You are on page 1of 11

Combining Images Across Multiple Subjects: A Study

of Direct Cortical Electrical Interference


Diana L. Miglioretti, Colin McCulloch, and Scott L. Zeger

This article introduces a Bayesian hierarchical model for combining information across multiple images. Our work was motivated by
an invasive functional brain mapping technique called direct cortical electrical interference that gives a sparse set of binary observations
of an underlying “true” region at multiple sites on the brain surface. To model region shapes that may vary widely across individuals,
we use mixtures of simple templates, for example, circles. These subject-speciŽ c templates are treated as random effects, governed by
a set of population templates that make up a population region. The numbers of subject-speciŽ c and population templates are treated
as unknown variables to be estimated from the data. Conditional on the subject-speciŽ c regions, the observed data are modeled using
logistic regression. To estimate the variability among images across patients, we develop a measure based on Baddeley’s error measure
for binary images. Because the dimension of the parameter space changes as the numbers of subject-speciŽ c and population templates
change, inference is made using reversible jump Markov chain Monte Carlo. Using a hierarchical approach, we may better estimate each
individual’s region by borrowing strength from other subjects’ data, we can estimate a population region by pooling information across
subjects, and we can use a collection of data from previous patients to predict the location of a future patient’s region of interest. The
approach is illustrated with DCE I data collected on 20 patients for two motor tasks: tongue and hand movements.
KEY WORDS: Binary image variation; Functional brain mapping; Hierarchical model; Reversible jump Markov chain Monte Carlo;
Template mixture model.

1. INTRODUCTION 1997; Lee 1997). Biostatistics now faces a new challenge to


develop hierarchical models for images and other complex
When considering the results of a screening test, a physi-
measures of health that are beginning to supplant traditionally
cian rarely relies on an individual’s test results alone. He or
used scalar measures in biomedical research.
she also considers the sensitivity and speciŽ city of the test
In this article we introduce a hierarchical approach for ana-
and prevalence of the disease among similar people. Simi-
lyzing multiple binary images that is an extension of the tem-
larly, when an experienced neurosurgeon considers resection
surgery to remove a seizure focus or brain tumor, he or she plate mixture model previously introduced for a single image
considers his or her prior knowledge obtained from previous (Miglioretti, McCulloch, and Zeger 2000). Our work was
patients along with any functional brain mapping results from motivated by a speciŽ c type of functional brain mapping data
that speciŽ c patient. Thus, when analyzing functional brain called direct cortical electrical interference (DCE I), reviewed
mapping data, it is natural to borrow strength across patients later; however, we seek a general approach that can be used
to estimate an individual’s brain map. In addition, combining for other brain mapping and spatial data. The goal of DCE I
information across patients is important for making popula- is to determine the cortical region in the brain that is nec-
tion inferences and predictions about future patients who have essary for a speciŽ c task. The underlying brain map may be
not yet undergone testing. Hierarchical models are a natural thought of as a binary image, where each location in the brain
approach for meeting these goals. is either necessary for the task or not. DCE I results in a sparse
The central idea in hierarchical modeling is to acknowl- set of binary measurements of this brain map at multiple sites
edge multiple levels of variability in data. In population and throughout the brain. The inferential goal is to estimate each
biomedical research, such modeling can account for variation individual’s image (i.e., brain map), along with a population
among individuals in their true characteristics as well as varia- image with speciŽ c interest in the locations and sizes of the
tion due to imprecise measurement. Hierarchical models, also region of interest (the area necessary for the task under study).
called random-effects, mixed-effects, or multilevel models, are
1.1 Direct Cortical Electrical Interference
well established for univariate continuous measures (e.g., Rao
1965; Lindley and Smith 1972; Laird and Ware 1982), binary DCE I, also known as cortical stimulation, is an invasive
responses (e.g., Stiratelli, Laird, and Ware 1984; Zeger and functional brain mapping method used in surgical resection
Karim 1991), and count data (e.g., Christiansen and Morris candidates. Most patients receiving DCE I have medically
intractable epilepsy. DCE I is used to determine if the parts
of the brain causing the seizures can be safely resected with-
Diana L. Miglioretti is Assistant ScientiŽ c Investigator, Center for Health out damaging regions necessary for important daily func-
Studies, Group Health Cooperative of Puget Sound, Seattle, WA 98101
(E-mail: miglioretti.d@ghc.org). Colin McCulloch is Senior Statistician, tions, such as motor control and language. For extraoperative
Applied Statistics Program, General Electric Corporate Research & Devel- DCE I, arrays of electrodes are subdurally implanted into the
opment, Schenectady, NY 12301 (E-mail: mcculloc@crd.ge.com). Scott L. brain surface and left in the patient’s brain for several days
Zeger is Professor and Chair, Department of Biostatistics, Johns Hop-
kins University, Baltimore, MD 21205 (E-mail: szeger@jhsph.edu). The to a week or more (Lesser et al. 1987; Gordon, Hart, Boat-
authors thank the associate editor for valuable comment s and Dana Boat- man, and Lesser 1996). During DCE I testing, small temporary
man, John Hart, and Barry Gordon from the Department of Neurology, functional “lesions” are created with a brief electrical current
Johns Hopkins University, for introducing us to this challenging prob-
lem and for providing the DCE I data along with helpful comments. This
work was partially supported by the National Institutes of Aging train-
ing grant T32 AG00247-01A 1 and the National Institutes of Health grant © 2002 American Statistical Association
R01-MH56639 . The program used for the example may be found at
Journal of the American Statistical Association
http://www.centerforhealthstudies.org /perpages/migliore/index.html.
March 2002, Vol. 97, No. 457, Applications and Case Studies
125
126 Journal of the American Statistical Association, March 2002

