Brain (2001), 124, 739–756
Altered brain functional connectivity and impaired short-term memory in Alzheimer’s disease Cheryl L. Grady,1 Maura L. Furey,2 Pietro Pietrini,5 Barry Horwitz3 and Stanley I. Rapoport4 1Rotman
Research Institute, Baycrest Centre for Geriatric Care, University of Toronto, Toronto, Ontario, Canada, 2Laboratory of Brain and Cognition, National Institute of Mental Health, 3Language Section, Voice, Speech and Language Branch, National Institute on Deafness and Other Communication Disorders, 4Section on Brain Physiology and Metabolism, National Institute on Aging, Bethesda, MD, USA and 5Department of Human and Environmental Sciences, University of Pisa, Italy
Correspondence to: Cheryl L. Grady, Rotman Research Institute, Baycrest Centre for Geriatric Care, 3560 Bathurst Street, Toronto, Ontario, Canada M6A 2E1 E-mail:
[email protected]
Summary To examine functional interactions between prefrontal and medial temporal brain areas during face memory, blood flow was measured in patients with Alzheimer’s disease and healthy controls using PET. We hypothesized that controls would show correlated activity between frontal and posterior brain areas, including the medial temporal cortex, whereas patients would not, although frontal activity per se might be spared or even increased compared with controls. We used a delayed match to sample paradigm with delays from 1 to 16 s. There was no change in recognition accuracy with increasing delay in controls, whereas patients showed impaired recognition over all delays that worsened as delay increased. Controls showed increased activity in the bilateral prefrontal and parietal cortex with increasing delay, whereas the patients had increased activity in the right prefrontal, anterior cingulate and left amygdala. Increased activity in the right prefrontal cortex was associated with better memory performance in both groups and activity in the left amygdala was correlated with better performance in the patients. Based on these task and behavioural effects, we examined functional connectivity of the right prefrontal cortex and left amygdala in both groups by determining those areas whose activity was correlated with activity in
these regions. In controls, activity in the right prefrontal cortex was positively correlated with blood flow in the left prefrontal cortex, bilateral extrastriate and parietal areas and the right hippocampus. In patients, activity in the right prefrontal cortex was correlated mainly with other prefrontal regions. Areas where activity was correlated with the left amygdala in patients included the bilateral posterior parahippocampal gyri, a number of left prefrontal regions, anterior and posterior cingulate, thalamus, and insula. Controls had a relatively restricted set of regions where activity correlated with the left amygdala, mainly temporal and occipital areas. These results support the idea of a functional disconnection between the prefrontal cortex and the hippocampus in Alzheimer’s disease and suggest that memory breakdown in early Alzheimer’s disease is related to a reduction in the integrated activity within a distributed network that includes these two areas. The unexpected finding of increased involvement of the amygdala suggests that the patients may have processed the emotional content of the faces to a greater degree than did the controls. Furthermore, the positive association between amygdala activity and memory performance in the patients suggests a possible compensatory role for an emotion-related network of regions.
Keywords: neuroimaging; dementia; face recognition; PET; memory Abbreviations: LV ⫽ latent variable; rCBF ⫽ regional cerebral blood flow
Introduction Dysfunction in episodic memory is one of the earliest and most devastating symptoms of Alzheimer’s disease (Grady et al., 1988; Price et al., 1993; Zec, 1993; Jacobs et al., 1995). Delayed memory is affected to a greater extent than © Oxford University Press 2001
is immediate memory, although the latter is clearly impaired (Moss et al., 1986; Hart et al., 1988; Welsh et al., 1991). Delayed memory is usually tested over a period of several minutes to an hour, but even delays of a few seconds are
740
C. L. Grady et al.
sufficient to cause a deficit in Alzheimer’s disease patients (Sahakian et al., 1988). This impairment of delayed memory in Alzheimer’s disease is consistent with the damage to medial temporal structures, including the hippocampus and entorhinal cortex, that is thought to occur early in the disease (Braak et al., 1993; Kemper, 1994). It is also consistent with the increasingly important role that the hippocampus plays in memory as the length of the retention interval increases (Squire, 1992). It has been shown that the volume of medial temporal regions is correlated with delayed but not immediate memory performance in Alzheimer’s disease patients (Kohler et al., 1998) and in non-demented elderly individuals (Golomb et al., 1994; Convit et al., 1997). Although these studies highlight the importance of memory and its dependence on medial temporal function early in the course of Alzheimer’s disease, the medial temporal areas are not the only brain regions that mediate memory function or that are damaged in Alzheimer’s disease. Widespread areas of cortex have been found to be involved in one or more types of memory in humans (for reviews see Grady, 1999; Cabeza and Nyberg, 2000) as well as in nonhuman primates (Fuster et al., 1985; Goldman-Rakic, 1990; Miller et al., 1991). Neuroimaging techniques, particularly PET, have shown that parietal and temporal cortices that are important for memory function are affected relatively early in the course of Alzheimer’s disease (Foster et al., 1984; Friedland et al., 1985; Haxby et al., 1985; Jagust et al., 1988; McGeer et al., 1990), whereas in many patients the frontal cortex is affected later in the course of the disease (Chase, 1987; Grady et al., 1990). This progression of brain abnormalities is reflected in the behavioural deficits that are seen, with problems of attentional function, presumably mediated by the frontal cortex, evident at more severe stages of the disease than temporally mediated memory deficits (Grady et al., 1988; Sahakian et al., 1990). These patterns of deficit suggest that memory function in early Alzheimer’s disease may be accompanied by a decrease of activity in the temporal cortex, both medial and lateral regions, but with relative sparing of prefrontal involvement. Functional neuroimaging studies of memory function in Alzheimer’s disease patients have provided some support for this idea, at least for verbal memory. In two studies of word recall, mildly demented patients showed increased activity in most of the areas activated in healthy elderly controls, although they failed to show activation in the hippocampus (Becker et al., 1996; Backman et al., 1999). However, the patients showed increased activity compared with controls in some regions of the prefrontal cortex, particularly in the left hemisphere. This additional increase in prefrontal activation in the patients was interpreted as a compensatory reallocation of cognitive resources during the memory task, although the patients were significantly impaired in performance on the word memory tasks. A study of overt verbal rehearsal of word lists also found that the prefrontal cortex was utilized to a greater degree in the patients compared with age-matched controls during the rehearsal
