- that traditional anatomical (T1-
neural activity and is thus an indirect gen consumption and oxygen supply weighted) MRI has. After several pre-

4™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™
Figure 2.
(A) Magnetic resonance signal. (1) In nonexcited tissue, protons spin around their axes, creating individual magnetic fields with
random directions. (2) When a vertical magnetic field is applied to the tissue, the protons align with it to create a net magnetic field
that is also vertical but is very small and difficult to detect. (3) A radio-frequency (RF) pulse creates a second magnetic field that
changes with time and gives rise to an electric current that is ultimately measured by magnetic resonance imaging (MRI). (4) The
net magnetic field can be divided into vertical and horizontal components; MRI measures the changes in these 2 components as the
protons respond to the applied magnetic fields and RF pulses. (B) Relaxation processes emphasized in MRI. A typical MRI sequence
begins when a subject is placed in a vertical magnetic field. With the protons aligned vertically, a horizontal RF pulse is applied to
the tip of the wobbling protons so that they rotate in the horizontal plane synchronously (in phase) with each other (1). The
horizontal pulse then is stopped (2), and the rotating protons begin to move out of phase with each other—they dephase. Dephasing
occurs relatively quickly and leads to a loss of horizontal magnetization. The time constant of decay is T2. After the withdrawal of
the horizontal pulse, the protons realign with the vertical magnetic field, and vertical magnetization is restored (3–5). This righting
of the protons occurs more slowly than dephasing and is measured indirectly. The time constant of the recovery of longitudinal
magnetization is T1. (Reprinted with permission of The McGraw-Hill Companies from Kandel E, Schwartz J, Jessell T. Principles of
Neural Science. 4th ed. New York, NY: McGraw-Hill; 2000.)

June 2007 Volume 87 Number 6 Physical Therapy f 675

Understanding Neuroimaging

processing steps, the brain areas

showing activity on images from
fMRI are coregistered with (overlaid
on) anatomical images to visualize
anatomical landmarks and provide
the differentiation of grey matter
from white matter. Figure 3 shows
the difference between the 2 types
of acquisition in the same subject;
Figure 3A shows an example of an
axial slice of the brain during a T1-

Downloaded from https://academic.oup.com/ptj/article-abstract/87/6/670/2747230 by guest on 22 June 2020

weighted structural scan, and Figure
Figure 3.
3B shows an example of a T2*-
Examples of 2 axial scans from the same subject. (A) The left image is an anatomical
weighted functional scan. Although (T1-weighted) image used for the visualization of anatomical landmarks and for the
the typical spatial resolution of func- differentiation of grey matter from white matter. (B) The right image is a functional
tional scans is 3 to 5 mm, newer image obtained with blood oxygen level– dependent contrast (T2*-weighted image).
techniques with stronger magnets This image is used to produce the activation map that is overlaid on the anatomical
image.
are enabling experimenters to view
structures in the submillimeter range
by using BOLD contrast.30,31

Considerations in Study
Design
Given the dependence of BOLD con-
trast on blood flow, it is intuitive that
there would be a delay between neu-
ral activity and the signal that can be
detected (Fig. 4).32 This delay, which
is referred to as a hemodynamic de-
lay, limits the temporal resolution of
this technique and has conse-
quences for study design. Because of
the approximate 3- to 5-second he-
modynamic delay, fMRI is most use-
ful for measuring the cortical activity
associated with tasks that require
several seconds to complete. More
complex study designs can be used Figure 4.
to correlate cortical activity with A single-trial functional magnetic resonance imaging time course (not averaged) dem-
movement parameters or to allow onstrating hemodynamic delay in the contralateral primary motor cortex (M1), bilateral
for fMRI investigations of tasks with premotor cortex (PM), and bilateral supplementary motor area (SMA) for a single
subject performing a motor task. The presentation of the task starts at time 0 and lasts
shorter durations, but these proce- for 2.1 seconds. The “go” signal for task execution is displayed at 9.1 seconds. The scale
dures are used infrequently in reha- of the y-axis indicates the fractional change relative to the baseline for each individual
bilitation research and thus are not time course shown. The electromyography (EMG) recording shows the absence of
discussed here. motion during the preparation period. Of interest is the increased signal intensity
immediately after the presentation of the task, indicative of motor planning, and again
after task execution (approximately 5 seconds after the start of EMG activity). (Re-
Block Design printed with permission of Lippincott Williams & Wilkins from Richter W, Andersen P,
Most fMRI studies have a block or Georgopoulos AP, Kim S-G. Sequential activity in human motor areas during a delayed
“boxcar” design, in which periods of cued finger movement task studied by time-resolved fMRI. NeuroReport. 1997;8:1257–
a control (or resting) state are inter- 1261.)
leaved with periods of task perfor-
mance, sensory stimulation, or both.
The resultant images are generated
by examining the statistical differ-

676 f Physical Therapy Volume 87 Number 6 June 2007

 Understanding Neuroimaging

Typically, the fMRI task or tasks are

repeated several times. Repeated tri-
als increase the statistical power by
increasing the number of compari-
sons of task and control conditions.
It is common to include control pe-
riods between task periods to allow
blood flow to return to the resting
state prior to the next task.

