Cerebral Cortex November 2009;19:2746--2754
Advance Access publication March 25, 2009
Engagement of Fusiform Cortex and
Disengagement of Lateral Occipital Cortex
in the Acquisition of Radiological
Erin M. Harley1, Whitney B. Pope3, J. Pablo Villablanca3,
Jeanette Mumford2, Robert Suh3, John C. Mazziotta3--5,
Dieter Enzmann3and Stephen A. Engel6
1Exponent Failure Analysis, Bellevue, WA 98007, USA,
2Department of Psychology, University of California, Los Angeles,
Los Angeles, CA 90095, USA,3Department of Radiological
Sciences and Pharmacology,4Department of Neurology,
5Ahmanson-Lovelace Brain Mapping Center, David Geffen School
of Medicine, University of California, Los Angeles, Los Angeles,
CA 90095, USA and6Department of Psychology, University of
Minnesota, Minneapolis, MN 55455, USA
The human visual pathways that are specialized for object
recognition stretch from lateral occipital cortex (LO) to the ventral
surface of the temporal lobe, including the fusiform gyrus. Plasticity
in these pathways supports the acquisition of visual expertise, but
precisely how training affects the different regions remains
unclear. We used functional magnetic resonance imaging to
measure neural activity in both LO and the fusiform gyrus in
radiologists as they detected abnormalities in chest radiographs.
Activity in the right fusiform face area (FFA) correlated with visual
expertise, measured as behavioral performance during scanning. In
contrast, activity in left LO correlated negatively with expertise,
and the amount of LO that responded to radiographs was smaller in
experts than in novices. Activity in the FFA and LO correlated
negatively in experts, whereas in novices, the 2 regions showed no
stable relationship. Together, these results suggest that the FFA
becomes more engaged and left LO less engaged in interpreting
radiographic images over the course of training. Achieving expert
visual performance may involve suppressing existing neural
representations while simultaneously developing others.
Keywords: diagnosis, expert, FFA, radiology, vision
Human visual cortex contains many distinct regions that
respond more strongly to images of objects than to more
simple patterns (for reviews, see Grill-Spector 2003; Op de
Beeck et al. 2008). These include the lateral occipital cortex
(LO) and more ventrally, the posterior fusiform cortex.
Together, these areas have been termed the lateral occipital
complex. The ventral surface of the anterior occipital lobe and
posterior temporal lobe is also object selective and includes
a region that is highly responsive to images of faces, the
fusiform face area (FFA, Kanwisher et al. 1997).
Object processing in human cortex appears to be hierarchi-
cal. Neurons in anterior regions show responses that are more
closely tied to conscious recognition of object identity than
neurons in LO (Bar et al. 2001; Grill-Spector et al. 2004).
Conversely, neural responses in LO are more affected by
stimulus transformations that do not affect object identity such
as image size and retinal location (Grill-Spector et al. 1999;
Eger, Kell, and Kleinschmidt 2008). Neurons in LO also
integrate information over a smaller portion of the object than
neurons in more anterior areas (Lerner et al. 2001, 2002), and
LO neural responses are most closely related to physical
measures of similarity between objects, whereas fusiform
responses are more closely related to subjective measures
of similarity (Haushofer et al. 2008). LO responses do not
distinguish between real and ‘‘nonsense’’ objects, whereas
? 2009 The Authors
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/2.0/uk/) which
permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
anterior responses do (Vuilleumier et al. 2002). Developmen-
tally, LO is relatively stable from ages 7--16, while areas on the
fusiform gyrus specialized for faces and places increase in size
(Golarai et al. 2007; Scherf et al. 2007; Grill-Spector et al. 2008).
Together, these results suggest that LO may contain a repre-
sentation of the shape of objects (Vinberg and Grill-Spector
2008), and the more ventral and anterior areas explicitly
encode object identity.
Object selective cortex retains considerable plasticity in the
adult, but the precise nature of this plasticity remains unclear.
Training subjects in the laboratory to recognize briefly
presented objects or to make fine discriminations between
artificial stimuli generally increases neural response strength to
the trained stimuli in both parts of fusiform cortex and parts of
LO (Gauthier et al. 1999; Grill-Spector et al. 2000; Op de Beeck
et al. 2006; Jiang et al. 2007; but see Yue et al. 2006). Shape
discrimination training also appears to narrow neural tuning in
LO (Yue et al. 2006; Jiang et al. 2007). Complementing these
with years of experience in a given domain. Expertise in birds,
cars, and butterflies produces increased activity for objects of
expertise primarily in ventral cortex (Gauthier et al. 2000;
Rhodes et al. 2004; Xu 2005; but see Grill-Spector et al. 2004).
Most of these studies, however, focused upon the FFA and did
not localize a priori regions of interest in LO, and one study that
examined LO responses post hoc actually found a decrease in
activity with expertise (Gauthier et al. 2000).
The goal of the present work is to examine responses in
both fusiform cortex and LO across the development of real-
world expertise. We conducted a cross-sectional study of
radiologists, who, in addition to their knowledge of disease
processes, have documented visual expertise in detecting
features of radiographic images (e.g., Sowden et al. 2000). We
used functional magnetic resonance imaging (fMRI) to com-
pare neural activity in first-year radiology residents (n = 7),
fourth-year radiology residents (n = 6), and practicing expert
thoracic or general radiologists (n = 5 and 2, respectively) as
they detected abnormalities in chest radiographs. We first
identified regions of interest in cortex, including the FFA, the
part of LO that responded to chest radiographs, and retinotopic
visual areas (see Fig. 1A,B and Materials and Methods, below).