phase, although activity in other areas, such as the premotor cortex, was reduced (Woodard et al., 1998). This increased prefrontal activation was accompanied by normal rehearsal rates in the patients, but their ability to remember the words later was impaired. Thus, all of these experiments suggest that early in the course of Alzheimer’s disease, prefrontal activation is maintained and may even be increased above normal levels in some memory tasks. However, this redistribution of cognitive resources may not be sufficient to maintain performance at normal levels. Increased utilization of the frontal cortex also has been found in healthy elderly adults in comparison with young adults during memory tasks (Cabeza et al., 1997; Grady et al., 1998; Madden et al., 1999; Reuter-Lorenz et al., 2000). To date there have been no reports of how brain activity is altered during visual non-verbal memory tasks in Alzheimer’s disease patients. In the current experiment we were interested in examining the changes in brain function that accompany the decline in memory for faces in early Alzheimer’s disease as well as the effect of increasing delay between stimulus presentation and recognition. A focus of this work is the delineation of the functional interactions among brain areas during memory processing, i.e. the functional connectivity. The assessment of functional connectivity is based on the assumption that memory and other cognitive abilities are the result of the integrated activity in networks of regions, rather than activity in any one region in isolation (Friston et al., 1993; Horwitz, 1994; McIntosh and Gonzalez-Lima, 1994) and that these network interactions are disrupted by the neural degeneration of Alzheimer’s disease. We found previously that the face perception network was altered in Alzheimer’s disease patients such that the interaction between the posterior face processing regions and the frontal cortex was disrupted (Horwitz et al., 1995). In healthy older controls, activity in a region of the right prefrontal cortex was highly correlated with activity in occipitotemporal regions during the face perception task, whereas in the Alzheimer’s disease patients this frontal area was only correlated with other prefrontal regions, suggesting an alteration of frontally mediated processing even during perception. When a memory delay was introduced into this match-tosample task, both young and old healthy adults showed increased activity in prefrontal regions with increasing delay, although only the young adults showed an increase in hippocampal activity (Haxby et al., 1995; Grady et al., 1998). We hypothesized that Alzheimer’s disease patients would be impaired on this delayed face memory task, particularly as delay increased and would show increased activity in prefrontal cortex, but not in the hippocampal region. We also predicted that there would be decreased interactions between prefrontal and posterior regions of the brain as we had seen for face perception, including interactions with medial temporal areas. In the current experiment these predictions were confirmed in that the patients showed reduced functional interactions between right hemisphere prefrontal and medial temporal regions. An unexpected finding was that the patients
Brain activity during face memory utilized a group of regions not engaged by healthy elderly adults, which included the left amygdala, other limbic areas, and the left prefrontal cortex.
Methods Participants The participants in this experiment were 21 healthy elderly individuals (nine women and 12 men; mean age ⫽ 66.1 ⫾ 4.5 years, range 57–74 years) and 11 mildly demented patients with a diagnosis of Alzheimer type dementia, three with possible Alzheimer’s disease (based on their cognitive decline being limited to memory) and eight with probable Alzheimer’s disease (four women, seven men; mean age ⫽ 68.5 ⫾ 11 years, range 49–84 years; Mini-Mental State Examination ⫽ 24 ⫾ 4). Both groups had a mean of 16 years of education; the Full Scale IQ (Wechsler, 1955) of the control group was 129 ⫾ 7 and that of the Alzheimer’s disease group was 109 ⫾ 14. All participants were screened (Duara et al., 1984) to rule out any diseases (other than Alzheimer’s disease in the patients) or medications that might alter cerebral function, including those commonly used for the treatment of Alzheimer’s disease. Sixteen of the healthy controls were part of a previously published study of face memory in young and old adults (Grady et al., 1998). All patients and controls gave written informed consent to participate in this study which was approved by the ethics committee of the National Institute on Aging (USA).
PET scanning PET scans, with injections of 37.5 mCi of H215O each and separated by 12 min, were performed on all participants. Scans were performed on a Scanditronix PC2048–15B tomograph (Uppsala, Sweden), which has a reconstructed resolution of 6.5 mm in both transverse and axial planes. This tomograph allows 15 planes, separated by 6.5 mm (centre to centre), to be acquired simultaneously. Emission data were corrected for attenuation by means of a transmission scan obtained at the same levels as the emission scans. Head movement during the scans was minimized by using a thermoplastic mask that was molded to each subject’s head and attached to the scanner bed. Each task started 1 min prior to isotope injection and continued throughout the scanning period. Scanning began when the radioactive count rate in the brain reached a threshold value and continued for 4 min. Arterial blood sampling was initiated at the time of injection and continued throughout the scanning period. Cerebral blood flow was calculated using a rapid least squares method (Holden et al., 1981; Carson et al., 1987)
Stimuli and tasks All subjects performed a sensorimotor control task (presented at the beginning of the scanning session) and four delayed
741
match-to-sample tasks using faces as stimuli (with the order of conditions counterbalanced across subjects). The delay interposed between the presentation of the sample and choice faces was either 1, 6, 11 or 16 s (Haxby et al., 1995). All matching tasks utilized a forced-choice, two-alternative format and each stimulus array consisted of three squares, one on the top of the array and two on the bottom. Each trial in the delayed matching tasks began with the presentation of an unfamiliar face in the upper square of the stimulus array and ended with the presentation of two choice faces in the bottom squares of the array. One of the choice faces was the face that had been presented at the beginning of the trial and one was a distracter. The delay between sample and choice faces was filled with 0–3 blank stimulus arrays, in which all three squares contained smaller grey squares. All stimulus arrays were presented for 4 s, separated by a 1 s interval. The subject indicated which of the two choice faces was a correct match by pressing a button with either the right or left thumb, depending on whether the correct choice stimulus was on the right or left side. The control task consisted of a single stimulus that was presented repeatedly and subjects were instructed to alternate right and left hand responses to repeated presentations of the control stimulus. This stimulus consisted of a noise pattern (with visual complexity similar to the faces) that was placed in all three squares of the stimulus array. A Macintosh computer was used to control the presentation of the stimuli as well as record reaction times and accuracy. All subjects with less than 20/20 vision uncorrected wore glasses during the scan, either their usual corrective lenses, or lenses custom-made to correct the subject’s vision for the viewing distance used during the scanning session (55 cm).