Experiments involving fMRI can be

Downloaded from https://academic.oup.com/ptj/article-abstract/87/6/670/2747230 by guest on 22 June 2020

conducted with virtually any MRI
scanner, as manufacturers offer this
as an option with their equipment.
Figure 5.
Schematic of functional magnetic resonance imaging setup. The subject lies inside the
Most hospitals have a 1.5- or 3-T
bore of the magnet. A mirror is positioned above the head coil to allow viewing of a scanner, as 1.5 or 3 T is the field
computer projector screen that displays the task. An electrogoniometer, joystick, button strength used for most fMRI studies,
press pad, or other device is attached to the joint of interest to monitor subject particularly for clinical populations.
responses. High-field-strength MRI (4 T or
greater) typically is performed at
MRI research centers and can be
ence in BOLD signals between the which to localize the activation that done with field strengths of 4 T, 7 T,
control and task periods.23 In other is acquired by fMRI. For the struc- and higher. A higher field strength,
words, the control period serves as a tural procedure, a subject lies down although associated with greater
baseline activation level during and is moved into the bore of the technical challenges, provides a
which all of the extraneous stimula- MRI scanner. No functional experi- closer spatial match between neural
tion (eg, noise of scanner, arousal ment is performed during this step activity and the resultant record of
state, and general sensory input) is because it is for anatomical purposes activation and thus better spatial
accounted for, so that only the cor- only. Following this step, the func- resolution.30
tical activity that is different from the tional part of fMRI is done. During
activity at baseline is considered to this part of the study, the subject Preprocessing of Data
be part of the task. This method is remains in the bore of the magnet The raw fMRI data need to be pre-
known as the “subtractive meth- but performs a task as images are pared before statistical comparisons
od.”33 Although this method is used acquired. Often, instructions for task can be performed. An important step
often, it is important to realize the performance are presented on a in this process is to account for the
assumptions that are being made small projector screen visible to the fact that each slice in an fMRI vol-
with this design. The subtractive subject with a mirror positioned ume (collection of slices) was ac-
method assumes that the task phase above the eyes. A computer located quired at a slightly different time.
adds cognitive processes to the rest- outside the scanning room is dedi- Thus, slice-timing corrections are
ing phase secondary to the addition cated to presenting the experimental performed on the data so that it ap-
of a new behavior. In actuality, how- paradigm and recording the re- pears that all voxels (defined below)
ever, the neural processing between sponses of the subject. The task con- within one volume were acquired at
the control phase and the task phase ditions that can be used are limited exactly the same time.35 Additional
may not be additive; rather, some by the space considerations of the preprocessing removes various arti-
cognitive processing may be re- magnet bore and the noise of the facts in the data, such as rifts in the
placed during the task phase.33 scanner but are nevertheless numer- data or movement artifacts caused by
ous. They include sensory or visual subject motion. Numerous brain im-
Procedure stimulation, cognitive tasks, and mo- aging software packages are avail-
An fMRI study protocol involves tor performance. Figure 5 shows an able to perform the preprocessing
both BOLD functional sequences example of a setup in which an MRI- and statistical analysis steps in the
and structural MRI. As mentioned compatible electrogoniometer was experiment.
above, the structural scan is neces- used to record finger motion while
sary to visualize the brain and ulti- subjects tracked a sine wave.34
mately to serve as a template upon

June 2007 Volume 87 Number 6 Physical Therapy f 677

Understanding Neuroimaging

Quantification of Activation and what is not, and often this numbers of activated voxels in the
The basis for detecting brain activa- threshold is not subjected to correc- areas of interest are summed, and the
tion lies in the statistical comparison tions for multiple comparisons; not percentage of the total activation is
of the BOLD signals of the task subjecting the threshold to such cor- determined for each area of interest.
blocks and the control blocks for rections could result in false-positive For example, if the primary sensory
each voxel within the magnetic res- findings.36 Although most investiga- cortex has 250 voxels, the primary
onance image and therefore the tors set the threshold a priori, it is motor cortex has 100 voxels, the
brain. A voxel is the unit volume of a still an arbitrary number, meaning supplementary motor area has 30
magnetic resonance image and de- that the use of a different threshold voxels, and the premotor cortex has
fines its spatial resolution. Each mag- will produce different results. A sec- 120 voxels, then the total activation
netic resonance image can be ond problem with a volume analysis is 500 voxels. The percent contribu-