We then acquired a separate data set in which participants
performed a diagnosis task, judging whether a specified
location in a radiograph contained an abnormality or ‘‘nodule’’
(Fig. 1C). We averaged fMRI activity within each region of
interest (ROI) and compared that activity with behavioral
performance in the diagnosis task. We hypothesized that more
anterior areas of ventral cortex would show greater activity in
Materials and Methods
Participants were recruited at the Department of Radiology at
University of California, Los Angeles (UCLA) Medical Center from
3volunteersubjectpopulations:practicingthoracicradiologists(n = 7;3
females; mean age = 51.6), fourth-year radiology residents (n = 7;
3 females; mean age = 30.9), and first-year radiology residents (n = 7; 2
females; mean age = 28.6). Practicing radiologists had at least 10 years
experience postresidency (mean = 18.9 years). One fourth-year resident
had large uncorrectable motion artifacts in his fMRI data and was
residency in Radiology. All participants had normal or corrected-to-
normalvision.Informedconsentwas received fromeachparticipant,and
all experimental procedures were approved by the UCLA Office for the
Protection of Research Subjects.
Data were collected inside a magnetic resonance imaging (MRI) device
while participants underwent functional scans of their brains. Stimuli
were displayed on MR-compatible goggles controlled by a Macintosh
frame buffer. The goggle display subtended 31 degrees of visual angle
horizontally and 23 degrees vertically. While viewing stimuli, partic-
ipants indicated their responses with an MR-compatible button box
equipped with 4 buttons.
Stimuli were normal and abnormal chest radiographs obtained from
a commercially available CD published by the Japanese Society of
Radiological Technology (JSRT) in cooperation with the Japanese
Radiological Society. Each abnormal radiograph contained a single lung
nodule, a potentially malignant round lesion located in the lung field.
The abnormal radiographs were further divided by the JSRT publishers
into 3 difficulty levels. An expert thoracic radiologist independently
evaluated each image and excluded images with the most difficult-to-
detect nodules and those of poor image quality. This resulted in
a usable set of 92 normal and 100 abnormal radiographs.
Two versions of each radiographic image were used in the main part
of the experiment: intact and scrambled (Fig. 1C). The 2 radiograph
types, normal and abnormal, were crossed with the 2 image types,
intact and scrambled, to yield 4 total stimulus conditions. Scrambled
radiographs were created by dividing each image into 25 squares and
randomly shuffling the location of all but one of those squares. The one
square that remained in its original location in each image was, for the
abnormal radiographs, the square that contained the lung nodule and,
for the normal radiographs, the square that was cued as a possible
nodule location. Cue locations for the normal radiographs were
matched to actual nodule locations in the abnormal radiographs. To
ensure that each cue was a plausible nodule location, cue location
assignment for the normal radiographs was determined by a postresi-
dency radiologist at UCLA Medical Center who did not participate as
a subject in the study.
Because scrambling the radiographs introduced vertical and hori-
zontal lines in the images, grids were drawn on all stimuli, both intact
and scrambled. Each lung nodule was entirely contained within a square
and never obscured by the grid. For sample intact and scrambled
nodule-containing radiographs, see Figure 1. All images were square
and subtended a visual angle of approximately 19 degrees.
Each participant completed one 80-min scanning session followed by
an expertise posttest completed outside of the scanner. The scanning
session contained several anatomical scans, 3 localizer scans, and 3
rapid event-related scans. Localizer scans were used to define regions
of interest. Examples of images used in the localizer scans are shown in
To define visual cortical areas selective for processing faces, partic-
ipants viewed 8 blocks of faces in alternation with 8 blocks of objects.
Note that rest or fixation blocks were not placed between the blocks of
faces and objects. Nine images were shown per stimulus block; 72 faces
and 72 objects were displayed in total. Each image was presented for
1.7 s with 0.3-s interstimulus intervals yielding a block duration of 18 s.
To control for attention, participants performed a 1-back task. On one
in every 9 stimulus presentations, on average, the same image was
shown on successive trials. Participants were instructed to press
a button when an image repeated in this fashion. Performance on the
one back task was 81% correct for novices and 85% correct for experts;
these means did not differ reliably.
To identify regions of object selective cortex, including LO, that
might be specialized for processing radiographs, participants viewed
Figure 1. Experimental methods. (A) Example face and object used in face localizer scan and sample results from one participant. Arrow indicates the FFA; voxels were
thresholded at r [ 0.25. (B) Example images used in radiograph localizer scan and results from one participant. Arrow indicates left LO; voxels were again thresholded at
r [ 0.25. (C) Upper: example intact and scrambled radiographs used in the diagnosis scan; arrow (not seen by subject) indicates location of lung nodule. Lower: stimulus
sequence in diagnosis scans: 500 ms stimulus presentations were preceded by 1000 ms and followed by 1500 ms fixation periods.
Cerebral Cortex November 2009, V 19 N 11 2747
8 blocks of intact radiographs in alternation with 8 blocks of scrambled
radiographs. Note again, that rest or fixation blocks were not placed
between the blocks of intact and scrambled radiographs. Scrambled
radiographs for the localizer scan were finely scrambled; each image
contained 100 equal-size squares (Figure 1B). Nine images were shown
per stimulus block; 32 unique radiographs were shown during the
localizer scan, repeated as necessary to yield a total of 72 intact
presentations and 72 scrambled presentations. Each image was
presented for 1.7 s with 0.3-s interstimulus intervals yielding a block
duration of 18 s. Radiographs shown in the localizer scan were not
shown in any event-related diagnosis scans. To control for attention,
participants performed a 1-back task as described above for the face
localizer scan (note that participants did not perform a diagnosis task on
these images). Performance on the one back task was 96% correct for
novices and 99% correct for experts; these means did not differ reliably.
We also employed standard retinotopic mapping procedures (Engel
et al. 1994; Sereno et al. 1995; DeYoe et al. 1996) to identify retinotopic
visual areas (e.g., V1, V2, and V3). Participants fixated on a square
positioned in the center of the display. A wedge filled with a high
contrast temporally reversing checkerboard pattern rotated around
fixation, completing one rotation every 30 s. To keep participants
fixated and attending, the fixation square intermittently changed from
black to white or vice versa. Participants were instructed to press
a button as rapidly as possible each time fixation changed color.