Data analysis Reaction time and accuracy (percentage correct) were analysed with repeated measures ANOVAs (analysis of variance). These analyses were carried out on 19 controls and 10 Alzheimer’s disease patients (some behavioural data were missing due to technical problems). PET data were registered using AIR (Woods et al., 1992), spatially normalized to the Talairach and Tournoux atlas coordinate system (Talairach and Tournoux, 1988) and smoothed (to 10 mm) using SPM95 (Frackowiak and Friston, 1994). The regional cerebral blood flow (rCBF) values for each subject were converted to ratios (to global CBF) and three types of analysis were carried out on the data: (i) an examination of the effects of memory delay on rCBF; (ii) an examination of the correlation between brain activity and behaviour; and (iii) an analysis of functional connectivity. The first analysis examined the effect of the five task conditions (memory and control) on rCBF using a multivariate technique known as ‘partial least squares’ (McIntosh et al., 1996). ‘Partial least squares’ is a multivariate analysis that identifies groups of brain regions distributed over the entire brain that together covary with some aspect of the experimental design, in
742
C. L. Grady et al.
contrast to the more typically used univariate analysis that assesses the significance of each region separately. ‘Partial least squares’ examines the covariance between brain voxels and the experimental design such that a new set of variables is identified (so-called Latent Variables or LVs, similar to principal components) that optimally relate the two sets of measurements. Each LV extracted represents an experimental effect and identifies both the pattern of task differences and the brain voxels showing that pattern. A ‘brain score’ calculated for each subject in each condition for each LV is an indication of how strongly the brain activity pattern identified by each LV is expressed in that subject. Comparison of these scores across conditions shows the change in rCBF characterized by each LV. Statistical significance of each LV as a whole was assessed by means of a permutation test (Edgington, 1980; McIntosh et al., 1996) and the stability of each brain voxel’s contribution to the pattern seen on each LV was determined through bootstrap resampling (Efron and Tibshirani, 1986; Sampson et al., 1989). A reliable contribution was made by voxels whose weight on a given pattern was greater than twice the estimated standard error of that weight, which corresponds to 95% confidence limits (Sampson et al., 1989). Locations of the maxima for each reliable region (containing at least 20 contiguous voxels) are reported in terms of brain region, or gyrus, and Brodmann area as defined in the Talairach and Tournoux atlas. The analysis of the effects of the task on rCBF was carried out on controls and Alzheimer’s disease patients separately as well as on the two groups combined. The correlations between the LVs obtained from the combined-group analysis and those from the two within-group analyses were calculated to obtain an indication of whether one group contributed primarily to a specific combined-group pattern or if both groups contributed equally. Group ⫻ Task interactions were examined by contrasting the within-group brain scores from those LVs that correlated with each combined LV using an ANOVA. The separately determined scores were compared rather than the scores from the combined analyses because having Group entered into the design matrix as a variable biases these latter scores. It is important to note that since the ANOVAs were carried out on the brain scores, which reflect activity across the whole image, any between-group differences apply to the entire pattern of activity and not just to any single region. The brain-behaviour and functional connectivity analyses were also carried out using ‘partial least squares’ (McIntosh, 1999). The brain/behaviour analysis examined the correlations between the behavioural measures obtained during the four delay conditions and brain activity in all voxels. This was done with both behavioural measures in the patient group (accuracy and reaction time) and with reaction time only in the control group (correlations with accuracy could not be validly carried out in this group since at least half the controls performed at ceiling on each of the delay conditions, thus restricting the range of values). For the functional connectivity analysis, the correlations between rCBF in a reference region
Fig. 1 Performance measures for Alzheimer’s disease patients (open squares) and controls (filled circles) in the four delayed memory conditions. Values are mean ⫾ standard error.
and all other brain voxels were examined in the four delay conditions of patients and controls (see Results for regions used as the reference voxels). For both of these analyses, correlations were computed between the variable of interest and rCBF in all brain voxels within each of the delay conditions, and then these correlations were compared using ‘partial least squares’ across the conditions.
Results Memory performance The behavioural results from the four memory conditions are shown in Fig. 1. There were significant effects of delay [F(3, 84) ⫽ 15.1, P ⬍ 0.001] and group [F(1, 28) ⫽ 31.0, P ⬍ 0.001] on face recognition accuracy, with the patients showing impaired face memory overall. However, the Task ⫻ Group interaction was also significant [F(3, 84) ⫽ 7.5, P ⬍ 0.001]. Separate analyses on controls and patients revealed that the control group showed no significant effect of delay on recognition accuracy [F(3, 54) ⫽ 1.7, P ⬎ 0.10], whereas the patients did show a significant effect of delay [F(3, 27) ⫽ 8.2, P ⬍ 0.001]. There were significant effects
Brain activity during face memory of delay [F(3, 84) ⫽ 76.1, P ⬍ 0.001] and group [F(1, 28) ⫽ 13.3, P ⫽ 0.001] on reaction times as well, but the interaction of delay and group was not significant. Thus, only the Alzheimer’s disease patients showed a decline in recognition accuracy with increasing memory delay, but both groups had increased response times as delay increased.