( 5 sec) passed between two electrodes about 1 cm apart and sites found to be associated with the task under study, typi-
(Gordon et al. 1996). During this time, the patient is asked cally after registering each brain to a common atlas (e.g., Boat-
to perform a task, such as name an object shown on a com- man, Lesser, Hall, and Gordon 1994; Boatman, Lesser, and
puter screen or move his or her Ž ngers. By comparing the Gordon 1995; Boatman, Hall, Goldstein, Lesser, and Gordon
patient’s performance with and without electrical interference, 1997; Gordon et al. 1996; Nii, Uematsu, Lesser, and Gordon
the clinician makes a judgement as to whether the task has 1996). In addition, the brain atlas may be divided into arbitrary
been temporarily impaired by current applied to this electrode zones and the proportion of associated sites within each zone
pair. This testing results in binary data (impaired or intact) at calculated across subjects (e.g., Malow et al. 1996; Ojemann,
multiple sites throughout the brain. Examples of DCE I data Ojemann, Lettich, and Berger 1989; Uematsu et al. 1992,
from two motor tasks are displayed later (see Figs. 4 and 5). Urasaki, Uematsu, Gordon, and Lesser 1994). In this article,
Because the electrical current during DCE I testing is dis- we introduce a new approach for analyzing DCE I data that
tributed to a region beneath and around the electrode pair, overcomes the obvious limitations of these previously used
cortical processing is affected in an area rather than just at a methods. We use a hierarchical approach that borrows strength
point location. It is important to take this into account during across subjects to estimate a brain map for each individual
analysis, because associating the observed value with a sim- and combines information across subjects to estimate a single
ple point location would result in overlocalizing the region of population brain map.
interest—namely, underestimating the areas in the brain nec- For DCE I, there is little prior information about the shape,
essary for the task. There is also evidence that the probability location, and size of the region of interest and limited spa-
of producing a deŽ cit during interference is a function of the tial resolution in the data. In addition, the region shapes may
strength of the current. Under this assumption, DCE I would vary widely across individuals. Therefore, we require a  exi-
be less likely to cause a task impairment if the region of inter- ble method that can model a wide range of region shapes using
est were located farther away from the electrode site where a small number of parameters that may be pooled across indi-
the current strength is weaker. viduals. To meet these requirements, we model the individual
Nathan, Sinha, Gordon, Lesser, and Thakor (1993) studied and population regions of interest using mixtures of subject-
the distribution of the electrical current using Ž nite element speciŽ c and population templates of simple geometrical shapes
analysis. They found that the current density peaks immedi- (e.g., circles). We treat the subject-speciŽ c templates as ran-
ately beneath the electrodes and declines rapidly to almost dom effects, governed by the population templates. We treat
no current approximately .5 cm away from the electrode pair. the numbers of population and subject-speciŽ c templates as
Just below the cortical surface, the current midway between unknown variables to be estimated from the data. Because the
the two electrodes is about half the strength of the peak cur- dimension of the parameter space depends on the numbers of
rent. We closely approximate the electrical current distribution templates, we make inference using reversible jump Markov
given by Nathan et al. (1993) using a mixture of three bivari- chain Monte Carlo (RJMCMC). Given the data, we summa-
ate Gaussian densities (Fig. 1). Two of the Gaussian densities rize our region analysis by producing, for each individual and
are circular and centered at each electrode in the site. The third the population, a pixelwise map of the posterior probability
density is elliptical and centered at the midpoint between the that each location in the image is contained in the region of
two electrodes with orientation equal to that of the electrode interest. These probabilities are implied by the posterior dis-
pair. tribution on the template parameters, which is marginalized
Methods for analyzing DCE I data in multiple subjects have over the numbers of templates. Under this marginalization, the
been mostly limited to displaying the locations of tested sites region estimates are not restricted to be a union of circles, but
instead can take on a wide range of shapes. We formulate the
model for the observed data using a generalized linear model
framework, so the method can be applied to varied types of
measurements.
In Section 2 we develop the template mixture model for
DCE I with multiple subjects. In Section 3 we describe the
algorithm used for model Ž tting and inference. We introduce
a measure of image variability in Section 4. In Section 5 we
present an example using data from DCE I studies conducted
at the Johns Hopkins Medical Institutions. We end with a dis-
cussion of our methodology.

2. THE TEMPLATE MIXTURE MODEL


The template mixture model for multiple subjects consists
of three main parts: (1) a set of prior assumptions about the
true image (brain map) for the population and (2) for each
individual, and (3) a speciŽ cation of the probability distribu-
Figure 1. Approximation of the Current Density Beneath the Elec- tion of the observations given the true images. In the follow-
trode Site. The two electrodes are located at ( x, y ) D ( .5, 1) and ing model formulation, we indicate subject-speciŽ c parameters
( 1.5, 1) . with subscript i and population parameters with superscript ü .
Miglioretti, McCulloch, and Zeger: Direct Cortical Electrical Interference 127

We let u denote a two-dimensional location on the surface tion template has six associated parameters: a two-dimensional
of the brain. The true image for the ith individual fi 4u5, location Ìük , a radius ük , and three precisions, Ž1k , Ž2k , and
i D 11 : : : 1 N , and the population image f ü 4u5 are assumed to Ž3k , used to deŽ ne individual distributions in the next level of
be binary such that the hierarchy. Here Ž1k and Ž2k correspond to the two location
( parameters Œü1k and Œü2k , and Ž3k corresponds to the template
1 if u 2 Si radii ük . The population template locations Ìü1 1 : : : 1 ÌüK ü are
fi 4u5 D
0 otherwise independently distributed uniformly over the image space A,
and the radii ü1 1 : : : 1 üK ü are independently distributed uni-
and
( formly over an a priori known reasonable range of values
ü 1 if u 2 S ü 4min 1 max 5, deŽ ned in Section 5. The distributions for the
f 4u5 D
0 otherwise1 population precisions 8Žp1 1 : : : 1 ŽpK ü 3 p D 11 21 39 are assumed
to be independent gammas. We do not constrain the precisions
where Si and S ü are the subject-speciŽ c and population regions for the two location parameters to be equal, facilitating extra
of interest. Thus each location on the cortex is assumed to be  exibility in the template parameter variability.
either needed for the task under study or not, and the regions Given the population parameters, the model for the subject-
of interest are the areas needed for the task. speciŽ c regions is deŽ ned as follows. We model the number of
We model the subject-speciŽ c regions of interest Si as mix- individual templates K1 1 : : : 1 KN as independent Poisson4‹5
tures of Ki simple geometrical shapes, or templates, Tik 2 Si D variables truncated to Ki µ K ü , where the mean number of
SKi
kD1 Tik . We use circular templates Tik D 8u2 —u ƒ Ìik — < ik 9, templates ‹ follows a gamma distribution. We make the con-
where Ìik is the two-dimensional location parameter indicat- straint that each individual can have at most one template
ing the center of the circle, ik is its radius, and —¢— repre- associated with each of the population templates. For the ith
sents Euclidean distance. The numbers of templates for each individual, the labels indicating to which population template
individual K1 1 : : : 1 KN are taken to be unknown and possi- each individual template corresponds are considered latent
bly different. Each template Tik is assumed to come from allocation variables zi1 1 : : : 1 ziKi , which are distributed uni-
a distribution centered near one of K ü population templates, formly on the permutations of Ki elements chosen from K ü
which are also taken to be circular with parameters Ìük and elements. The individual template two-dimensional locations
ük , k D 11 : : : 1 K ü . These population templates make up the Ì i1 1 : : : 1 ÌiKi and radii i1 1 : : : 1 iKi are independently dis-
S ü
population region of interest, S ü D KkD1 Tkü . tributed as Gaussians with population means 4Ìü 1 Ðü 5 and
The full hierarchical model is given in Figure 2. The num- precisions 4Î5 determined by the allocation vector zi and
ber of population templates K ü is assumed to follow a Pois- constrained to be in the same range as the population param-
son distribution with Ž xed mean parameter ‹ü . Each popula- eter space.

Figure 2. Full Hierarchical Model.