Downloaded from https://academic.oup.com/ptj/article-abstract/87/6/670/2747230 by guest on 22 June 2020

thought of as a checkerboard with is that a given anatomical area of in- tion of the primary motor cortex
rows and columns of boxes with a terest (also called a “region of inter- therefore is 20% of the total activa-
depth or slice thickness. That is, the est” [ROI]) may not be equivalent tion. Preliminary work in our labora-
dimensions of the box define the size between subjects. For example, 2 tory has suggested that this index
of the voxel. The activation obtained subjects both may have 200 acti- produces more reproducible results
in an experiment can be quantified vated voxels in the primary motor across scanning sessions (unpub-
in terms of either spatial extent (vol- area, but given variations in brain lished data, 2007).
ume), which is expressed in terms of size, that quantity may represent dif-
the number of voxels, or the inten- ferent percentages of the total area Intensity Method
sity of activation within the voxels. available. Volume measures of voxel An alternative to expressing activa-
activation also have been shown to tion in terms of the number of acti-
There are several methods for deter- have lower levels of reproducibility vated voxels (volume method) is to
mining and quantifying cortical acti- than other methods of reporting.37 examine the intensity of an activated
vation. Multiple t tests or F tests cortical region. This examination
based on the assumptions of the gen- To address problems of variability, can be done for an entire ROI, re-
eral linear model generally are used, indexes of activation can be calcu- gardless of the areas of activation
requiring a voxel-by-voxel compari- lated to normalize the activation within it (thus eliminating the
son throughout the brain (for a more within a subject. A commonly used threshold problem), or just for those
detailed discussion, see Smith35). A one is the laterality index.38 This cal- voxels that were deemed active be-
cross-correlation analysis also can be culation is used to describe to what cause they exceeded the given statis-
used; this analysis involves compari- degree the contralateral or ipsilateral tical threshold. The intensity method
sons of a time course of the signal cortical areas are activated relative to allows the examiner to determine
strength of each voxel with an ideal- the stimulus or movement, as fol- the degree of intensity change dur-
ized function of the task and control lows: LI⫽(C⫺I)/(C⫹I). In this equa- ing the task. That is, in a comparison
periods. For most methods, a statis- tion, LI is the laterality index, C is the of 2 groups of subjects (such as a
tical threshold of activation is used as number of voxels in the contralateral control group and a treatment
a cutoff for what is considered ac- hemisphere, and I is the number of group), there may not be a differ-
tive. For example, P⫽.05 may be voxels in the ipsilateral hemisphere. ence in the volume of activation
used as the threshold, all voxels be- This equation produces a range of (number of voxels or laterality of ac-
low this predefined statistical thresh- values from ⫺1 to ⫹1, with a nega- tivation), but the intensity of the
old will be considered active, and tive number indicating primarily ip- BOLD signal may be stronger in 1
those above this threshold will not. silateral activation and a positive group. Intensity measures have been
number indicating primarily con- found to be less variable than volume
Volume Method tralateral activation. This index is measures.37,42– 44
A volume analysis is based on a vol- particularly useful for describing the
ume measure of activation; that is, control of movement in a subject The intensity of a response is derived
the numbers of voxels activated dur- with a brain lesion.34,39 – 41 from the BOLD signal time course,
ing the experimental task are which provides the raw average
summed and then localized to a Another method, referred to as the level of baseline activation during a
given region. Volume analysis is sub- “percent contribution method,” control phase and the ␤ weight (or
ject to several difficulties. A primary takes into account the general level change in signal intensity) during a
problem is that a threshold is used to of excitability or brain activation in a task phase. The ␤ weight and the
determine what is considered active subject. With this method, the total level of baseline activation are used

678 f Physical Therapy Volume 87 Number 6 June 2007

 Understanding Neuroimaging

to determine the percent change in cated. Comparisons of groups or metal fragments (such as those en-
signal intensity for each task period. tasks can be made by performing sta- countered by metal workers), and
The raw numbers can vary greatly tistical comparisons of the anterior surgical implants or clips. More un-
between people or testing ses- and posterior, the right and left, and usual sources of concern are medica-
sions,37 making direct comparisons the superior and inferior coordinates tion patches and tattoos. These is-
difficult. Normalizing the task data to to determine whether the location of sues require further investigation
the resting data allows for inter- activation has shifted within a partic- and vary on the basis of the year in
subject and intrasubject compari- ular region. For example, a treatment which they were acquired and the
sons.45,46 The percent change value technique may result in a shift of materials used. Generally, if a sub-
then can be used in standard statisti- activation in the anterior direction. ject has had any prior surgical pro-
cal comparisons. cedure, it should be investigated,

Downloaded from https://academic.oup.com/ptj/article-abstract/87/6/670/2747230 by guest on 22 June 2020