Task effects on brain activity The controls showed two significant LVs related to the memory tasks (LV1, P ⬍ 0.0001; LV2, P ⬍ 0.025), the Alzheimer’s disease patients showed only one significant LV (P ⬍ 0.01) and the combined group analysis identified two significant LVs (both P values ⬍ 0.0001). The first combinedgroup LV (Fig. 2) was correlated with the first LV from the control subjects (r ⫽ 0.98) and with the second non-significant LV from the Alzheimer’s disease patients (r ⫽ 0.58). In addition, there was a significant Task ⫻ Group interaction on the scores [F(4, 120) ⫽ 9.3, P ⬍ 0.001]. The regions identified by this LV differentiated the short memory delay conditions (1 s and 6 s) from the control task in the healthy elderly adult group. Since this pattern characterized the shortest delay conditions in which the greatest number of faces was seen, it represented brain activity associated with both face processing and memory over very short delays. Increased activity during the short-delay conditions compared with the control task was found in bilateral areas of the ventral and dorsal extrastriate cortex, the prefrontal cortex (both ventral and dorsal regions), the parietal cortex and premotor regions (Table 1). The control task was accompanied by increased activity, relative to the 1 s and 6 s delay conditions, in perisylvian areas bilaterally and in the medial prefrontal cortex. It can be seen from Table 1 that this pattern was entirely due to the contribution from the control group, as none of the regions identified by this LV were reliably found in the Alzheimer’s disease patients. The second pattern (Fig. 3) was correlated with the second LV from the control subjects (r ⫽ 0.55) and with the significant LV from the Alzheimer’s disease patients (r ⫽ 0.89). There was also a significant Task ⫻ Group interaction on this LV [F(4, 120) ⫽ 12.0, P ⬍ 0.001]. As can be seen from these correlations and the reliability ratios shown in Table 2, the Alzheimer’s disease patients contributed more strongly to this pattern, which identified brain areas whose activity differentiated the longer delay conditions from the control condition and the 1 s delay (Fig. 3). The areas with reliably increased activity during the longer delays in the patients included the right prefrontal cortex (an inferior region and a dorsolateral one), the left amygdala, anterior cingulate and the cuneus (Table 2). Only one of these areas, the ventral prefrontal region, showed reliably increased activity in the control participants at the longer delays, although there was a tendency for increased activity in most of the regions. Interestingly, although not identified by this combined group LV, the within group analysis of the Alzheimer’s disease patients also showed a reliable increase
743
in the left hippocampus at the longer delays (see Fig. 3; X ⫽ –36, Y ⫽ –26, Z ⫽ –12; reliability ratio ⫽ 2.55) which was not reliable in the controls (ratio ⫽ 0.35). Areas with increased activity during low or no memory demand were in perisylvian regions and the posterior extrastriate cortex. The areas of maximal increase at the longer delays in the control group (identified from the within-group analysis) were in bilateral regions of anterior prefrontal cortex, a ventral anterior cingulate region and an area in the left temporal cortex (Fig. 3 and Table 2). The Alzheimer’s disease patients did not show reliable changes in these areas (Table 2).
Brain/behaviour correlations Each of the three brain/behaviour analyses (one in the controls and two in the patients) resulted in a single significant LV (all P values ⬍ 0.02) that identified a set of brain regions with similar correlations across the four delay conditions. In order to limit the focus of these analyses to those areas that showed both a task effect and a behavioural effect, we report here those maxima from the behavioural analyses that were located within 2 cm or less from regions showing an increase in activity at the longer delays. Table 3 shows the coordinates of the reliable areas that were identified by both the task analyses (taken from the within group analyses to best characterize each group) and the analyses of brain/behaviour correlations. In the control group there were two prefrontal areas and a left temporal region that showed both an increase in activity at the longer delays and a relation with reaction time, such that increased activity was correlated with a reduction in reaction time. In the Alzheimer’s disease patients the right cuneus and the left amygdala region showed a reliable contribution to the task analysis and both behavioural analyses (Table 3). Blood flow in these regions was increased in the delay conditions compared with the control task, was positively correlated with accuracy on the tasks (except for the 6 s condition) and negatively correlated with reaction time in the patients. An additional area, in the right ventral prefrontal cortex, showed an increase in activity with longer memory delays in the patients and a positive correlation with accuracy, but was not correlated with reaction time (Table 3).
Functional connectivity The regions chosen for the functional connectivity analysis were those from Table 3 that showed the largest task effects in each group. In the control group this was a right anterior prefrontal region and in the patients it was the left amygdala (shown in italics in Table 3). Thus, the connectivity analysis was carried out on regions where there was increased activity during the longer delay conditions and where this increase was associated with better performance on the tasks. Since both controls and patients showed areas in the right prefrontal cortex with this pattern, the specific prefrontal coordinates for each group were used in the analysis (see Table 3). The
744
C. L. Grady et al.
Fig. 2 Brain areas identified by the first task LV from the combined–group analysis are shown in (A) and from the first LV in the control group analysis in (B) (brain areas meeting the bootstrap criterion of reliability ratio (艌2.0). Areas are shown on a standard MRI beginning at –28 mm relative to the AC–PC (anterior commissure–posterior commissure) line (top left of image) and extending up to 28 mm above the AC–PC line (bottom right of each image). The graph on the right of the figure shows the mean brain scores for each condition on the combined group LV for controls and Alzheimer’s disease patients (AD). These scores are analogous to factor scores in a principal component analysis and each subject receives a score for each condition. Brain regions shown in yellow and red have greater activity during those conditions with positive mean brain scores (i.e. these regions have positive weights on the LV) and areas shown in blue have greater rCBF during those conditions with negative mean brain scores (negative weights). There was a significant Task ⫻ Group interaction on the scores (see text). Local maxima of these regions can be found in Table 1.