128 Journal of the American Statistical Association, March 2002

The third part of the model describes the observed data, 3. MODEL FITTING
which we assume follow a generalized linear model. Let yij
If we assume that the numbers of individual and population
represent the binary observation at the jth testing site of the
templates are unknown, then the dimension of the parameter
ith individual, j D 11 : : : 1 Ji 1 i D 11 : : : 1 N . Thus yij equals
space changes as the number of templates changes. One  ex-
1 if the task is impaired during DCE I and 0 if the task is
ible technique for estimating parameters in Bayesian models
functional. Let xij D 4xij 1 1 xij2 5 be the pair of two-dimensional
of varying dimension is the RJMCMC methodology proposed
coordinates for the two electrodes in the jth testing site for
by Green (1995). One can explore the combined model space
individual i. Given the subject-speciŽ c parameters, we model
by setting up a hybrid sampler (Tierney 1994) that proposes
the observed values using logistic regression,
to update the parameters within a Ž xed-dimensional subspace
in one type of move and proposes to jump between models of
yij —fi Bernoulli4 ij 51 different dimensions in another type of move. In this way, we
and use RJMCMC to simulate from the joint posterior distribu-
tion deŽ ned in (3). An estimate of each subject-speciŽ c region
logit4 ij 5 D  0 C  1 H 4fi 1 xij 51 (1)
of interest Si may be calculated using the posterior mean of
R the ith individual’s brain map, marginalizing over the num-
where H 4fi 1 xij 5 D u2²2 I4—u ƒ Ìik — < ik for any k D
ber of templates Ki and population parameters, p4u 2 Si —y5 ²
11 : : : 1 Ki 5Dxij 4u5 du. Here I 4¢5 is the indicator function and
E4fi 4u5—y5. An estimate of the population region of interest
Dxij 4u5 is the electrical current distribution centered around
S ü may be calculated similarly as p4u 2 S ü —y5 ² E4f ü 4u5—y5.
xij1 and xij 2 , assumed known. Thus H 4fi 1 xij 5 is a measure of
We estimate E4fi 4u5—y5 by the usual Monte Carlo estimate,
the amount of current at the jth electrode site that overlaps
the region of interest Si , constrained to be a value between " Kit
#
O 1 XT
[
0 and 1. We estimate H 4fi 1 xij 5 using Monte Carlo integra- fi 4u5 D I u2 Tik 4Ìikt 1 ikt 5 1 (4)
tion by simulating J locations dij from the current distribution T ƒ t0 tDt0 C1 kD1

Dxi 4u5 and setting H b4fi 1 xij 5 equal to the proportion of these
where Ti1 4Ìi1t 1 i1t 51 : : : 1 TiKit 4ÌiKit t 1 iKit t 5 are the templates
locations that fall within any of the circular templates.
from the tth realization of the Markov chain, T is the total
The regression parameters  0 and  1 in (1) describe the
number of iterations, and t0 is the number of burn-in iterations.
error rates for the task under study. SpeciŽ cally,  0 is the log
The population posterior mean E4f ü 4u5—y5 may be estimated
odds of observing an impairment during DCE I testing given
similarly, replacing the subject-speciŽ c parameters in (4) with
that the electrical current does not overlap the region needed
the corresponding population parameters. These estimates are
for the task. The slope parameter  1 describes the change in
preferred over a measure of the region with maximum a pos-
log odds of an impairment as the amount of current overlap-
teriori density, because they result in a smoothed estimate of
ping the region of interest increases. The prior distributions
the region of interest, as opposed to a union of circles.
for the regression coefŽ cients are taken to be Gaussian for  0
For the multiple-subject template mixture model, we set up
and  1 .
a hybrid sampler with six steps: (1) update the population
Observed values yij are assumed to be conditionally inde-
template parameters and precisions, (2) update the mean num-
pendent given the true underlying image fi . It follows that the
ber of individual templates, (3) update the individual template
likelihood is simply
parameters, (4) update the regression parameters, (5) add or
N Ji
remove a template for each individual, and (6) birth or death of
Y
N
Y Y
p4yi —Ìi 1 Ði 1 Ki 1 Á5 µ
y
 ijij 41 ƒ  ij 51ƒyij 1 (2) a population template and associated individual templates. For
iD1 iD1 jD1 update steps 1–4, each parameter is updated in turn by condi-
tioning on all other parameters in the style of Gibbs sampling.
where logit4 ij 5 D  0 C 1 H 4fi 1 xij 5. The posterior distribution Because we cannot sample directly from the full conditional
is written as distributions, we use Metropolis–Hastings steps. For step 1,
to update the population template locations and correspond-
p4K ü 1 ‹1 K1 z1 Ìü 1 Ðü 1 Î1 Ì1 Ð1 Á—y5 ing precisions, we take the candidate distributions to be the
full Gaussian and gamma conditional distributions under the
D Z ƒ1 p4Á5p4K ü 5p4‹5p4Ìü 1 Ðü —K ü 5p4ΗK ü 5 assumption of no truncation in the subject-speciŽ c parameter
" # distributions. These distributions closely approximate the full
Y
N
€ ü ü ü ü
p4Ki —K 5p4zi —Ki 1 K 5p4Ìi 1 Ði —Ki 1 zi 1 Ì 1 Ð 1 Î5 conditionals if the distributions of the subject-speciŽ c template
iD1 parameters are not highly skewed (i.e., if there is minimal
N Ji
truncation). Because there is likely to be signiŽ cant trunca-
Y Y tion in the subject-speciŽ c radii distributions given the prior
€ p4yij —Ìi 1 Ði 1 Ki 1 Á51 (3)
iD1 jD1 range speciŽ ed in Section 5, we use a random walk Metropo-
lis sampler (Gilks, Richardson, and Spiegelhalter 1996) for
where Z is an unknown normalizing constant. The Ž rst line of the population template radii and corresponding precisions.
(3) contains all hyperparameters and population distributions, Similarly, for step 2, we also use a random walk sampler to
the second line gives the individual distributions, and the third update the mean number of individual templates. For steps 3
line is the likelihood component. and 4, to update the subject-speciŽ c template parameters and
Miglioretti, McCulloch, and Zeger: Direct Cortical Electrical Interference 129

the regression parameters, we use grid-based candidate distri- 4. IMAGE VARIATION