Although the normalization step is and questions regarding the nature
This result of examining the intensity necessary, it has obvious disadvan- of the surgery should be directed to
of activation in an ROI can be ex- tages. When a subject’s brain is a physician familiar with MRI exper-
pressed as an intensity index, which warped into an idealized brain, spa- iments.
is defined as follows: [(task intensity tial specificity is reduced. This prob-
⫺resting intensity)/resting inten- lem is of particular concern in the Given the enclosed nature
sity]⫻100. In this equation, task in- examination of a subject with a brain of the MRI scanner, subjects should
tensity is the average raw signal in- lesion, as the nonlinear transforma- be screened for claustrophobia, as
tensity during the task periods and tion associated with the “warp” will they may become anxious. Finally,
resting intensity is the average raw be particularly sensitive to drastic im- consideration should be given to
signal intensity during the resting pe- age density changes (such as those the medical history of the subject.
riods.44 This method is useful for de- associated with a brain lesion).48 In Many medications have effects on
termining whether, within an area of addition, brain atrophy secondary to the brain and its vessels, and these
interest, there was a change in the injury or age may distort normal anat- effects may influence brain activity,
intensity of the BOLD signal be- omy, making comparisons with a task performance, or both. These
tween a control phase and a task normal coordinate system more dif- medications include antidepressants,
phase. ficult. There is currently no widely anxiolytics, anticonvulsants, anti-
embraced alternative method; how- psychotics, antihypertensives, anti-
All indexes have a limitation of spec- ever, given the variability in human cholinergics, and the treatments
ificity. That is, each time a mathemat- brain size and shape, some sort of used in Parkinson disease and Alzhei-
ical manipulation of the data is made, processing always will be required to mer disease. Although these medica-
subtleties are lost. This limitation is allow for comparisons or averaging tions are not necessarily exclusion
an issue that researchers need to de- of data across subjects. criteria, their effects are an issue that
termine a priori on the basis of their each investigator should consider
scientific question and the study Potential Pitfalls of fMRI carefully. Also to be considered is
design. Subject Selection whether the disease history varies
Careful selection of volunteers is the within the study group and includes
Localization of Activation first issue to consider in the design disorders that may affect the hemo-
Once activation has been quantified, of an fMRI experiment. Because of dynamic response, such as vascular
the specific location of the activation the high magnetic field, most fer- disease.
must be determined. Localization rous metals are not allowed in the
can be done on the basis of an indi- scanner, because of the potential for Extraneous Movement
vidual’s anatomy, but it is most com- movement or heating of implanted Head movement. As with any
monly performed with Talairach co- devices. Items that are contraindi- highly technical scientific technique,
ordinates because of the need for cated include aneurysm clips, pace- several issues require attention when
comparisons and averaging of data makers, and cochlear implants. experiments are being conducted.
across subjects.47 During the MRI Many other items may be contraindi- Possibly the most vexing of these in
preprocessing step, a subject’s brain cated but, under certain circum- fMRI is excessive head motion. Un-
is “warped” into Talairach space. stances, may be allowed. Consider- fortunately, it is quite common for a
The coordinates then can be used to ation should be given to a broad subject to move his or her head
determine where in the brain (eg, range of metals, including jewelry, while performing a task in the MRI
Brodmann areas 1–3 or cingulate gy- body piercing, dental metal, frag- scanner. This motion may involve a
rus) the center of activation is lo- ments from injury, occupational startle reaction to stimulus onset or

June 2007 Volume 87 Number 6 Physical Therapy f 679

Understanding Neuroimaging

the beginning of scanner noise. It bar” to guarantee stabilization, but lar control. Ideally, electromyogra-
may involve a tracking motion of the this may not be well tolerated by phy (EMG) would be used to moni-
head that is associated with the subjects and may lead to difficulty tor for muscular activity in both the
task,49 or it may simply be attribut- with recruitment. Shorter experi- performing limb and the opposite
able to the inability of the subject to mental times or mock scanner train- limb. The use of EMG in the mag-
remain still for an extended period of ing sessions also may reduce the pro- netic environment is hampered by
time. This problem is of particular pensity for movement.51 Algorithms difficulties associated with interfer-
concern when patients, older peo- also are used during image acquisi- ence between the MRI scanner and
ple, or children are being studied, as tion to perform real-time motion cor- the EMG amplifier. Some researchers
the ability of these populations to rection to help minimize this address this interference by using
inhibit motion may be reduced. problem.52 EMG recording during a mock scan-

Downloaded from https://academic.oup.com/ptj/article-abstract/87/6/670/2747230 by guest on 22 June 2020