same coordinates for the left amygdala were used for both patients and controls. The brain areas where blood flow was correlated with activity in the right prefrontal cortex (P ⬍ 0.001) in the controls are shown in Fig. 4A. Positive correlations across
the four delay conditions were found in widespread areas of prefrontal cortex bilaterally, including both ventral and dorsal portions (Table 4). Positive correlations were seen also in the bilateral extrastriate and parietal cortices, right hippocampus, and the posterior cingulate. The negative
Brain activity during face memory
745
Table 1 Local maxima of brain areas with differential activity during the short-delay conditions versus the control task (LV1 from group analysis) Region
Gyrus/Nuc
Hem
BA
x
y
z
Controls*
AD patients*
Short-delay ⬎ Control task Prefrontal GOb GOb GFi GFm GFd Premotor GPrC GPrC Extrastriate GF GF Parietal LPi LPi Thalamus
R L R R M R L R L R L R
11 11 45 46 6/32 6 6 19 19 39 39
20 –18 42 26 2 40 –32 44 –30 26 –28 18
32 26 26 44 8 4 –4 –68 –80 –68 –68 –18
–12 –16 20 8 44 28 32 –20 –12 32 32 4
3.80 5.95 4.54 3.92 3.12 4.78 4.13 7.91 6.41 9.14 4.73 3.17
0.22 1.24 0.67 –0.47 1.53 0.75 0.43 1.75 1.59 1.18 1.74 1.00
Control task ⬎ Short-delay Prefrontal GFd GFm GFm Temporal GTs GTs GTm GTm Parietal LPi LPi
L R L R L R L R L
10 9 9/46 42 22 21 21 40 39
–10 22 –30 50 –54 50 –56 48 –44
54 38 44 –30 –38 –8 –14 –46 –64
12 36 32 16 16 –8 –4 28 24
–5.19 –4.09 –3.41 –10.69 –5.83 –3.56 –3.30 –2.90 –3.85
–0.41 –0.45 –0.62 –0.53 –1.15 –1.78 –1.63 –1.29 –1.70
Coordinates and estimated Brodmann areas (BA) from Talairach and Tournoux (1988). x (right/left): negative values are in the left hemisphere; y (anterior/posterior), negative values are posterior to the zero point (located at the anterior commissure); z (superior/ inferior), negative values are inferior to the plane defined by the anterior and posterior commissures. *Reliability ratio for these regions in the separate analyses of control subjects and Alzheimer’s disease (AD) patients. Abbreviations for all tables: Nuc ⫽ nucleus; Hem ⫽ hemisphere; R ⫽ right; L ⫽ left; M ⫽ midline (艋5 mm from 0 point in x dimension); BA ⫽ Brodmann area; Cu ⫽ cuneus; GF ⫽ fusiform gyrus; GL ⫽ lingual gyrus; GOi ⫽ inferior occipital gyrus; GOm ⫽ middle occipital gyrus; GH ⫽ parahippocampal gyrus; GF(s, m, i, d) ⫽ frontal gyrus (superior, middle, inferior, medial); GOb ⫽ orbitofrontal gyrus; GC ⫽ cingulate gyrus; GPrC ⫽ precentral gyrus; GPoC ⫽ postcentral gyrus; Gs ⫽ subcallosal gyrus; Gsm ⫽ supramarginal gyrus; GT(s, m, i) ⫽ temporal gyrus (superior, middle, inferior); GTT ⫽ transverse temporal gyrus; LPi ⫽ inferior parietal; LPc ⫽ paracentral lobule; LPs ⫽ superior parietal; P ⫽ pulvinar nucleus of the thalamus; Pcu ⫽ precuneus.
correlations were mainly in ventral occipital, anterior temporal and sensory regions. In Alzheimer’s disease patients, the right prefrontal cortex was correlated (P ⬍ 0.001) with other prefrontal areas, particularly in the right hemisphere, showing considerable overlap with those prefrontal areas correlated with the right prefrontal cortex in the controls (Fig. 4B). Other areas of positive correlation were seen in the extrastriate and temporal cortex and the insula (Table 4). Areas where activity was negatively correlated with that in the right prefrontal cortex of the patients included posterior visual areas and some subcortical regions, including the putamen, where activity was positively correlated with the right prefrontal cortex in controls (Table 4). The areas where rCBF was positively correlated with activity in the left amygdala across all delay conditions in the control group (P ⬍ 0.001) were found mainly in temporal regions, including the posterior portion of both parahippocampal gyri and the fusiform gyri (Fig. 5A and Table 5). Positive correlations in the anterior cortex were restricted to the anterior cingulate in the left hemisphere and ventral prefrontal regions in the right hemisphere. Negative correlations with the left amygdala were seen in the bilateral
dorsal frontal, parietal and posterior extrastriate regions, as well as in the dorsal cingulate. The Alzheimer’s disease patients showed a pattern of positive correlations (P ⬍ 0.01; Fig. 5B) that included the bilateral posterior parahippocampal gyri, the left pulvinar and insula, a large area of the left prefrontal cortex that included ventral and dorsal regions and both anterior and posterior cingulate (Table 5). Negative correlations of rCBF were seen in the posterior extrastriate cortex and the right dorsolateral prefrontal cortex.