The diagnosis scans were event-related scans in which participants
viewed intact and coarsely scrambled radiographs and judged whether
a cued region in each radiograph contained a lung nodule (Fig. 1C). A
new trial occurred every 3.0 s. The sequence of a single trial was as
follows: 1) a grid the same size as the radiographs was displayed for
1000 ms. The grid contained a fixation cross in the unit of the grid that
the subject was to judge. Participants were instructed to move their
eyes to the fixation cross and then to remain fixated at that location for
the remainder of the trial. 2) The grid, but not the fixation cross,
remained on the screen and a radiograph was displayed for 500 ms. 3)
The radiograph was removed, but the grid and fixation remained for
Each scan consisted of 124 trials composed of 25 radiographs in each
of the 4 stimulus conditions: 1) intact, normal; 2) scrambled, normal; 3)
intact, abnormal; 4) scrambled, abnormal, and 24 rest trials in which no
radiograph was presented. Condition order was counterbalanced using
an m-sequence (Buracas and Boynton 2002).
Functional MRI data were acquired using a blood oxygen level--
dependent contrast-weighted echo-planar pulse sequence (3T; time
echo = 25; time repetition = 3 s; flip angle = 90; field of view = 20 3 20
cm; voxel size = 3.125 3 3.125 3 4 mm; 36 slices parallel to the
anterior-posterior commissure line). High-resolution conventional
anatomical images were acquired coplanar to the functional data, and
T1-weighted volumetric scans were acquired for cortical flattening.
fMRI Data Analysis
Localizer scans were analyzed by simple correlation of each voxel’s
time series with a sinusoid at the stimulus frequency. Because the face
localizer and radiograph localizer scans did not include rest blocks, this
correlation effectively measured the reliability of the difference
between activity in the 2 conditions (faces vs. objects or intact vs.
scrambled localizers). Active pixels were identified as those with
a correlation above a threshold whose activity was in-phase with the
stimulus. Different threshold levels were tested for area identification;
the overall pattern of results did not depend upon threshold.
Event-related scans were analyzed using the general linear model and
the ‘‘Finite Impulse Response’’ approach. The design matrix contained
regressors for each time point in the response for each condition.
Estimates for each regressor were computed using ordinary least squares
from the average timecourse of active voxels in each visual area. The
amplitude of the response for each condition was computed as the peak
of the estimated impulse response, excluding the first and last time
points. This peak was computed independently for each subject, and so
the time point at which it occurred varied slightly between subjects.
Other estimates of response amplitude (e.g., area under the curve and fit
model hemodynamic response) yielded similar results.
Following standard techniques, the FFA was identified as voxels in the
mid-fusiform gyrus that were active in the face localizer scan
(Kanwisher et al. 1997). The occipital face area (OFA) was identified
as voxels on the posterior lateral surface of the occipital lobe that were
active in the face localizer scan. Areas on the lateral aspect of the
occipital lobe that were consistently active in the radiograph localizer
scan were defined as LO. Additional contiguous areas on the ventral
surface of the occipital lobe were also identified; this region
corresponds to the posterior fusiform gyrus and has been annotated
pFus as in other studies. Note that the regions in LO and pFus identified
in this way correspond to only a subset of the larger lateral occipital
complex that is often identified by comparing images of other types of
objects to textures (Malach et al. 1995).
Correlating Behavior and Neural Activity
We computed Pearson correlation coefficients between participants’
estimated response amplitudes and behavioral performance on the
diagnosis task. Corresponding Student’s t values were also computed.
To test whether our results were due to low-performing subjects, we
removed subjects from the analysis when their performance was below
a d# of 0.75. Two subjects were below this threshold for scrambled
radiographs only; these data were removed from the correlation
analysis and only data for intact radiographs (both activity and
performance) were used. One subject was below d# of 0.75 for both
intact and scrambled objects, and their data were completely removed
from the analysis.
To compare correlations between groups, we conducted a random-
ization analysis in which participants were assigned to groups at
random, and the correlation between behavior and FFA activity was
computed for each group. The difference between group correlations
was then computed. This randomization was repeated 1000 times,
yielding a null distribution of group correlation differences from which
p values were computed.
General Expertise Posttest
To provide a measure of general radiology expertise, each participant
completed a posttest outside of the scanner. The posttest consisted of
15 chest radiographs selected from radiology board certification
training images. Test items ranged in degree of diagnosis difficulty,
for example, pneumopericardium, aortic aneurysm, and mitral valve
calcification. Participants were allowed to view each image freely under
no time constraints and were asked to provide a written diagnosis for
each. An expert radiologist at UCLA Medical Center who was not
a participant in the study scored expertise posttests.
Practicing radiologists and fourth-year residents performed
better than first-year residents on the test of general radiological
expertise, conducted after scanning (see Materials and Meth-
ods). Group average scores were 86.67%, 70.00%, and 38.10%
correct, respectively. Four years of intense training appears to
have allowed the fourth-year residents to approach knowledge
levels of the practicing radiologists. Because the difference
between fourth-year residents and practicing radiologists was
relatively small, we combined these 2 into one group, called
‘‘experts.’’ Experts reliably outperformed novices on the test of
general radiological expertise (t = 6.96, p < 0.001).
Visual expertise, measured using the diagnosis task performed
in the scanner, showed a similar pattern (see Fig. 2). Practicing
and fourth-year radiologists’ performance was again relatively
close and above that of first-year radiologists (d# = 1.26, 1.19,
and 0.97 for the 3 groups, respectively). The difference between
experts and novices was reliable (t = 2.1, p < 0.03).
Neural Bases of Radiological Expertise
Harley et al.