butions (Tierney 1994), which we found much more efŽ cient
In functional brain mapping, the cortical regions associated
than a generic proposal distribution.
with some tasks are more variable than the regions associ-
To explore the combined parameter space, we include two
ated with other tasks. A quantitative measure of this vari-
birth/death steps designed to allow jumps between models
ability would be useful for describing how the locations of
with different numbers of individual and population templates.
regions needed for different functions vary within populations.
For the individual birth/death step, performed separately for
For example, we might expect language regions to be more
each individual, we attempt to add or remove one subject-
variable in epilepsy patients with seizure foci in the temporal
speciŽ c template. For each individual, a random choice is
lobe if language functions are displaced to areas outside this
made between proposing the birth or death of a template with
region due to the disease (see, e.g., Ojemann and Whitaker
probabilities bKi and dKi D 1 ƒ bKi . We take bKi D dKi D 05.
1978). Measuring variability in this case is not straightforward,
Appendix A gives a detailed exposition of the algorithm along
because each individual’s dataset is an image as opposed to
with derivations of the acceptance probabilities. Brie y, for
one or even a few numbers. In this section we propose a plau-
the ith individual’s birth step, a population template is selected
sible measure of variability in a set of binary images.
from those for which that individual does not have an associ-
Following Baddeley (1992), we deŽ ne d4x1 S5 as the short-
ated template. A new set of individual template parameters is
est distance between a pixel in an image x 2 X and a region
drawn from their conditional distributions given the selected
in the image A 2 X2 d4x1 S5 D min84x1 a52 a 2 A9, where
population template. The new template is accepted with prob-
4x1 a5 is some measure of distance, such as the Euclidean
ability 4Ki 1 Ki C 15 D min811 A9, where
distance, between the pixel and the point a. We propose the
‹ p4yi —Ki C 11 Ìi 1 Ði 1 Ì0i 1 Ð0i 1 Á5 following estimate of variation among N images:
AD (5)
Ki C 1 p4yi —Ki 1 Ìi 1 Ði 1 Á5
V 4S1 1 : : : 1 SN 5
and 4Ì0i Ð0i 5 is the vector of proposed new template parameters. 1 X
D var6d4x1 S1 51 : : : 1 d4x1 SN 57
If the number of individual templates Ki equals the number of M x2X
population templates K ü at that sweep, then nothing is done. ³ ´ N
For the reverse (death) step, an individual template to delete 1 X 1 N ƒ1 X X
D 6d4x1 Si 5 ƒ d4x1 Sj 572 0
is randomly chosen. If individual i has no templates at that M x2X 2 2 iD1 j>i
sweep, then nothing is done. The individual death step mov-
ing from Ki C 1 to Ki templates is accepted with probability The term inside the large braces is motivated by an unbiased
4Ki C 11 Ki 5 D min811 Aƒ1 9, where A is deŽ ned in (5). estimate of the variance of d4x1 Si 5, because the expectation
For the population birth/death step, we propose to add or of the squared difference between two independent and iden-
delete a population template and associated individual tem- tically distributed random variables y1 and y2 is an estimate
plates. A random choice is made between the birth and death of twice the variance of yi . Thus V 4Si 5 is a measure of the
steps with probabilities bK ü and dK ü D 1 ƒ bK ü , where we take average variance of the distance of a pixel x to Si , averaged
bK ü D dK ü D 05. Appendix B outlines the algorithm in detail. over all M pixels in the image X. This estimator is closely
For the birth step, a new set of population template parameters related to Baddeley’s ã2 (Baddeley 1992), which is deŽ ned as
is selected from their prior distributions. A random number the square root of the average squared difference between the
of individuals are chosen for which new subject-speciŽ c tem- shortest distance from a pixel to a region in one image and the
plates associated with the new population template are pro- distance from the same pixel to a region in a second image,
posed (with the possibility of no individuals being selected). averaging over all pixels in the image.
New template parameters for these individuals are drawn
from their conditional distribution given the new population 5. EXAMPLE
template. The new population and individual templates are At the Johns Hopkins Medical Institutions, DCE I is used
accepted with probability 4K ü 1 K ü C 15 D min811 LR € B9, for functional brain mapping in surgical resection candidates
where LR is the likelihood ratio, to identify cortical areas critical for motor and language func-
tions to avoid postresection deŽ cits (Gordon et al. 1996). Elec-
" #" #
‹ü Y ‹ YN
p4ki µ K ü —K ü 5 trodes are arranged in rectangular arrays composed of 2.3-mm
BD ü diameter exposed platinum iridium disks spaced 1 cm apart
K C 1 i2I Ki C 1 iD1 p4ki µ K ü C 1—K ü C 15
(Lesser et al. 1987). These arrays are surgically placed into
Q
iyI 4K ƒ K i C 15
ü the patient’s cortical surface so they cover the expected loca-
€ Q 1 (6) tion of the seizure focus to be resected and surrounding areas.
4K ü C 15N i2I 4Ki C 15
(For details on the clinical DCE I testing protocol used, see
and I is the set of individuals for which a new template is Lesser et al. 1987 and Uematsu et al. 1990.)
proposed. For the reverse (death) step, a population template To demonstrate our methodology, we use the template mix-
and associated individual templates to delete are randomly ture model to estimate the regions associated with movement
chosen. The acceptance probability for the death step mov- of the tongue and hands during DCE I testing in 20 patients.
ing from K ü C 1 to K ü population templates 4K ü C 11 K ü 5 is The raw data are displayed in Figures 4 and 5. Black dots
min811 LR € B ƒ1 9, where B is deŽ ned in (6). with lines represent electrodes associated with a positive motor
130 Journal of the American Statistical Association, March 2002