Discussion The behavioural results of this experiment are consistent with previous studies of memory impairment in Alzheimer’s disease that have shown worsening memory performance with delay and provide further evidence that even delays of a few seconds are sufficient to degrade performance. This was in sharp contrast to the recognition accuracy found in the healthy controls, which did not change across the delay conditions. This difference in performance can be viewed as an increased vulnerability of visual representations over short periods of time in Alzheimer’s disease patients, represen-
746
C. L. Grady et al.
tations that are readily maintained in healthy elderly adults. This increased vulnerability is probably related to the degenerative changes that are presumably occurring in the
medial temporal regions, and the cortical regions with which they connect. Consistent with the behavioural data, the brain activity pattern associated with memory delay in the
Brain activity during face memory
747
Table 2 Local maxima of areas with activity characterizing the longer delay conditions Region
Gyrus/Nuc
Hem
BA
x
y
z
Controls*
AD patients*
Long-delay ⬎ Control task in Alzheimer’s disease patients* Prefrontal GFi R 47 GFi L 11 GFm R 46 GFm R 6 Cingulate L 32 Insula L Amygdala L Extrastriate Cu R 19 Globus pallidus R
32 –32 24 22 –16 –38 –24 2 22
30 26 28 –2 30 –12 –6 –74 –14
4 –12 28 44 32 –8 –12 32 4
2.17 1.71 1.83 1.96 1.34 1.24 1.23 0.89 1.75
2.16 1.78 2.92 1.93 3.02 1.85 2.65 2.34 1.96
Long-delay ⬎ Control task in control subjects† Prefrontal GFm R GFm L Cingulate M Temporal GTm L
30 –30 2 –50
52 52 38 –14
8 4 0 –8
2.57 2.39 2.47 2.49
–0.72 0.20 –0.32 0.56
10 10 24 21
Abbreviations can be found in the legend to Table 1. *Regions from LV2 of the combined group analysis. †Regions from LV2 of the within-group analysis of the controls. ‡Reliability ratio for these regions in the separate analyses of control subjects and Alzheimer’s disease (AD) patients.
Table 3 Comparison of task and behavioural analyses Analysis Controls Prefrontal cortex Task Reaction time Task Reaction time Temporal cortex Task Reaction time Alzheimer’s disease patients Left amygdala region Task Reaction time Accuracy Right prefrontal Task Accuracy Right cuneus Task Reaction time Accuracy
x
y
z
1 s*
6s
11 s
16 s
Ratio†
30 34 –30 –32
52 54 52 58
8 16 4 12
1.7 –0.31 –1.0 –0.52
5.0 –0.47 1.0 –0.32
4.4 –0.27 3.0 –0.18
6.4 –0.27 4.0 –0.44
2.57 3.74 2.39 4.36
–50 –48
–14 –16
–8 –12
–3.3 –0.56
–1.7 –0.72
0.9 –0.36
1.3 –0.44
2.49 5.59
–24 –32 –22
–6 –12 –8
–12 –12 –20
4.6 –0.69 0.70
8.5 –0.83 0.13
9.3 –0.39 0.43
7.3 –0.73 0.56
2.65 6.86 2.39
32 36
36 24
–4 –4
3.3 0.58
7.5 –0.11
4.3 0.48
6.6 0.40
2.50 2.48
4 0 –2
–76 –76 –72
32 32 32
3.8 –0.74 0.66
7.9 –0.81 –0.03
4.5 –0.67 0.45
5.2 –0.80 0.59
2.59 5.20 2.66
Coordinates shown in italics indicate the regions chosen for the functional connectivity analysis. *Values for the Task analysis are the percentage increase above the control task level at the specified coordinates. The coordinates for the task are from the within group analyses and are the same as in Table 2 for the control subjects. For the Alzheimer’s disease patients the coordinates are those closest to the maxima shown in Table 2. Values for the behavioural analyses are the correlations between the behavioural measures and rCBF at the given coordinates (which are maxima from the behavioural analyses). †Ratio ⫽ reliability ratio for these regions from the specified analysis.
Fig. 3 Brain areas identified by the second combined–group task LV (A), by the first LV in the Alzheimer’s disease patients (B), and by the second LV in the controls (C) are shown (brain areas meeting the bootstrap criterion of reliability ratio (艌2.0). Areas are shown on four slices from a standard MRI with the position of each slice relative to the AC–PC line shown above the slice. The graph at the bottom of the figure shows the mean brain scores for each condition on the combined–group LV for controls and patients. In all images brain regions shown in yellow and red have greater activity during the conditions with positive brain scores and areas shown in blue have greater rCBF during the conditions with negative brain scores. There was a significant Task ⫻ Group interaction on the scores (see text). Local maxima of these regions can be found in Table 2.
748
C. L. Grady et al.
Fig. 4 Brain areas where rCBF is correlated with activity in right prefrontal cortex in controls and patients. Areas are shown on a standard MRI beginning at –28 mm relative to the AC–PC line (top left slice) and extending up to 40 mm above the AC–PC line (bottom right slice). White areas are those where activity is positively correlated with activity in the prefrontal regions (indicated by arrows) across the four delay conditions and black areas are those with negative correlations. (A) Areas of correlation with right prefrontal cortex (x ⫽ 30, y ⫽ 52, z ⫽ 8) in the control group. (B) Areas correlated with right prefrontal cortex (x ⫽ 32, y ⫽ 36, z ⫽ –4) in Alzheimer’s disease patients. Local maxima of regions of correlated activity can be found in Table 4.
Alzheimer’s disease patients also suggested a decline in function with increasing delay, as it was most prominent in the 6 s condition. Thus, both the memory performance and the brain activity data in the patients suggest difficulty in maintaining a face representation for more than about 5 s.