On average, performance on the diagnosis task was lower
when radiographs were scrambled compared with intact; this
scrambling effect was small but reliable (d# = 1.05, 1.23 for
scrambled and intact, F1,17= –13.49, p < 0.01). Scrambling the
images did not affect the groups differently (F2,27= 0.02), and so
most analyses combine data from intact and scrambled trials.
Fusiform Face Area
A discrete region of activity in the right fusiform gyrus was
identified in all subjects (see Fig. 1 and Table 1). Corresponding
activity in the left hemisphere was evident in all but 2 subjects,
and this region was on average much smaller than in the right
hemisphere. During the radiograph diagnosis scans, overall
activity was moderate in the right FFA (hereafter, simply FFA)
and did not differ reliably between novices and experts (Fig. 3).
To examine in detail the relationship between visual expertise
and neural activity, we correlated performance during diagnosis
scans with the amplitude of the FFA response, across partic-
ipants. The fMRI response amplitude (hereafter referred to as
activity) in the FFA correlated reliably with visual expertise as
measuredbydiagnosisperformance(r = 0.55,p < 0.01).Figure3
shows the scatter plot of FFA activity against expertise.
The correlation between visual expertise and FFA activity
was stable. It was evident for correlations conducted separately
on intact and scrambled images (r = 0.49, 0.54). The correla-
tion was also not due to subject age. To test this, we first
correlated performance with age and then correlated the
residuals from this regression with FFA activity. The FFA
activity correlated reliably with the residuals (r = 0.56,
p < 0.01). The correlation was additionally not explainable by
general responsiveness of the FFA. Regressing out activity in
the FFA localizer scans revealed a reliable residual correlation
with performance (r = 0.55, p < 0.01).
Because our measure of expertise was performance during
scanning, it was important to verify our results did not simply
reflect experts’ greater success at the task. To control for this,
the radiographs were chosen from 3 levels of difficulty
determined a priori. For each subject, we examined FFA
activity for the set of trials at the single level of difficulty where
performance was closest to d# = 1.25. For experts, this tended
to be the more difficult levels, whereas for novices, it tended to
be the easier levels. Even for these trials, where the groups’
success at the task was matched (average d# = 1.1 for novices
and 1.09 for experts), activity in the FFA correlated with our
behavioral measure of visual expertise, overall performance in
the scanner (r = –0.50, p < 0.02).
Three subjects—2 experts and 1 novice—performed rela-
tively poorly on the diagnosis task. This may have been due to
difficulty with the speeded task, the visual display apparatus,
or the scanner environment, all of which differed from tradi-
tional diagnosis. Regardless, the relationship between FFA
activity and performance was not due to these potential
outliers, as can be seen by inspecting Figure 3B. Removing data
Figure 2. Average performance (d#) on nodule detection task in intact and
scrambled radiographs for 3 subject group expertise levels: practicing radiologists,
fourth-year residents (Y4), and first-year residents (Y1). Error bars are ±1 standard
error of the mean.
Average ROI sizes (mm3)
(n 5 7)
(n 5 6)
(n 5 7)
Note: Entries are the average ROI sizes for each group. The numbers in parenthesis indicate
the number of participants from each group for whom no ROI could be identified; these
participants were not included in the averages. Except for the left LO, ROI sizes did not differ
reliably between groups.
MR Signal (% change)
FFA Activity (% change)
Figure 3. Activity in the FFA during diagnosis scans. (A) Average fMRI time courses
for each group; the expert group consisted of both fourth-year residents and
practicing radiologists. Error bars indicate one standard error of the mean. (B) Scatter
plot of the amplitude of fMRI response with performance of the diagnosis task during
scanning for first years and experts. Colors are as in subplot (A).
Cerebral Cortex November 2009, V 19 N 11 2749
when performance was below a threshold d# of 0.75 still
yielded a reliable correlation (r = 0.47, p < 0.02).
Responses in the FFA were not reliably affected by stimulus
condition. Activity did not reliably differ between images that
contained nodules and those that did not. It also did not differ
reliably between intact and scrambled radiographs. This
pattern held for all ROIs examined.
Activity in the left FFA during the diagnosis scans was weak
and did not differ between groups. It also did not correlate with
Occipital Face Area
The OFA was identifiable reliably only in the right hemispheres
of our subject population. Weak activity was seen in the left
hemisphere generally, but clear ROIs were identifiable in less
than half of our participants. The right OFA was more active in
first-year residents than in experts during the diagnosis scans
(F1,16= 5.99, p < 0.05). This activity did not correlate with
expertise (r = –0.08).
LO and Posterior Fusiform Gyrus
All participants showed activity in LO during the radiograph
localizer scan (Fig. 1; Table 1). We identified separable foci of
activity in LO and the posterior fusiform gyrus and analyzed
the data separately for each region in each hemisphere. The
focus in left LO was larger in novices than in experts (Table 1;
F1,16= 5.88, p < 0.05) and showed a trend toward higher
activity in novices than in experts in the diagnosis scans
(F1,16= 2.8, p < 0.11; Fig. 4).
Activity in left LO showed a negative correlation with visual
expertise (Fig. 4; r = –0.53, p < 0.02). This correlation held for
both intact and scrambled stimuli (r = –0.57, –0.44). The correla-
tionalsocouldnotbeaccountedforbyage(r = –0.47onresiduals,
p < 0.05). The correlation also remained after removing potential
outlier data below a d# of 0.75 (r = –0.49, p < 0.05).
The negative correlation between left LO activity and visual
expertise was not due to general effort or attention. It is in
theory possible that subjects who performed better on the
task did so with less effort or attention, which in turn led to
reduced activity in left LO. Such an explanation, however,
would predict a similar pattern in other visual areas that are
known to be modulated by attention. Our findings of an
opposite pattern in FFA and no relationship between perfor-
mance and robust activity in right LO (see below) argue against
a general effort account.