response during DCE I testing (i.e., yij D 15. Gray lines rep-
resent tested electrode sites with no motor response (i.e.,
yij D 05. During DCE I testing, a patient is asked to stick out
and wiggle the tongue or to hold out the hands and wiggle
the Ž ngers. A positive motor response is deŽ ned as either
(1) an induced motor response, in which the tongue or hand
jerks or twitches during DCE I, or (2) motor arrest, in which
movement of the tongue or Ž ngers stops during DCE I. Before
analysis, each individual’s electrode locations were registered
to the Talairach atlas (Talairach and Tournoux 1988). Regis-
tration was performed manually by comparing the electrode
locations to structural features of the brain using photographs
taken at surgery, skull x-rays, and a three-dimensional com-
puted tomography (CT) scan of the electrodes superimposed
onto a three-dimensional brain reconstruction from volumetric Figure 3. Two Hundred Fifty Samples from  i j Induced by the Prior
magnetic resonance imaging (MR I) scans. Distribution on (  0 ,  1 ) . The prior for  0 is Gaussian with mean - 12
and standard deviation .5, and the prior for  1 is Gaussian with mean
Because there is only one observation per electrode site per
70 and standard deviation 15.
patient, there is little information to estimate the regression
coefŽ cients (see Miglioretti et al. 2000). However, there is
strong prior information about these parameters. The data con-
40,000 MCMC sweeps, discarded the Ž rst 20,000 sweeps, and
sist of the binary clinical judgement as to whether there was
saved every Ž fth sweep for the results.
a positive motor response at each electrode site, which appear
To assess convergence, we ran multiple chains starting at
to have very low error rates, because any uncertainty about the
different initial values and noted no signiŽ cant difference in
results at a particular site simply prompts continued testing
the results. Acceptance rates were between 22% and 65% for
until a conŽ dent decision can be made. In addition, it is known
the individual birth/death step and between 7% and 9% for
that a small overlap of the electrical current with the corti-
population birth/death step. The mean number of population
cal area responsible for motor control should cause a positive
templates is 8.0 (range 4–16) for tongue movements and 6.4
motor response. Thus the probability of a positive response
(range 4–14) for hand movements. The mode of the mean
should be very close to 0 when H 4fi 1 xij 5 ’ 0 (i.e., the elec-
trical current does not overlap with the region of interest) and number of individual templates ‹ is 7.9 for tongue movements
very close to 1 when even a small portion of the current den- [95% highest posterior density interval (HPD) D 406–14.3] and
sity is over the region of interest, say when H 4fi 1 xij 5 > 025. 4.9 for hand movements (95% HPD D 2.6–11.2).
In addition, the probability should increase quickly from 0 We examined model Ž t by comparing the observed numbers
to 1. There is some uncertainty about the regression coefŽ - of electrode sites associated with impairments for each indi-
cients, because it is not known how much current overlap is vidual to the expected frequencies simulated from the poste-
required to cause an impairment, although it is likely that the rior predictive distributions for the two motor tasks (Gelman,
amount of overlap needed is small. To take this uncertainty Carlin, Stern, and Rubin 1995). In addition, we overlaid spa-
into account, we chose the priors for  0 and  1 to include a tial plots of the observed data with the posterior probability
range of coefŽ cients that meet the aforementioned conditions. contours for each individual (Figs. 4 and 5). Both methods
We set the prior for  0 to Gaussian with mean ƒ12 and stan- indicate no lack of Ž t of the model.
dard deviation .5 and the prior for  1 to Gaussian with mean Figures 4 and 5 show the individual posterior probability
70 and standard deviation 15. Figure 3 shows the associated contours for the regions associated with tongue movements
uncertainty in the probability of impairment as a function of and hand movements. Positive tongue responses associated
the amount of current overlapping the region. with DCE I testing were found in 19 of the 20 patients. In
The prior for K ü was taken to be Poisson(3), and the prior the patient without an observed tongue response (patient 20),
for ‹ was gamma 41051 055 to encourage a small number of it appears that the electrode grid was likely implanted below
templates. We set the prior range for Ìük to be the range of the region associated with tongue movement. Hand responses
the electrode grids. The prior for kü was taken to be uniform were observed in 14 of the 20 patients during DCE I test-
from .1 to 1 cm, with the maximum being the typical spac- ing, and it appears again that the grids were likely implanted
ing between two electrodes. The limits were chosen to keep below the areas associated with hand movement in the other
the circles reasonably sized to better approximate the regions. patients. This is consistent with the literature, because the
Given the goals of analysis, circles more than 1 or 2 cm apart cortical regions associated with hand movements are consis-
should come from different distributions. Taking this into con- tently shown to be located superior and slightly posterior to
sideration, we chose the prior parameters for Î as follows. those for tongue movements (Crone, Miglioretti, Gordon, and
We took the precisions for Œü1 and Œü2 to be gamma 451 35, Lesser 1998a; Crone et al. 1998b; PenŽ eld and Boldrey 1937;
to restrict circles from the same population to be reasonably Pfurtscheller, Flotzinger, and Neuper 1994; Uematsu et al.
close to each other, and we took the precisions for ü to 1992). The results also agree with recent studies showing that
be gamma 451 15. The sampler was programmed in SAS/ IML the areas associated with the hand and the tongue overlap and
(SAS Institute, Inc. 1990). We implemented each model for are diffusely spread throughout the sensorimotor cortex (Crone
Miglioretti, McCulloch, and Zeger: Direct Cortical Electrical Interference 131

Figure 4. Individual Posterior Probability Contours for the Regions Associated With Tongue Movement in 20 Patients. Black dots with lines
represent electrodes associated with a positive motor response during DCEI testing. Gray lines represent tested sites with no motor response.
From lightest to darkest, posterior contours correspond to .05 µ p < .10, .10 µ p < .20, and p > .20.

et al. 1998a; Pascual-Leone, Cohen, Brasil-Neto, and Hallett the testing area. If the information from the other patients had
1994; Schlaug, Knorr, and Seitz 1994). been ignored, then the area associated with hand and tongue
There are several interesting items to note about the movements would be smaller in these patients. Finally, in
individual posterior contours. First, several patients had no many patients, there is shrinkage of the area estimates toward
observed positive motor responses during DCE I testing. By the population mean. For example, for patients 1, 2, 6, and 7
borrowing strength across patients, the template mixture for the tongue task and patients 1, 3, 5, and 7 for the hand
model gives a coherent estimate of the areas associated with task, there are electrode pairs where the posterior probability
hand and tongue movements. If information from the other is more concentrated near one electrode–the one closest to the
subjects had been ignored, then the data would indicate that population region.
these individuals had no area associated with tongue and hand The next set of Ž gures highlights the richness of results
movements. In addition, most subjects have areas estimated to that we can obtain from this type of analysis, because we can
be associated with both hand and tongue movements outside calculate posterior distributions on any function of the model

Figure 5. Individual Posterior Probability Contours for the Regions Associated With Hand Movement in 20 patients. Black dots with lines
represent electrodes associated with a positive motor response during DCEI testing. Gray lines represent tested sites with no motor response.
From lightest to darkest, posterior contours correspond to .05 µ p < .10, .10 µ p < .20, and p > .20.
132 Journal of the American Statistical Association, March 2002

Figure 6. Posterior Densities for the Individual Region Sizes for


Hand (Black) and Tongue (Gray) Movements. The solid lines corre- Figure 7. Posterior Densities for the Intersubject Image Variation for
Tongue (Gray) and Hand (Black) Movements. The solid lines corre-
spond to posterior modes, and the dashed lines correspond to the 95%
highest posterior density intervals. spond to the posterior mode, and the dashed lines correspond to the
95% HPD intervals.