The control group showed brain activity patterns that characterized both the short-and long-delay conditions. One pattern was related both to short-delay conditions and face perception and included a number of regions previously found to be involved in these processes, including extrastriate
Brain activity during face memory Table 4 Correlations with activity in right prefrontal cortex Region Controls Positive correlations Prefrontal
Extrastriate
Temporal Hippocampus Parietal
Gyrus/Nuc
GFi GFi GFi GFm GFm GFm GFm GF GF GF GL GTT GH LPi LPi Pcu Pcu
Cingulate Caudate/putamen Negative correlations Temporal GTi GTi GTs Extrastriate GF GF GL GOm Parietal LPs Sensory GPoC GPoC Alzheimer’s disease patients Positive correlations Prefrontal GFi GFi GFi GFd GFm GFm Extrastriate GL GL GF/GH GF/GH Motor GPrC Temporal GTs GTm Insula Parietal LPi Negative correlations Temporal GTs Extrastriate Cu GF Cingulate GC Thalamus P P Putamen L
Hem
BA
x
y
z
Ratio*
R L L R R L L R L L R L L R R L R L L R L
47 47 45 9/10 9/46 46 9 37 37 18 18 41 36
24 –32 –42 24 40 –36 –38 42 –46 –32 12 –34 –16 28 54 –50 10 –24 0 18 –20
20 20 22 48 28 46 34 –36 –44 –84 –90 –24 –36 –30 –36 –46 –72 –62 –36 14 12
–4 –8 16 24 36 8 36 –12 –16 –4 –4 12 –16 –12 24 32 24 20 36 8 4
4.25 3.21 6.12 9.87 8.84 10.68 5.70 3.88 6.43 5.20 3.62 3.79 3.54 3.49 3.53 4.30 3.12 5.56 3.16 4.22 2.83
R L R R L L R L R L
20 20 22 19 37 18 18 7 2 2
52 –38 50 34 –48 –6 30 –38 40 –44
–32 –28 –16 –72 –62 –82 –88 –70 –24 –26
–28 –24 0 –28 –24 0 8 40 32 36
3.10 3.91 3.24 4.57 3.14 3.26 3.69 5.28 3.07 6.97
R R L R R L R L R L R R R R L R
47 46 47 8 9 10 18 18 36 36 4 42 21
40 20 –24 6 34 –38 18 –18 36 –28 44 50 50 38 –34 24
16 42 18 44 20 50 –74 –74 –24 –28 –14 –24 –40 –16 –6 –54
0 20 –20 36 36 0 –4 0 –20 –24 40 16 –8 –4 0 40
11.88 10.99 4.44 9.29 10.07 7.45 6.83 3.04 3.53 3.25 6.94 4.85 3.18 6.94 5.27 5.14
L L L R R L R
42 18 37 23
–36 –12 –44 8 16 –18 22 –18
–34 –94 –38 –58 –28 –26 0 6
16 8 –12 12 12 12 4 0
9.00 4.84 5.57 4.69 2.68 3.94 4.01 4.10
40 40 18 31 31
40
Abbreviations can be found in the legend to Table 1.* Ratio ⫽ reliability ratio for these regions.
749
750
C. L. Grady et al.
Fig. 5 Brain areas where rCBF is correlated with activity in the left amygdala (x ⫽ –24, y ⫽ –6, z ⫽ –12) in controls (A) and patients (B). Areas are shown on a standard MRI beginning at –28 mm relative to the AC–PC line (top left slice) and extending up to 40 mm above the AC–PC line (bottom right slice). White areas are those where activity is positively correlated with activity in the left amygdala (indicated by arrow in B) across the four delay conditions and black areas are those with negative correlations. Local maxima of regions of correlated activity can be found in Table 5.
regions and the ventral prefrontal cortex (Haxby et al., 1994, 1995; Kanwisher et al., 1997). The other pattern identified regions with increased activity in the longer-delay conditions, mainly in the prefrontal cortex, similar to that previously described in a subset of these older adults (Grady et al., 1998) and consistent with other reports of prefrontal activation during face memory (Haxby et al., 1995; Courtney et al., 1997; Furey et al., 1997). This finding is expanded in the
current analysis to show that these prefrontal areas show strong positive interactions with one another in both shortand long-delay conditions, within each hemisphere as well as across hemispheres. In addition, activity in the right anterior prefrontal cortex correlates positively with activity in the extrastriate cortex and with the right hippocampus across the memory delays. It is worth noting that the role of the hippocampus in this memory task was revealed only
Brain activity during face memory Table 5 Correlations with activity in the left amygdala Region Controls Positive correlations Prefrontal
Gyrus/Nuc
GFi GFd
Cingulate Premotor Extrastriate Temporal
GPrC GF GF GF GTi GTi GH GH
Amygdala Insula Negative correlations Prefrontal GFi GFd GFm GFm Cingulate GC Temporal GTm GTm GTs Extrastriate GL Cu GOm GOm Parietal Gsm LPs LPs Alzheimer’s disease patients Positive correlations Prefrontal GFd GFd GFm GFm GFm Premotor GPrC Cingulate GC GC GC Insula Temporal GH GH GTm GTs Thalamus P Caudate nucleus Cerebellum Negative correlations Prefrontal GFm Temporal GTs Extrastriate GL GL GF Cu
Hem
BA
x
y
z
Ratio*
R R L L L R R L R L R L R L
45 10 32 32 6 19 37 19 20 20 36 35
24 12 –16 –18 –48 26 42 –22 44 –44 18 –18 30 –30
42 50 18 42 –2 –66 –38 –62 –6 –32 –32 –42 –2 –4
0 12 –8 4 8 –20 –16 –12 –24 –16 –8 –8 –12 16
4.36 4.83 7.90 3.26 3.93 3.62 4.60 3.35 4.05 3.08 3.06 4.66 4.00 3.09
L R L R R R L L L R R L R R L
9/44 9 9 6 32 21 39 22 18 18 19 19 40 7 7
–38 14 –24 26 2 60 –42 –54 –8 20 44 –30 34 22 –22
14 38 36 2 18 –48 –70 12 –98 –88 –76 –90 –46 –64 –66
28 32 32 48 40 4 16 4 –12 4 4 16 36 44 40
5.54 4.53 7.22 10.70 10.58 5.03 5.39 4.06 4.16 4.99 5.59 4.34 5.68 7.05 6.96
L L R L L L R L L L R L L L L R L L
11 10 10 46 9/46 6 24 24 23
–8 –18 22 –36 –28 –38 6 –14 –6 –34 12 –34 –58 –52 –20 8 –10 –10
40 48 46 44 30 2 –12 34 –52 –22 –42 –38 –12 10 –34 10 20 –46
–12 12 4 4 28 20 36 4 16 4 0 –4 –16 0 8 –4 –4 –28
2.85 4.22 2.57 2.77 4.00 4.05 5.08 4.74 4.35 3.38 3.68 3.92 3.68 3.04 4.59 4.09 3.96 3.56
R L R L L R
9 22 18 18 19 19
28 –56 10 –16 –22 18
48 –20 –66 –94 –64 –72
28 4 4 –12 –8 32
4.51 3.39 5.13 4.03 4.25 2.56
35 35 21 22
Abbreviations can be found in the legend to Table 1. *Ratio ⫽ reliability ratio for these regions.