Nevertheless, to rule out both effort and task success as an
account for our results, we again analyzed a subset of our data
that were matched for task performance. The data were chosen
using the 3 levels of difficulty assigned to the stimuli; for each
subject, we examined LO activity in the subset of trials at the
single level of difficulty where performance was closest to
d# = 1.0. Even for these trials, where presumably effort was
greater for novices than experts (average d# was 0.874 for
novices and 0.868 for experts), activity in left LO correlated
with visual expertise (r = –0.51, p < 0.02).
Although active left LO was larger in novices than in experts,
this difference did not account for the observed correlation
between expertise and activity. A control analysis that used
data only from the most active voxels in each subject still found
a reliable correlation between expertise and performance
(r = –0.50, p < 0.02; 40 most active voxels).
Activity in left LO was also negatively correlated with activity
in the FFA (Fig. 5; r = –0.47, p < 0.05). Subjects in which LO
activity was relatively high showed FFA activity that was
relatively low and vice versa.
Left pFus showed a weak negative correlation with visual
expertise (r = –0.18, n.s.). Right LO and pFus showed no
correlation with expertise (r = 0.10, –0.08, respectively). None
of these areas showed overall activity differences between
novices and experts.
Retinotopic Visual Areas
Retinotopic areas did not show reliable correlations between
activity and expertise; however, they generally did show trends
for negative correlations and greater activity in novices than in
experts during the diagnosis task. This latter difference was
reliable only in V1 (F1,18= 4.475, p < 0.05). V1 activity did not
account for the negative correlation found in left LO (r = –0.51,
p < 0.02 on residuals).
Our results reveal a striking double dissociation between
radiological visual expertise and cortical area activation.
Activity in the FFA correlated positively with expertise,
whereas activity in left LO correlated negatively. One in-
terpretation of this pattern is that the acquisition of expertise
MR Signal (% change)
Left LO Activity(% change)
Figure 4. Activity in left LO during diagnosis scans. (A) Average fMRI time courses
for each group; error bars indicate one standard error of the mean. (B) Scatter plot of
activity with performance of the diagnosis task during scanning for first years and
experts. Colors are as in subplot (A).
Neural Bases of Radiological Expertise
Harley et al.
involves suppressing preexisting neural representations in LO
as well as developing new ones in the FFA.
Expert Object Processing in LO and Fusiform Cortex
Our results agree with and extend prior studies of object-
specific cortex. LO appears to contain a general representation
of object shape; short-term adaptation there depends upon
shape (Kourtzi and Kanwisher 2001), and LO responses are
greater for closed shapes than for visual surfaces (Vinberg and
Grill-Spector 2008). The spatial pattern of activity in LO
correlates with physical shape (Haushofer et al. 2008) and
distinguishes between members of an object category that
differ in shape (Eger, Ashburner, et al. 2008). Consistent with
this role, training to discriminate complex shapes increases the
overall level of activity in LO, changes its spatial pattern, and
narrows its tuning but does not lead to the development of
category-specific regions of cortex (Grill-Spector et al. 2000;
Op de Beeck et al. 2006; Yue et al. 2006; Jiang et al. 2007).
Perhaps surprisingly, we found that both the size of and
activity level in the part of left LO most responsive to
radiographs were negatively correlated with expressed exper-
tise. One prior study of expertise also reported greater activity in
LO for novices than for experts (Gauthier et al. 2000). Our
results extend this work to show a continuous correlation
between LO activity and expertise. In addition, our results show
that the negative correlation with expertise arises in the part of
LO that is most active to radiographs generally. The simplest
interpretation of our data is that the left LO’s general
representation of shape is less important for diagnosing radio-
graphs in expert radiologists than it is for novices.
Neural representations of objects in more ventral cortex
differ from those in LO in a number of ways. Responses in these
regions are more invariant with respect to image size and
image location and integrate information across a larger
portion of objects (Grill-Spector et al. 1999; Lerner et al.
2001, 2002; Eger, Kell, and Kleinschmidt 2008). Responses also
correlate better with subjective similarity of shapes than
physical similarity (Haushofer et al. 2008). In addition, neurons
in ventral cortex are closely tied to identification of objects in
familiar categories (Gauthier et al. 1997; Bar et al. 2001; Grill-
Spector et al. 2004). Areas specialized for the particular
categories of letter strings, faces, and body parts all have been
found anterior to LO (Downing et al. 2001; McCandliss et al.
2003; Rhodes et al. 2004).
Our data show a positive correlation of activity in ventral
cortex during diagnosis as a function of expertise in radiology.
These findings agree with prior findings that experts in cars,
birds, and butterflies all show increased activation in ventral
cortex for the objects of their expertise (Gauthier et al. 2000;
Rhodes et al. 2004; Xu 2005, but see Grill-Spector et al. 2004).
The simplest interpretation of our data is that more expert
radiologists make use of specialized neural representations in
anterior ventral visual cortex that are engaged in the diagnosis
Interactions between LO and Fusiform Cortex
One novel aspect of our results is the negative correlation
between activity in the FFA and left LO, which suggests that
training in radiology alters the interaction between cortical
areas. Consistent with this idea, the negative correlation
between activity in LO and the FFA was larger in experts than
in novices and only reliable in experts (Fig. 5; for experts r =
–0.71, p < 0.01; for novices r = 0.35, n.s.). To compare directly
the group correlations, we conducted a randomization analysis
in which we repeatedly randomly assigned participants to
groups and recomputed the correlation between LO and FFA
activity (see Materials and Methods). The observed difference in
correlation between first-year residents and experts was reliable;
differences of that magnitude or greater were obtained in less
than 3.5% of randomized samples (n = 1000).