ner performance of the task55 or by
The problem with unwanted head Eye movement. In addition to introducing pauses in the scanning
motion is two-fold. Excessive motion movement of the head, unwanted protocol to allow for brief periods of
results in a false activation pattern eye movement is another concern. EMG recording.56 Other methods for
because the fMRI signal changes For example, when cortical activity observing mirror movement involve
caused by the hemodynamic re- as a result of motor task performance monitoring gross motion (such as
sponse are smaller than the apparent is being investigated, activation may with an electrogoniometer or a joy-
signal differences that result from be attributable to the eye movement stick)34,57 or visual observation, but
head movement. Additionally, move- associated with the task and not to these methods are limited also, as
ment adds to the variance of the sig- the target task. Instructing the sub- there could be extraneous EMG ac-
nal, which in turn reduces the sensi- ject to maintain eye fixation or mon- tivity of insufficient magnitude to
tivity of the statistical tests to detect itoring eye motion with an eye- produce observable motion.
true activation, resulting in a possi- tracking device can assist in
ble type II error.50 Typically, brain dissociating the potential confound- Emotional Status
activation related to motion artifacts ing effects of unwanted eye mo- BOLD signal activation has been
is most evident on the brain surface tion.53 There are several difficulties shown to be affected by emotional
(rim activation) and at the interface with the use of eye tracking in an status, such as depression58,59 or anx-
between cerebrospinal fluid and MRI scanner, but eye tracking is the iety.60,61 As such, a comparison of
brain tissue (such as that seen most quantifiable method for mea- groups of people who may differ in
around the ventricles). suring subject eye motion and these affective domains requires con-
should be considered if eye motion sideration. Additionally, within the
Because of the pervasiveness of this is a likely confounding variable. same person, changes in affect from
issue, fMRI software packages per- one testing session to another may
form motion correction during the Muscular activation. The muscu- be reflected in changes in the mag-
preprocessing phase of data analysis. lar activation strategy (pattern of nitude of the BOLD signal. To in-
It is important to examine all data for muscle use) and mirror movement crease the probability that the
motion and to set an a priori thresh- (movement of the opposite hand [or changes seen are attributable to task
old for data that should be excluded. foot]) during the performance of a performance, emotional status needs
For example, movement greater than task are other issues that require con- to be documented prior to image
or equal to the voxel resolution is sideration when fMRI experiments acquisition. Although changes in
discarded. That is, if the voxel reso- are being conducted or when fMRI emotional status (such as anxiety or
lution is typically 3 mm3, any trial data are being interpreted. Mirror depression) would not nullify neuro-
with movement of 3 mm or greater is movement is especially important imaging findings per se, such
discarded. In our experience, data when patients are being studied, as changes should be taken into ac-
for up to 20% of subjects may need inhibition of motion may be a prob- count during post hoc analysis. This
to be excluded for this reason. lem in this population.54 The muscu- objective could be achieved by cate-
lar activation strategy during task gorizing subjects on the basis of
Several methods can be used to re- performance is applicable to all stud- emotional status and searching for
duce head movement. Head cushion- ies that require movement. For ex- differences between groups, per-
ing (such as stabilization with cush- ample, cortical activation pattern haps by using the Beck Depression
ions or straps) is a safe deterrent that changes after a treatment could be Inventory62 or by adding emotional
typically is well tolerated by all sub- attributable to either a change in status as a covariable.
jects. Other methods include a “bite neural control or a change in muscu-

680 f Physical Therapy Volume 87 Number 6 June 2007

 Understanding Neuroimaging

Task Load subjects. A small number of subjects, has been identified. This conclusion
There is growing evidence that task the inherent variability of fMRI, and is not entirely correct. In the inter-
load, or the degree of task demand the variability associated with pa- pretation of neuroimaging data, it is
during an experiment, may influence tient populations (such as stroke lo- important to remember that activa-
the magnitude or extent of brain ac- cation and size and atrophy because tion in a given brain region means
tivation.63– 65 In a comparison of of age) can make finding significant that that region is associated with
BOLD signal data from a healthy pop- results difficult. Is a negative result the performance of that task. The
ulation with data from a patient pop- truly representative of no difference, observed activation alone likely is in-
ulation, task load can be of particular or is it attributable to a type II error sufficient to produce movement, and
concern, because a given task may that could be solved with more it cannot be determined whether the
be less demanding for people who power? This problem is not unique observed activation is attributable to

Downloaded from https://academic.oup.com/ptj/article-abstract/87/6/670/2747230 by guest on 22 June 2020

are healthy. Indeed, even imagina- to fMRI but requires caution in the the activation of excitatory or inhib-
tion of a movement can produce sig- interpretation of experiments with itory neurons.70 It is easier to keep
nificant levels of activation.57,66 – 68 negative results.52 this cautious interpretation in per-
spective when one remembers the
It is impractical to control com- Overcoming Artifacts multiple constraints associated with
pletely for task load a priori, but Although it is true that several issues neuroimaging, such as the indirect
pilot data on task performance alone can lead to artifacts (errors) in sig- measurement of neural activity,
(without neuroimaging) would assist nals, most of them can be addressed “warping” of the brain into Talairach
in determining whether perfor- with careful planning of an experi- space, the use of a threshold for ac-
mance varies between groups. This mental design. Specifically, for ex- tivation, and the averaging of signals
situation also can be addressed by amining a treatment approach in a over a period of many seconds. Ad-
using measures of accuracy or per- pretest-posttest design, it is imper- ditionally, some of the effects of er-
formance time for post hoc covari- ative to use a randomized control ror may be even more pronounced
ables in the analysis. Another group. This group should be in the interpretation of fMRI investi-
method of equalizing task load is to matched to the treatment group gations of patient populations. In-
“titrate” the task, that is, modify task with regard to age, sex, and handed- deed, more work needs to be done
difficulty so that all participants ness. Comparisons of the 2 groups to optimize this technique for pa-
maintain the same level of perfor- then can be made with confidence, tient populations.
mance (eg, a 75% accuracy of re- because the error associated with
sponses).69 Previous work34,44,57 has task performance can be assumed to The silver lining in the cloud out-
minimized the problem of task load be equally distributed across the shines the grey, however. Neuro-
through the use of age-matched con- groups. Likewise, for examining imaging already has yielded numer-
trol subjects in studies of subjects changes within a subject (versus ous noteworthy works in the field of
with stroke, thereby removing the group data), a multiple-baseline de- rehabilitation, as highlighted in this
effect of age on performance. In in- sign may be considered to account Special Series. With proper study de-
vestigations of the cortical effects as- for the variability that will occur sign and widespread critical inter-
sociated with a treatment or tech- from one testing session to the next, pretation, neuroimaging will be-
nique, a control group allows a independent of any treatment effect. come an increasingly powerful tool
comparison of groups. A significant Functional MRI always will have er- that will help to shape the under-
difference in a controlled design ror, but it is not always convenient standing of mechanisms of recovery
lends confidence that the results are or even possible to measure all po- and influence physical therapy inter-
not attributable to variability of the tential sources of error. Careful plan- vention of the future.
testing measure but rather represent ning of an experimental design can
a true difference between groups. lead to greater confidence that any
Both authors provided concept/idea/project
effect seen is attributable to the ex- design and writing.
Power perimental conditions and not to
This article was received June 1, 2006, and was
Functional MRI is an expensive re- ubiquitous artifacts. accepted January 10, 2007.
search tool. The high costs are justi-
fied by the technical maintenance Interpreting Data DOI: 10.2522/ptj.20060149
and highly trained personnel re- It is tempting to conclude from “lit
quired to keep the device operating. up” areas of activation in the cortex
These high costs often result in an- that the location of activation re-
other problem: a small number of sponsible for the observed behavior