parameters. For example, the region size can be estimated at


population posterior distributions overlap more than the dis-
each sweep by calculating the size of the area covered by at
tributions for the individual region sizes. Although the indi-
least one template. The variability can be calculated using our
vidual hand regions are smaller than the tongue regions, there
measure deŽ ned in Section 4.
is some evidence of greater variability in the regions associ-
Figure 6 shows the posterior densities for the average
ated with hand movements in this set of patients (see Fig. 7).
individual region sizes for hand (black) and tongue (gray)
This combination results in population regions that are more
movements. The solid lines correspond to the modes, and
similar in size.
the dashed lines correspond to the 95% HPD intervals. The
modal region size for hand movements in the patients stud-
ied is 4.4 cm2 (95% HPD D 301–7.0 cm2 ). The region size 6. DISCUSSION
for tongue movements is 50% larger (mode D 607 cm2 , 95% In this article, we have extended the template mixture model
HPD D 5.2–8.8 cm2 ). proposed in an earlier article (Miglioretti et al. 2000) to mul-
There is some evidence that the region associated with tiple subjects by embedding the single-subject model into a
tongue movement is larger than the region associated with larger hierarchical model with a population component. We
hand movement in these patients (posterior probability D 092, model the individual and population regions of interest using
posterior mode for difference D 200 cm2 , 95% HPD D ƒ083– a mixture of subject-speciŽ c and population templates. The
4.7 cm2 ). However, the difference could be due to the fact that subject-speciŽ c templates are treated as random effects, gov-
the cortical area associated with tongue movements is inferior erned by the population templates. To model the observed
to that for hand movements, and the electrode grids in these data, we use a generalized linear model. SpeciŽ cally, for mod-
patients tended to cover more of the inferior portion of the eling binary DCE I data, we use logistic regression. Because
sensorimotor cortex. the dimension of the parameter space depends on the num-
The posterior densities for the intersubject image variation bers of templates, we Ž t the model using the reversible jump
for the two tasks is displayed in Figure 7. In this study popu- methodology of Green (1995).
lation, the region associated with hand movements is slightly
more variable than the region associated with tongue move-
ments (posterior probability D 074). The image variability for
the regions associated with the hand is 1.5 (95% HPD D 071–
3.1), whereas that for the tongue is 1.3 (95% HPD D 055–2.0).
The posterior contours for the population regions associated
with tongue and hand movements are displayed in Figure 8.
Across the population, there is considerable overlap between
the regions associated with hand and tongue movements; how-
ever, the cortical area associated with hand movements tends
to be superior and posterior to tongue movements.
Figure 9 shows the posterior densities for the popula- Figure 8. Population Posterior Probability Contours for the Regions
tion region sizes for the hand (black) and tongue (gray) Associated With Tongue Movement and Hand Movement in the Study
Population. Black dots with lines represent electrodes associated with
movements. The population region associated with tongue a positive motor response during DCEI testing in 20 patients. Gray lines
movements (mode D 905 cm2 , 95% HPD D 502–13.5 cm2 ) is represent electrodes with no motor response. From lightest to dark-
slightly larger than the region associated with hand movements est, posterior contours correspond to .05 µ p < .10, .10 µ p < .20, and
(mode D 701 cm2 , 95% HPD D 302–12.3 cm2 ); however, the p > .20.
Miglioretti, McCulloch, and Zeger: Direct Cortical Electrical Interference 133

DCE I analysis, because the length of time an electrode grid


can be left in a patient for testing is limited by cost concerns,
greatly limiting the amount of research testing that is possible.
The RJMCMC method allows us to avoid prespecifying the
numbers of templates used in the model. More templates are
automatically Ž t for individuals with larger regions, and fewer
templates Ž t for those with smaller regions. Without this  ex-
ibility, if the number of templates were Ž xed to be too small,
then regions of interest would not be modeled adequately, and
if the number of templates were set too large, then the addi-
tional templates not needed to model the region either would
be distributed throughout the region where there is no infor-
mation (i.e., where there are no electrodes) or would overlap
Figure 9. Posterior Densities for the Population Region Sizes for
with other templates. By modeling the joint posterior distri-
Hand (Black) and Tongue (Gray) Movements in 20 Patients. The solid
lines correspond to posterior modes, and the dashed lines correspond bution for the numbers of individual and population templates
to the 95% HPD intervals. and the parameters, and by using a Poisson prior for the num-
bers of templates with a low mean, RJMCMC discourages the
addition of templates that do not increase the likelihood. Thus
To estimate each individual image and the population
a parsimonious model is automatically encouraged.
image, we use an estimate of the posterior mean of the region.
When implementing the RJMCMC methodology, develop-
Because these estimates are marginalized over the number of
ment of the dimension-changing move types requires some
templates, the region shapes are not restricted to be a union of
care to ensure that the model mixes well across the differ-
circles. Thus this method is useful when the analytic goal is to
ent number of templates. In their article on modeling mix-
model binary regions using a small number of parameters, but
there is minimal information about the shapes and locations of tures with an unknown number of components, Richardson
the areas of interest in the underlying image. The model’s par- and Green (1997) implemented a birth/death step with empty
simony is essential for combining information across subjects. clusters only. In our model, this would correspond to popula-
By using mixtures of subject-speciŽ c templates, this approach tion templates with no associated individual templates. Origi-
can model a wide range of shapes across subjects. nally, we tried using this birth/death step instead of the more
For DCE I and other functional brain mapping techniques, complicated population birth/death step described in Section 3.
including multiple subjects in a single hierarchical model has However, we found that there were rarely any empty popula-
several advantages. First, we may better estimate an individ- tion templates to propose deleting, especially if the number of
ual’s brain map by borrowing strength from other subjects’ population templates, K ü , was close to the number of individ-
data. DCE I testing for a single subject is limited by both the ual templates, Ki . As a result, the algorithm would not mix
location of the electrode grid and the fact that all electrode well for K ü close to any Ki . Similar behavior was noted by
pairs are not tested. By using information from other patients, Castelloe (1998, 1999) in his model for spatial Poisson cluster
we can make reasonable estimates about the areas associated processes.
with the task under study outside the testing area in that spe-
ciŽ c patient. This was demonstrated in our example, where
the regions associated with hand and tongue movements were APPENDIX A: INDIVIDUAL BIRTH/DEATH STEP
at least partially outside the testing area for many subjects. ALGORITHM AND ACCEPTANCE PROBABILITIES
A second advantage of a hierarchical approach is that infor- To simplify notation, let Èi D 4Ìi 1 Ði 5 and Èü D 4Ìü 1 Ðü 5 represent
mation from past patients can be used to predict where to the subject-speciŽ c and population template parameters. The indi-
test a new patient. This is becoming an important objective of vidual birth and death step is detailed in Table A.1. The acceptance

Table A.1. Individual Birth/Death Step

For i D 11 : : : 1 N, with probability .5, go to the birth step. Otherwise, go to the death step.
Birth step
1. Select a population template v U({ 11 : : : 1 K ü } n{ z i 1 1 : : : 1 z i Ki } ).
2. Draw new individual templates parameters u p(Èi —Èüv 1 Îv ).
3. Draw an index for the new individual template l U({ 11 : : : 1 Ki C 1} ).
4. With probability (Ki 1 Ki C 1), update the parameters for individual i as follows:
For k < l , keep Èi k and z i k .
For k D l , set Èi k D u and z i k D v.
For k > l , set Èi k C1 D Èi k and z i k C1 D z i k .
Otherwise, keep the current values.
Death step
1. Randomly select an individual template l U({ 11 : : : 1 Ki } ).
2. With probability (Ki 1 Ki ƒ 1), delete template Èi l and allocation variable z i l .
Otherwise, keep the current values.
134 Journal of the American Statistical Association, March 2002

probability for the individual birth step is tion birth and death steps are detailed in Table B.1. For the popula-
tion birth step, the prior ratio is given by
4Ki 1 Ki C 15 D min411 prior ratio € likelihood ratio µ ¶
p4K ü C 15 Y
N
p4Ki0 —K ü C 15 p4zi0 —Ki0 1 K ü C 15
prior ratio D
€ proposal ratio € Jacobian50 p4K ü 5 iD1 p4Ki —K ü 5 p4zi —Ki 1 K ü 5