The negative correlation of LO and FFA activity may indicate
a competitive interaction between different neural representa-
tions engaged in the task, as has been proposed for memory
systems (e.g., Poldrack and Packard 2003). It is possible, for
example, that left LO contains a more parts-based representa-
tion that is suppressed as a more ‘‘holistic,’’ ‘‘global,’’ or
‘‘configural’’ representation develops in the FFA (Gauthier
et al. 1999; 2000; Lux et al. 2004; Busey and Vanderkolk
2005). Similarly, the left hemisphere may contain viewpoint
invariant representations (Vuilleumier et al. 2002) that are less
useful for diagnosis than viewpoint dependent ones in the right
Alternatively, the negative correlation could reflect ‘‘explain-
ing away,’’ in which higher level representations suppress
corresponding lower level representations in order to aid
processing of the remaining image regions (Murray et al. 2002).
For example, the fusiform region may subserve recognition of
normal anatomical features in the radiograph and suppress
lower level representations of them in LO. This in turn would
leave abnormal features of the radiograph isolated in LO, which
could aid their identification.
Origins of the Correlations between Activity and Expertise
The correlations we observed between expressed expertise
and neural activity (Figs 3B and 4B) have 2 possible origins.
They could reflect the hypothesis that 1) as subjects became
expert in radiology, mean activity levels during diagnosis
increased in the FFA and decreased in left LO. Under this
interpretation, training affected the mean level of activity in
each region. Inspecting the figures closely, however, also
reveals stronger correlations for experts than for novices. Thus,
our results could also reflect the hypothesis that 2) as subjects
0 0.2 0.40.6
Left LO Activity (% change)
FFA Activity (% change)
Figure 5. Scatterplot of activity in FFA versus activity in left LO during diagnosis
scans. Colors are as in previous figures.
Cerebral Cortex November 2009, V 19 N 11 2751
became expert in radiology, activity became more correlated
with expressed expertise in the FFA and more anticorrelated in
left LO. Under this interpretation, training affected the
relationship between expressed expertise and activity. Note
that these 2 hypotheses are not mutually exclusive. Evaluating
both hypotheses requires evaluating effects of training, rather
than expressed expertise, on neural activity.
The first hypothesis predicts that training groups should
differ in their mean levels of activity in the 2 regions, but our
results did not show reliable differences in overall FFA or LO
activity. This pattern has 2 likely causes. First, the novices in
our study were already on their way to becoming expert,
which could have moved some into the expert range. More
critically, our measure of visual expertise was performance on
the diagnosis task while in the scanner. The slightly unnatural
nature of the task may have prevented some subjects from fully
expressing their expertise, including the 3 subjects that were
outliers in performance. Subjects with higher expressed
expertise, as measured by performance on the diagnosis task,
did show reliably higher activity in the FFA (t = 2.3, p < 0.05
for comparison of 10 highest performers to remainders) and
a strong trend toward the opposite pattern in left LO (t = 2.1,
p < 0.06). Additionally, removing outlier performance data
yielded a trend toward greater activity for novices than experts
in LO (t = 1.7, p < 0.11), though not the opposite trend in FFA
(t = 0.55; n.s.). Given these suggestive patterns, it seems
plausible that group differences in mean activity were at least
one source of the correlations we observed between expertise
and neural activity. But, because our data do not show overall
group differences in neural activity, we cannot rule out that
mean activity in the FFA and LO was unaffected by training.
The second hypothesis, that training affects the correlation
between activity and expertise, predicts that the expert group
of subjects should show higher correlations than the novices.
Visually inspecting the data suggested that this may be the case
and computing correlations coefficients separately for each
group yielded higher values for experts (FFA, r = 0.77; left LO,
r = –0.68) than for novices (FFA, r = –0.10; left LO, r = 0.61). To
test this more formally, we conducted additional randomization
analyses. The differences in correlation between first-year
residents and experts were reliable; for both regions, differ-
ences of that magnitude or greater were obtained in less than
3% of randomized samples (n = 1000). These results should be
interpreted cautiously, however, because of the small sample
size and limited variance of the novice group as well as the
presence of potential outlier subjects in both groups. Future
work with larger sample sizes can overcome these limitations
and also examine potential differences between fourth-year
residents and practicing radiologists.
Nevertheless, the current results are consistent with the
hypothesis that training increased the correlation between
expressed expertise and FFA activity and decreased the
correlation between it and activity in left LO. These differences
in correlation between training groups have several possible
functional interpretations. One possibility is that the greater
correlation for experts was simply due to greater effort and/or
success at the task. This explanation was ruled out by our
analysis that examined neural activity for a subset of trials
on which performance was matched (see Results). In the
performance matched trials, success was equated by definition,
and novices were presumably putting forth even more effort
than experts because the task was more difficult for them.
Activity in both ROIs still correlated with overall expressed
expertise even for these trials.
Our data are more consistent with the interpretation that
training changed neural representations in FFA and LO that
were important for the diagnosis task. These learning effects
were stronger in some subjects than others, which produced
the observed correlation between activity and expressed
expertise in experts. The training-induced changes could have
been ‘‘bottom-up,’’ reflecting relatively permanent alterations in
receptive fields in FFA and LO, though the low performance of
some experts suggests that external factors were able to
influence their expression. Alternatively, the learning could
have occurred primarily in higher cortical regions and affected
FFA and LO activity through ‘‘top-down’’ mechanisms. For
example, experts may have learned a task strategy, such as
focusing attention on particular learned feature combinations
in the image. Such learning could also vary in strength between
subjects and could have been disrupted by external factors;
perhaps, the low-performing experts adopted a different
strategy under the unusual conditions in the scanner. Our
present data do not allow us to distinguish between top-down
and bottom-up effects of learning, but they do identify the FFA
and left LO as important targets of training.