June 2007 Volume 87 Number 6 Physical Therapy f 681

Understanding Neuroimaging

References 16 Khedr EM, Ahmed MA, Fathy N, Rothwell 33 Sartori G, Umilta C. How to avoid the fal-
JC. Therapeutic trial of repetitive transcra- lacies of cognitive subtraction in brain im-
1 Goyder DG. My Battle for Life: The Auto- nial magnetic stimulation after acute isch- aging. Brain Lang. 2000;74:191–212.
biography of a Phrenologist. London, emic stroke. Neurology. 2005;65:
United Kingdom; Simpkin, Marshall, and 34 Carey JR, Kimberley TJ, Lewis SM, et al.
353–354.
Co; 1857. Analysis of fMRI and finger tracking train-
17 Siebner H, Mentschel C, Auer C, Conrad B. ing in subjects with chronic stroke. Brain.
2 Berker EA, Berker AH, Smith A. Transla- Repetitive transcranial magnetic stimula- 2002;125:773–788.
tion of Broca’s 1865 report: localization of tion has a beneficial effect on bradykinesia
speech in the third left frontal convolu- 35 Smith SM. Overview of fMRI analysis. Br J
in Parkinson’s disease. Neuroreport.
tion. Arch Neurol. 1986;43:1065–1072. Radiol. 2004;77:S167–S175.
1999;10:589 –594.
3 Fritsch G, Hitzig E. Über die elektrische 36 Genovese CR, Lazar NA, Nichols T.
18 Siebner H, Tormos JM, Ceballos-Baumann
Erregbarkeit des Grosshirns. Arch Anat Thresholding of statistical maps in func-
AO, et al. Low-frequency repetitive trans-
Physiol Wissenschaftl. 1870;37:300 –332. tional neuroimaging using the false discov-
cranial magnetic stimulation of the motor ery rate. Neuroimage. 2002;15:870 – 878.
4 Walsh V, Cowey A. Transcranial magnetic cortex in writer’s cramp. Neurology.
stimulation and cognitive neuroscience. 1999;52:529 –537. 37 Cohen MS, DuBois RM. Stability, repeat-

Downloaded from https://academic.oup.com/ptj/article-abstract/87/6/670/2747230 by guest on 22 June 2020