In this case, the Jacobian is 1, because the generated random vari- p4Èü 0 1 Î0 —K ü C 15 Y
N
p4È0i —Ki0 1 zi0 1 Èü 0 1 Î0 5
€
ables are simply equal to the proposed new template parameters. The p4Èü 1 ΗK ü 5 iD1 p4Èi —K1 zi 1 Èü 1 Î0 5
likelihood is deŽ ned in (2). The prior ratio is given by µ ¶µ N ¶
‹ü Y ‹ Y p4ki µ K ü —K ü 5
D ü
p4Ki C 1—K ü 5 p4z0i —Ki C 11 K ü 5 K C 1 i2I Ki C 1 iD1
p4ki µ K ü C 1—K ü C 15
prior ratio D Q
p4Ki —K ü 5 p4zi —Ki 1 K ü 5 i62 I 4K
ü
ƒ Ki C 15
€ p4È0ülü 1 Î0l ü 5
p4È0i1 1 : : : 1 È0ikC1 —z0i 1 Ki C 11 Èü 5 4K ü C 15N
€ Y
p4Èi1 1 : : : 1 ÈiKi —zi 1 Ki 1 Èü 5 € p4È0ili —zili D lü 1 Èülü 0 1 Î0l ü 50
‹ 1 i2I
D p4È0il —zil D v1 Èüv 1 Îv 51
Ki C 1 K ü ƒ Ki The proposal ratio is given by

where z0i and È0ik are the candidate values. The proposal ratio is proposal ratio
given by
bkC1 hd 4lü —K ü C 15
D Q
dKi C1 hd 4l—Ki C 15 dk hb 4uü 1 vü 5hb 4lü —K ü 5 i2I 6hb 4li —K ü 1 Ki 5hb 4ui —uü 1 v ü 57
proposal ratio D Q
bKi h b 4v—zi 1 Kü 5hb 4l—Ki 5hb 4u—z0il D v1 Èüv 1 Îv 5 i2I 4Ki C 15
D Q 0
ü
K ƒ Ki p4Èül 0 1 Î0l 5 i2I p4È 0 0 ü 0 0
ili —K i 1 zili D l 1 Èlü 1 Îl ü 5
ü
D 1
p4È0il —zil D v1 Èüv 1 Îv 5
The Jacobian is equal to 1. Thus the acceptance probability for the
where hb and hd are the candidate distributions for the birth and death birth step 4K ü 1 K ü C 15 is min811 LR € B9, where
steps. Thus the acceptance probability simpliŽ es to 4Ki 1 Ki C 15 D
µ ¶µ N ¶
min811 A9, where A D 4‹=Ki C 154p4yi —Ki C 11 È0i Á5=p4yi —Ki 1 Èi 1 Á55. ‹ü Y ‹ Y p4ki µ K ü —K ü 5
BD
For the death step moving from Ki C 1 to Ki 1 the acceptance proba- K C 1 i2I Ki C 1
ü
iD1 p4ki µ K C 1—K C 15
ü ü

bility is 4Ki C 11 Ki 5 D min811 Aƒ1 9. Q ü


i62I 4K ƒ Ki C 15
€ Q
APPENDIX B: POPULATION BIRTH/DEATH STEP 4K ü C 15N i2I 4Ki C 15
ALGORITHM AND ACCEPTANCE PROBABILITIES
and LR is the likelihood ratio. The likelihood is deŽ ned in (2). Simi-
To simplify notation, let Èi D 4Ìi 1 Ði 5 and Èü D 4Ìü 1 Ðü 5 represent larly, the acceptance probability for the death step moving from K ü C
the subject-speciŽ c and population template parameters. The popula- 1 to K ü population templates 4K ü C 11 K ü 5 is min811 LR € B ƒ1 9.

Table B.1. Population Birth/Death Step

With probability, .5, go to the birth step. Otherwise, go to the death step.
Birth step
1. Draw new population template parameters uü p(Èü ) and v ü p(Î).
2. Draw a population template index l ü U({ 11 : : : 1 K ü C 1} ).
3. Draw an integer M U({ 01 : : : 1N } ).
4. If M > 0,
Draw a set of individual indices I D { i 1 1 : : : 1 i M } without replacement from { 11 : : : 1 N } .
For i 2 I:
Draw new template parameters ui p(Èi —Èüv 1 Îv ).
Draw an index for the new template l i U({ 11 : : : 1 Ki C 1} ).
5. With probability  (K ü 1 K ü C 1):
Update the population parameters as follows:
For k < l ü , keep Èük and Îk .
For k D l ü , set Èük D uü and Îk D vü .
For k > l ü , set ÈükC1 D Èük and Îk C1 D Îk .
For i 2 I, update the individual parameters as follows:
For k < l i , keep Èi k and z i k .
For k D l i , set Èi k D u and z i k D l ü .
For k > l i , set Èi k C1 D Èi k and z i kC1 D z i k .
For i 62 I, keep the current values.
Otherwise, keep the current values.
Death step
1. Randomly select a population template l U({ 11 : : : 1 Ki } ).
2. With a probability  (K1 K ƒ 1), delete population template Èül , the associated precisions Îl ,
and all individual template parameters Èi k and allocation variables z i k such that z i k D l .
Otherwise, keep the current values.
Miglioretti, McCulloch, and Zeger: Direct Cortical Electrical Interference 135