The FFA and Expert Visual Processing
The overall questions of whether visual expertise for objects
other than faces depends upon the FFA remain quite
controversial. Differences in methods and results have left
room for widely disparate overall interpretations of the
literature (Bukach et al. 2006; McKone et al. 2007). Addressing
this debate was not the primary goal of the present study, but
its results are nevertheless relevant. Four previous studies have
tested whether people with many years of expertise differ in
their pattern of FFA activity from novices: Two found greater
relative activity in the FFA for experts, as well as correlations
between expertise and activity (Gauthier et al. 2000; Xu 2005),
one found a nonsignificant trend for greater relative activity for
experts (Rhodes et al. 2004), and one found no difference
between experts and novices (Grill-Spector et al. 2004). Our
results support the idea that the FFA is important for visual
expertise for stimuli other than faces. Our study does have
limitations, however. We (like all previous fMRI studies) cannot
rule out that the expertise effects we see in the FFA are not
instead due to spread of activity from some very nearby area
that, for example, is specialized for body parts (Peelen and
Downing 2005; Schwarzlose et al. 2005). Similarly, this study
cannot determine whether the same neurons within the FFA
are responding to both radiographs and faces.
It is also clear from prior work that not all expert visual
processing depends upon the FFA (e.g., processing of letters and
words). Determining when the FFA is involved in expert visual
processing remains a challenge. Our scrambled stimuli were
one attempt to do so; they were intended to disrupt ‘‘holistic’’
processing in which the FFA might play a selective role.
Unfortunately, our scrambling manipulation had only a small
effect on behavior and no reliable effect on FFA activity. It seems
likely that the scrambling, which left relatively large-scale
anatomical features intact, was too coarse to provide a strong
test of the holistic hypothesis. Consistent with this idea, the
finely scrambled images in the radiograph localizer scan did
generate less activity in the FFA than the intact images (p < 0.05;
Neural Bases of Radiological Expertise
Harley et al.
tested by measuring fit of sinusoid in-phase with the intact
The Role of LO and the FFA in Radiograph Diagnosis
Diagnosis of chest radiographs is a complex task, one that
certainly depends upon specialized processing in other cortical
regions beyond those localized here. This initial study focused
on processing of the entire radiograph; we did not, for
example, run a ‘‘nodule localizer’’ scan to identify regions
containing neurons sensitive to the small variations in shape
that indicate potential tumors and other abnormalities in
medical images. The diagnosis task may depend heavily upon
such areas, and whether they lie near the FFA or in LO (or both
or neither) remains unknown. Our results do nevertheless
constrain possible models of the neural bases of radiological
expertise. What follows is one speculative account that is
consistent with our data.
processing of anatomical features of radiographs. Activity in the
right FFA could signal recognition of features that are common or
normal, and in experts may cause suppression of a shape-based
representation in left LO. This suppression in turn may aid task
performance by reducing the strength of representations that
are less useful for the task of identifying abnormalities. Visual
expertise, then, might involve not only the growth of new
knowledge but also the suppression of older forms of knowledge.
National Institutes of Health (EY11862, RR12169, RR13642,
and RR00865) and the UCLA Departments of Radiological
Sciences and Neurology.
The authors thank Yuhong Jiang for helpful comments. For generous
support the authors also wish to thank the Brain Mapping Medical
Research Organization, Brain Mapping Support Foundation, Pierson-
Lovelace Foundation, The Ahmanson Foundation, William M. and Linda
R. Dietel Philanthropic Fund at the Northern Piedmont Community
Foundation, Tamkin Foundation, Jennifer Jones-Simon Foundation,
Capital Group Companies Charitable Foundation, Robson Family, and
Northstar Fund. Conflict of Interest: None declared.
Address correspondence to Stephen Engel, Department of Psychol-
ogy, University of Minnesota, N218 Elliott Hall, 75 East River Road,
Minneapolis, MN, USA. Email: firstname.lastname@example.org.
Bar M, Tootell RB, Schacter DL, Greve DN, Fischl B, Mendola JD,
Rosen BR, Dale AM. 2001. Cortical mechanisms specific to explicit
visual object recognition. Neuron. 29:529--535.
Bukach CM, Gauthier I, Tarr MJ. 2006. Beyond faces and modularity: the
power of an expertise framework. Trends Cogn Sci. 10:159--166.
Buracas GT, Boynton GM. 2002. Efficient design of event-related fMRI
experiments using M-sequences. Neuroimage. 16:801--813.
Busey TA, Vanderkolk JR. 2005. Behavioral and electrophysiological
evidence for configural processing in fingerprint experts. Vision
DeYoe EA, Carman GJ, Bandettini P, Glickman S, Wieser J, Cox R,
Miller D, Neitz J. 1996. Mapping striate and extrastriate visual areas
in human cerebral cortex. Proc Natl Acad Sci USA. 93:2382--2386.
Downing PE, Jiang Y, Shuman M, Kanwisher N. 2001. A cortical area
selective for visual processing of the human body. Science. 293:
Eger E, Ashburner J, Haynes JD, Dolan RJ, Rees G. 2008. fMRI activity
patterns in human LO carry information about object exemplars
within category. J Cogn Neurosci. 20:356--370.
Eger E, Kell CA, Kleinschmidt A. 2008. Graded size-sensitivity of object-
exemplar-evoked activity patterns within human LO subregions.
J Neurophysiol. 4:2038--2047.
EngelSA, RumelhartDE, Wandell
Chichilnisky EJ, Shadlen MN. 1994. fMRI of human visual cortex.
Gauthier I, Anderson AW, Tarr MJ, Skudlarski P, Gore JC. 1997. Levels of
categorization in visual recognition studied using functional
magnetic resonance imaging. Curr Biol. 7:645--651.
Gauthier I, Skudlarski P, Gore JC, Anderson AW. 2000. Expertise for cars
and birds recruits brain areas involved in face recognition. Nat
Gauthier I, Tarr MJ, Anderson AW, Skudlarski P, Gore JC. 1999.