Nat Rev. 2000;1:73–79. ability, and the expression of signal mag-
19 Talelli P, Greenwood RJ, Rothwell JC. Arm nitude in functional magnetic resonance
5 Hämäläinen M, Hari R, IImoniemi RJ, et al. function after stroke: neurophysiological imaging. J Magn Reson Imaging. 1999;10:
Magnetoencephalography: theory, instru- correlates and recovery mechanisms as- 33– 40.
mentation, and applications to noninva- sessed by transcranial magnetic stimula-
sive studies of the working brain. Reviews tion. Clin Neurophysiol. 2006;117:1641– 38 Cramer S, Nelles G, Benson R, et al. A func-
of Modern Physics. 1993;65:413– 497. 1659. tional MRI study of subjects recovered
from hemiparetic stroke. Stroke. 1997;28:
6 Lounasmaa OV, Hämäläinen M, Hari R, 20 Buxton RB. Introduction to Functional 2518 –2527.
Salmelin R. Information processing in the Magnetic Resonance Imaging. Cam-
human brain: magnetoencephalographic bridge, United Kingdom: Cambridge Uni- 39 Cramer SC, Weisskoff RM, Schaechter JD,
approach. Proc Natl Acad Sci U S A. versity Press; 2001. et al. Motor cortex activation is related to
1996;93:8809 – 8815. force of squeezing. Hum Brain Mapp.
21 Huettel SA, Song AW, McCarthy G. Func- 2002;16:197–205.
7 Hari R. Electroencephalography: basic tional Magnetic Resonance Imaging.
principles, clinical applications, and re- Sunderland, Mass: Sinauer Associates; 40 Tamada T, Miyauchi S, Imamizu H, et al.
lated fields. In: Niedermeyer E, Lopes da 2004. Cerebro-cerebellar functional connectivity
Silva F, eds. Electroencephalography: revealed by the laterality index in tool-use
22 Kandel E, Schwartz J, Jessell T. Principles
Basic Principles, Clinical Applications, learning. Neuroreport. 1999;10:325–331.
of Neural Science. 4th ed. New York, NY:
and Related Fields. Baltimore, Md: Wil- McGraw-Hill; 2000. 41 Park SW, Butler AJ, Cavalheiro V, et al.
liams & Wilkins; 1993:1035–1061. Changes in serial optical topography and
23 Ugurbil K, Ogawa S, Kim SG, et al. Imag-
8 Forss N, Hietanen M, Salonen O, Hari R. TMS during task performance after
ing brain activity using nuclear spins. In:
Modified activation of somatosensory cor- constraint-induced movement therapy in
Maraviglia B, ed. Magnetic Resonance
tical network in patients with right- stroke: a case study. Neurorehabil Neural
and Brain Function: Approaches From
hemisphere stroke. Brain. 1999;122: Repair. 2004;18:95–105.
Physics. Amsterdam, the Netherlands: IOS
1889 –1899. Press; 1999:261–310. 42 Waldvogel D, van Gelderen P, Immisch I,
9 Meinzer M, Elbert T, Wienbruch C, et al. et al. The variability of serial fMRI data:
24 Friedland RP, Iadecola C. On the regula-
Intensive language training enhances correlation between a visual and a motor
tion of the blood-supply of the brain. Neu-
brain plasticity in chronic aphasia. BMC task. Neuroreport. 2000;11:3843–3847.
rology. 1991;41:10 –14.
Biol. 2004;25:20. 43 Carey JR, Anderson KM, Kimberley TJ,
25 DeYoe E, Bandettini P, Neitz J, et al. Func-
10 Cherry SR, Phelps ME. Imaging brain func- et al. fMRI analysis of ankle movement
tional magnetic resonance imaging of the
tion with positron emission tomography. tracking training in subject with stroke.
human brain. J Neurosci Methods.
In: Toga AW, Mazziotta JC, eds. Brain Exp Brain Res. 2004;154:281–290.
1994;54:171–187.
Mapping: The Methods. 2nd ed. San Di- 44 Kimberley TJ, Lewis SM, Auerbach EJ,
ego, Calif: Academic Press; 2002:485–511. 26 Ogawa S, Lee TM, Barrere B. The sensitiv- et al. Electrical stimulation driving func-
ity of magnetic resonance image signals of
11 Pascual-Leone A, Walsh V, Rothwell J. tional improvements and cortical changes
a rat brain to changes in the cerebral ve-
Transcranial magnetic stimulation in cog- in subjects with stroke. Exp Brain Res.
nous blood oxygenation. Magn Reson
nitive neuroscience: virtual lesion, chro- 2004;154:450 – 460.
Med. 1993;29:205–210.
nometry, and functional connectivity. 45 Georgopoulos AP, Whang K, Georgopou-
Curr Opin Neurobiol. 2000;10:232–237. 27 Logothetis NK, Pauls J, Augath M, et al. los MA, et al. Functional magnetic reso-
Neurophysiological investigation of the
12 Pascual-Leone A, Walsh V. Transcranial nance imaging of visual object construc-
basis of the fMRI signal. Nature.
magnetic stimulation. In: Toga AW, Mazzi- tion and shape discrimination: relations
2001;412:150 –157.
otta JC, eds. Brain Mapping: The Methods. among task, hemispheric lateralization,
2nd ed. San Diego, Calif: Academic Press; 28 Logothetis NK. The underpinnings of the and gender. J Cogn Neurosci. 2001;13:
2002:255–290. BOLD functional magnetic resonance im- 72– 89.
aging signal. J Neurosci. 2003;23:3963–
13 Wassermann EM. Risk and safety of repet- 46 Lewis SM, Jerde TA, Tzagarakis C, et al.
3971.
itive transcranial magnetic stimulation: re- Logarithmic transformation for high-field
port and suggested guidelines. Electroen- 29 Mukamel R, Gelbard H, Arieli A, et al. Cou- BOLD fMRI data. Exp Brain Res. 2005;
cephalogr Clin Neurophysiol. 1998;108: pling between neuronal firing, field poten- 165:447– 453.
1–16. tials, and fMRI in human auditory cortex. 47 Talairach J, Tournoux P. Co-planar Stereo-
Science. 2005;309:951–954.
14 O’Reardon JP, Blumner KH, Peshek AD, taxic Atlas of the Human Brain. New
et al. Long-term maintenance therapy for 30 Ugurbil K, Toth L, Kim DS. How accurate York, NY: Thieme; 1988.
major depressive disorder with rTMS. is magnetic resonance imaging of brain 48 Rorden C, Brett M. Stereotaxic display of
J Clin Psychiatry. 2005;66:1524 –1528. function? Trends Neurosci. 2003;26:108 – brain lesions. Behav Neurol. 2000;12:191–
114.
15 Takeuchi N, Chuma T, Matsuo Y, et al. 200.
Repetitive transcranial magnetic stimula- 31 Kim DS, Duong TQ, Kim S-G. High- 49 Hajnal JV, Mayers R, Oatridge A, et al. Ar-
tion of contralesional primary motor cor- resolution mapping of iso-orientation col- tifacts due to stimulus correlated motion
tex improves hand function after stroke. umns by fMRI. Nature. 2000;3:164 –169. in functional imaging of the brain. Magn
Stroke. 2005;36:2681–2686. 32 Kim SG, Richter W, Ugurbil K. Limitations Reson Med. 1994;31:289 –291.
of temporal resolution in functional MRI.
Magn Reson Med. 1997;37:631– 636.