[Received April 2000. Revised September 2001.] Models for Direct Cortical Electrical Interference Data,” Biostatistics, 1,
403–421.
REFERENCES Nathan, S. S., Sinha, S. R., Gordon, B., Lesser, R. P., and Thakor, N. V.
(1993), “Determination of Current Density Distributions Generated by
Baddeley, A. J. (1992), “An Error Metric for Binary Images,” in Robust Com- Electrical Stimulation of the Human Cerebral Cortex,” Electroencephalog-
puter Vision, Quality of Vision Algorithms, in Proceedings of the Interna- raphy and Clinical Neurophysiolog y, 88, 183–192.
tional Workshop on Robust Computer Vision, pp. 59–78. Nii, Y., Uematsu, S., Lesser, R. P. and Gordon, B. (1996), “Does the
Boatman, D., Hall, C., Goldstein, M., Lesser, R., and Gordon, B. (1997), Central Sulcus Divide Motor and Sensory Functions?,” Neurology, 46,
“Neuroperceptua l Differences in Consonant and Vowel Discrimination 360–367.
as Revealed by Direct Cortical Electrical Interference,” Cortex, 33, Ojemann, G. A. (1979), “ Individual Variability in Cortical Localization of
83–98. Language,” Journal of Neurosurger y, 50, 164–169.
Boatman, D., Lesser, R., and Gordon, B. (1995), “Auditory Speech Processing (1981), “ Intrahemispheric Localization of Language, Memory, and
in the Left Temporal Lobe: An Electrical Interference Study,” Brain and Motor Mechanisms in Human Cortex and Thalamus,” in New Perspec-
Language, 51, 269–290. tives in Cerebral Localization, ed. R. Thompson, New York: Raven Press,
Boatman, D., Lesser, R., Hall, C., and Gordon, B. (1994), “Auditory Percep- pp. 157–175.
tion of Segmental Features: A Functional Feuroanatomi c Study,” Journal Pascual-Leone, A., Cohen, L. G., Brasil-Neto, J. P., and Hallett, M. (1994),
of Neurolinguistics, 8, 225–234. “Non-invasive Differentiation of Motor Cortical Representation of Hand
Castelloe, J. M. (1998), “ Issues in Reversible Jump Markov Chain Monte Muscles by Mapping of Optimal Current Directions,” Electroencephalog-
Carlo and Composite EM Analysis, Applied to Spatial Poisson Clus- raphy and Clinical Neurophysiolog y, 93, 42–48.
ter Processes,” Ph.D. dissertation, University of Iowa, Department of PenŽ eld, W., and Boldrey, E. (1937), “Somatic Motor and Sensory Represen-
Statistics. tation in the Cerebral Cortex of Man as Studied by Electrical Stimulation,”
(1999), “Reversible Jump Markov Chain Monte Carlo Analysis of Brain, 60, 389–443.
Spatial Poisson Cluster Processes With Bivariate Normal Displacement,” Pfurtscheller, G., Flotzinger, D., and Neuper, C. (1994), “Differentiation
Computing Science and Statistics, 31, 306–315. Between Finger, Toe and Tongue Movement in Man Based on 40
Christiansen, C. L., and Morris, C. N. (1997), “Hierarchical Poisson Regres- Hz EEG,” Electroencephalograph y and Clinical Neurophysiolog y, 90,
sion Modeling,” Journal of the American Statistical Association, 92, 618– 456–460.
632. Rao, C. R. (1965), “The Theory of Least Squares When the Parameters
Crone, N. E., Miglioretti, D. L., Gordon, B., and Lesser, R. P. (1998a), “Func- are Stochastic and Its Application to the Analysis of Growth Curves,”
tional Mapping of Human Sensorimotor Cortex With Electrocorticographi c Biometrika, 52, 447–458.
Spectral Analysis II. Event-Related Synchronizatio n in the Gamma Band,” Richardson, S., and Green, P. J. (1997), “On Bayesian Analysis of Mixtures
Brain, 121, 2301–2315. With an Unknown Number of Components,” Journal of the Royal Statistical
Crone, N. E., Miglioretti, D. L., Gordon, B., Sieracki, J. M., Wilson, M. T., Society, Ser. B, 59, 731–792.
Uematsu, S., and Lesser, R. P. (1998b), “Functional Mapping of Human SAS Institute, Inc. (1990), SAS/ IML Software, Usage and Reference, Version
Sensorimotor Cortex With Electrocorticographi c Spectral Analysis I. Alpha 6, First Edition, Cary, NC: Author.
and Beta Event-Related Desynchronization,” Brain, 121, 2271–2299. Schlaug, G., Knorr, U., and Seitz, R. (1994), “ Inter-Subject Variability of
Gelman, A., Carlin, J. B., Stern, H. S., and Rubin, D. B. (1995), Bayesian Cerebral Activations in Acquiring a Motor Skill, a Study With Positron
Data Analysis, London: Chapman and Hall. Emission Tomograph y,” Experimental Brain Research, 98, 523–534.
Gilks, W. R., Richardson, S., and Spiegelhalter, D. J. (1996), Markov Chain Stiratelli, R., Laird, N., and Ware, J. H. (1984), “Random-Effects Models for
Monte Carlo in Practice. London: Chapman and Hall. Serial Observations With Binary Response,” Biometrics, 40, 961–971.
Gordon, B., Hart, J. J., Boatman, D., and Lesser, R. P. (1996), “Cortical Talairach, J., and Tournoux, P. (1988), Co-Planar Stereotactic Atlas of the
Stimulation ( Interference) During Behavior,” in Behavioral Neurology and Human Brain, New York: Thieme.
Neuropsychology, eds. T. E. Feinberg and M. Farah, Chicago: McGraw-Hill. Tierney, L. (1994), “Markov Chains for Exploring Posterior Distributions”
Green, P. J. (1995), “Reversible Jump Markov Chain Monte Carlo Computa- (with discussion), The Annals of Statistics, 22, 1701–1762.
tion and Bayesian Model Determination,” Biometrika, 82, 711–32. Uematsu, S., Lesser, R., Fisher, R. S., Gordon, B., Krauss, G., Vining,
Laird, N. M., and Ware, J. H. (1982), “Random-Effects Models for Longitu- E. P., and Webber, R. W. (1992), “Motor and Sensory Cortex in Humans:
dinal Data,” Biometrics, 38, 963–974. Typograph y Studied With Chronic Subdural Stimulation,” Neurosurgery,
Lee, H. S. (1997), “Random Effects Models for Count Data,” Communications 31, 59–72.
in Statistics, Part A—Theory and Methods, 26, 1893–1904. Uematsu, S., Lesser, R., Fisher, R., Krauss, G., Hart, J., Vining, E. P., Free-
Lesser, R. P., Lüders, H., Klem, G., Dinner, D. S., Morris, H. H., Hahn, man, J., and Gordon, B. (1990), “Resection of the Epileptogenic Area in
J. F., and Wyllie, E. (1987), “Extraoperative Cortical Functional Localiza- Critical Cortex With the Aid of a Subdural Electrode Grid,” Stereotactic
tion in Patients With Epilepsy,” Journal of Clinical Neurophysiolog y, 4, and Functional Neurosurger y, 54–55, 34–45.
27–53. Urasaki, E., Uematsu, S., Gordon, B., and Lesser, R. P. (1994), “Cortical
Lindley, D. V., and Smith, A. F. M. (1972), “Bayes Estimates for the Linear Tongue Area Studied by Chronically Implanted Subdural Electrodes—With
Model,” Journal of the Royal Statistical Society, Ser. B, 34, 1–41. Special Reference to Parietal Motor and Frontal Sensory Responses,” Brain,
Malow, B., Blaxton, T., Sato, S., Bookheimer, S., Kufta, C., Figlozzi, C., and 117, 117–132.
Theodore, W. (1996), “Cortical Stimulation Elicits Regional Distinctions in Zeger, S. L., and Karim, M. R. (1991), “Generalized Linear Models With
Auditory and Visual Naming,” Epilepsia, 37, 245–252. Random Effects: A Gibbs Sampling Approach,” Journal of the American
Miglioretti, D. L., McCulloch, C., and Zeger, S. L. (2000), “Template Mixture Statistical Association, 70, 138–144.

You might also like