Activation of the middle fusiform ‘face area’ increases with
expertise in recognizing novel objects. Nat Neurosci. 2:568--573.
Golarai G, Ghahremani DG, Whitfield-Gabrieli S, Reiss A, Eberhardt JL,
Gabrieli JD, Grill-Spector K. 2007. Differential development of high-
level visual cortex correlates with category-specific recognition
memory. Nat Neurosci. 10:512--522.
Grill-Spector K. 2003. The neural basis of object perception. Curr Opin
Grill-Spector K, Golarai G, Gabrieli J. 2008. Developmental neuro-
imaging of the human ventral visual cortex. Trends Cogn Sci. 12:
Grill-Spector K, Knouf N, Kanwisher N. 2004. The fusiform face area
subserves face perception, not generic within-category identifica-
tion. Nat Neurosci. 7:555--562.
Grill-Spector K, Kushnir T, Edelman S, Avidan G, Itzchak Y, Malach R.
1999. Differential processing of objects under various viewing
conditions in the human lateral occipital complex. Neuron.
Grill-Spector K, Kushnir T, Hendler T, Malach R. 2000. The dynamics of
object-selective activation correlate with recognition performance
in humans. Nat Neurosci. 3:837--843.
Haushofer J, Livingstone MS, Kanwisher N. 2008. Multivariate patterns
in object-selective cortex dissociate perceptual and physical shape
similarity. PLoS Biol. 6:e187.
Jiang X, Bradley E, Rini RA, Zeffiro T, Vanmeter J, Riesenhuber M. 2007.
Categorization training results in shape- and category-selective
human neural plasticity. Neuron. 53:891--903.
Kanwisher N, McDermott J, Chun MM. 1997. The fusiform face area:
a module in human extrastriate cortex specialized for face
perception. J Neurosci. 17:4302--4311.
Kourtzi Z, Kanwisher N. 2001. Representation of perceived object
shape by the human lateral occipital complex. Science. 293:
Lerner Y, Hendler T, Ben-Bashat D, Harel M, Malach R. 2001. A
hierarchical axis of object processing stages in the human visual
cortex. Cereb Cortex. 11:287--297.
Lerner Y, Hendler T, Malach R. 2002. Object-completion effects in the
human lateral occipital complex. Cereb Cortex. 12:163--177.
Lux S, Marshall JC, Ritzl A, Weiss PH, Pietrzyk U, Shah NJ, Zilles K,
Fink GR. 2004. A functional magnetic resonance imaging study of
local/global processing with stimulus presentation in the peripheral
visual hemifields. Neuroscience. 124:113--120.
Malach R, Reppas JB, Benson RR, Kwong KK, Jiang H, Kennedy WA,
Ledden PJ, Brady TJ, Rosen BR, Tootell RB. 1995. Object-related
activity revealed by functional magnetic resonance imaging
in human occipital cortex. Proc Natl Acad Sci USA. 92:
McCandliss BD, Cohen L, Dehaene S. 2003. The visual word form area:
expertise for reading in the fusiform gyrus. Trends Cogn Sci.
McKone E, Kanwisher N, Duchaine BC. 2007. Can generic expertise
explain special processing for faces? Trends Cogn Sci. 11:8--15.
Murray SO, Kersten D, Olshausen BA, Schrater P, Woods DL. 2002.
Shape perception reduces activity in human primary visual cortex.
Proc Natl Acad Sci USA. 99:15164--15169.
BA,Lee AT, Glover GH,
Cerebral Cortex November 2009, V 19 N 11 2753
Op de Beeck HP, Baker CI, DiCarlo JJ, Kanwisher NG. 2006.
Discrimination training alters object representations in human
extrastriate cortex. J Neurosci. 26:13025--13036.
Op de Beeck HP, Haushofer J, Kanwisher NG. 2008. Interpreting fMRI
data: maps, modules and dimensions. Nat Rev Neurosci. 9:123--135.
Peelen MV, Downing PE. 2005. Within-subject reproducibility of
category-specific visual activation with functional MRI. Hum Brain
Poldrack RA, Packard MG. 2003. Competition among multiple memory
systems: converging evidence from animal and human brain studies.
Rhodes G, Byatt G, Michie PT, Puce A. 2004. Is the fusiform face area
specialized for faces, individuation, or expert individuation? J Cogn
Scherf KS, Behrmann M, Humphreys K, Luna B. 2007. Visual category-
selectivity for faces, places and objects emerges along different
developmental trajectories. Dev Sci. 10:F15--F30.
Schwarzlose RF, Baker CI, Kanwisher N. 2005. Separate face and body
selectivity on the fusiform gyrus. J Neurosci. 25:11055--11059.
Sereno MI, Dale AM, Reppas JB, Kwong KK, Belliveau JW, Brady TJ,
Rosen BR, Tootell RB. 1995. Borders of multiple visual areas in
humans revealed by functional magnetic resonance imaging [see
comments]. Science. 268:889--893.
Sowden PT, Davies IR, Roling P. 2000. Perceptual learning of the
detection of features in X-ray images: a functional role for
improvements in adults’ visual sensitivity? J Exp Psychol Hum
Percept Perform. 26:379--390.
Vinberg J, Grill-Spector K. 2008. Representation of shapes, edges,
and surfaces across multiple cues in the human visual cortex.
J Neurophysiol. 99:1380--1393.
Vuilleumier P, Henson RN, Driver J, Dolan RJ. 2002. Multiple levels of
visual object constancy revealed by event-related fMRI of repetition
priming. Nat Neurosci. 5:491--499.
Xu Y. 2005. Revisiting the role of the fusiform face area in visual
expertise. Cereb Cortex. 15:1234--1242.
Yue X, Tjan BS, Biederman I. 2006. What makes faces special? Vision
Neural Bases of Radiological Expertise
Harley et al.