682 f Physical Therapy Volume 87 Number 6 June 2007

 Understanding Neuroimaging

50 Friston KJ, Holmes A, Poline J-B, et al. De- 57 Kimberley TJ, Khandekar G, Skraba LL, 65 Rypma B, Prabhakaran V, Desmond JE,
tecting activations in PET and fMRI: levels et al. Neural substrates of motor imagery et al. Load-dependent roles of frontal brain
of interference and power. Neuroimage. in severe hemiparesis. Neurorehabil Neu- regions in the maintenance of working
1995;4:223–235. ral Repair. 2006;20:268 –277. memory. Neuroimage. 1999;9:216 –226.
51 Seto E, Sela G, McIlroy WE, et al. Quanti- 58 Drevets WC. Functional anatomical abnor- 66 Nair DG, Purcott KL, Fuchs A, et al. Corti-
fying head motion associated with motor malities in limbic and prefrontal cortical cal and cerebellar activity of the human
tasks used in fMRI. Neuroimage. 2001;14: structures in major depression. Prog brain during imagined and executed uni-
284 –297. Brain Res. 2000;126:413– 431. manual and bimanual action sequences: a
functional MRI study. Brain Res Cogn
52 Desmond JE, Glover GH. Estimating sam- 59 Mayberg HS, Liotti M, Brannan SK, et al. Brain Res. 2003;15:250 –260.
ple size in functional MRI (fMRI) neuroim- Reciprocal limbic-cortical function and
aging studies: statistical power analysis. negative mood: converging PET findings 67 Luft AR, Skalej M, Stefanou A, et al. Com-
J Neurosci Methods. 2002;118:115–128. in depression and normal sadness. Am J paring motion and imagery-related activa-
Psychiatry. 1999;156:675– 682. tion in the human cerebellum: a functional
53 Niimi R, Shimamoto K, Sawaki A, et al. MRI study. Hum Brain Mapp. 1998;
Eye-tracking device comparisons of three 60 Hoehn-Saric R, Lee JS, McLeod DR, Wong 6:105–113.
methods of magnetic resonance image se- DF. Effect of worry on regional cerebral

Downloaded from https://academic.oup.com/ptj/article-abstract/87/6/670/2747230 by guest on 22 June 2020

ries displays. J Digit Imaging. 1997;10: blood flow in nonanxious subjects. Psychi- 68 Niyazov DM, Butler AJ, Kadah YM, et al.
147–151. atry Res. 2005;140:259 –269. Functional magnetic resonance imaging
and transcranial magnetic stimulation: ef-
54 Wittenberg GF, Bastian AJ, Dromerick AW, 61 Wang J, Rao H, Wetmore GS, et al. Perfu- fects of motor imagery, movement and
et al. Mirror movements complicate inter- sion functional MRI reveals cerebral blood coil orientation. Clin Neurophysiol.
pretation of cerebral activation changes flow pattern under psychological stress. 2005;116:1601–1610.
during recovery from subcortical infarc- Proc Natl Acad Sci U S A. 2005;102:17804 –
tion. Neurorehabil Neural Repair. 17809. 69 Hillary FG, Steffener J, Biswal BB, et al.
2000;14:213–221. Functional magnetic resonance imaging
62 Beck A, Mendelson M, Mock J. Inventory technology and traumatic brain injury re-
55 Ellingson ML, Liebenthal E, Spanaki MV, for measuring depression. Arch Gen Psy- habilitation: guidelines for methodological
et al. Ballistocardiogram artifact reduction chiatry. 1961;4:561–571. and conceptual pitfalls. J Head Trauma
in the simultaneous acquisition of auditory 63 Manoach DS, Gollub RL, Benson ES, et al. Rehabil. 2002;17:411– 430.
ERPS and fMRI. Neuroimage. 2004;22: Schizophrenic subjects show aberrant
1534 –1542. 70 Lauritzen M. Reading vascular changes in
fMRI activation of dorsolateral prefrontal brain imaging: is dendritic calcium the
56 Dai TH, Liu JZ, Sahgal V, et al. Relationship cortex and basal ganglia during working key? Nat Rev Neurosci. 2005;6:77– 85.
between muscle output and functional memory performance. Biol Psychiatry.
MRI-measured brain activation. Exp Brain 2000;48:99 –109.
Res. 2001;140:290 –300. 64 Rypma B, D’Esposito M. The roles of pre-
frontal brain regions in components of
working memory: effects of memory load
and individual differences. Proc Natl Acad
Sci U S A. 1999;96:6558 – 6563.

June 2007 Volume 87 Number 6 Physical Therapy f 683