751
752
C. L. Grady et al.
through the analysis that examined functional connectivity and not in the task analysis, highlighting the usefulness of this approach to neuroimaging. The correlated activity between the right prefrontal cortex and the extrastriate cortex during these face memory tasks replicates our previous finding of correlated activity between these areas during face perception without a memory component (Horwitz et al., 1995). These two sets of results together indicate that healthy older adults engage a network of regions that includes posterior visual areas and the prefrontal cortex for face processing in general, and additionally recruit the hippocampus for representing faces in short-term memory. The importance of the prefrontal cortex for working memory function in the elderly is further underlined by the association between activity in prefrontal regions and better performance on the tasks, as measured by reaction time, a correlation that also has been reported in older individuals by Rypma and D’Esposito (Rypma and D’Esposito, 2000). The Alzheimer’s disease patients showed marked differences from the controls in brain activity patterns during memory as well as in functional connectivity. In the patients, prefrontal activity during the memory tasks was found only in the right hemisphere and the correlations between this region and other brain areas were also limited mainly to the frontal lobes. There was no correlated activity between the right prefrontal cortex and the hippocampus in the patients during the delay conditions. This is consistent with our hypothesis of anterior–posterior disconnection in these patients and indicates that the prefrontally mediated network used by healthy elderly adults for this task is disrupted in early Alzheimer’s disease. In addition, there was limited correlation between activity in the prefrontal cortex and visual areas seen in the patients, particularly the face-sensitive ventral visual areas, suggesting that there is a functional disconnection of these areas consistent with that found previously during face perception (Horwitz et al., 1995). The absence of increased left prefrontal activity during the longer memory delays in the patients is in contrast to that previously found in healthy young adults (Haxby et al., 1995) and to that seen in the control group in this experiment. The increase of left prefrontal activity during this task has been interpreted as a shift from a purely maintenance strategy at short delays to strategies relying on more elaborate processing of the faces at longer delays. Since the patients show no such increase, this may be an indication that they fail to engage these more elaborate strategies. In addition, we did not find an above-normal increase of prefrontal activity in the Alzheimer’s disease patients such as that reported in patients during verbal memory paradigms (Becker et al., 1996; Woodard et al., 1998; Backman et al., 1999), although this difference could be due to the inherent differences between verbal and nonverbal tasks as well as the differences between recognition and recall. Nevertheless, the maintained activity in the right prefrontal cortex was consistent with our hypothesis that prefrontal activity would be relatively spared during these tasks. However, the functional disconnection of
the prefrontal cortex with the posterior cortical regions indicates that an examination of task-related rCBF differences alone is not sufficient to rule out alterations of function in any given area. Perhaps the most interesting, and unexpected, difference between patients and controls was the increased activity in left medial temporal regions seen at the longer delays in the patients, despite the fact that these areas are affected relatively early in Alzheimer’s disease (Braak et al., 1993). In addition, the amygdala in patients was found to have functional connections with a widespread area of the left prefrontal cortex, as well as the insula, anterior cingulate, pulvinar, and posterior portions of the parahippocampal gyri. These results highlight two important points about medial temporal activation in Alzheimer’s disease patients during cognitive tasks. The first is that the medial temporal regions may show increased activity during cognitive tasks, despite presumed damage to these areas, although others have shown such activation to be less than normal (Small et al., 1999). The second point is that one medial temporal area, such as the hippocampus, may be functionally disconnected to the prefrontal cortex in Alzheimer’s disease patients, compared with controls, whereas another region, such as the amygdala, may have increased connections to the prefrontal cortex. This suggests that dynamic changes in the functional roles of medial temporal regions can occur in demented patients and are reflected in how activity in these regions covaries with the rest of the brain. The amygdala is anatomically connected to all of the regions with which it showed correlated activity (Amaral et al., 1992) and many of these areas, including the amygdala, are involved in the emotional processing of faces. For example, the amygdala is thought by some to mediate the processing of facial emotion, particularly fear (Adolphs et al., 1994; Young et al., 1995; Broks et al., 1998; Morris et al., 1998; Whalen et al., 1998a), even when the emotional processing is implicit or unconscious (Whalen et al., 1998a; Morris et al., 1999). This region may also be active during nonemotional processing of unfamiliar faces (Dubois et al., 1999). The insula (Phillips et al., 1997), pulvinar (Morris et al., 1999), and inferior prefrontal cortex (Hornak et al., 1996) are also involved in emotional processing of faces and the ventral part of the anterior cingulate is activated in a number of emotional tasks (Mayberg, 1997; Taylor et al., 1998; Whalen et al., 1998b). The finding of correlated activity among all these areas in the patients, but not in controls, may indicate that the patients have accessed a network that mediates the processing of facial emotion. If so, this access could have been either explicit or implicit. That is, the patients could have placed more emphasis on the emotional expressions of the faces in their attempt to remember them, whereas the controls could have relied more on other facial characteristics, such as the particular features. This conjecture of feature analysis in controls is consistent with their selfreports of strategies used during the task. Although none of the patients reported using emotion as a mnemonic strategy,
Brain activity during face memory they may not have had sufficient insight into their strategies or may not have remembered them. On the other hand, access of emotional information by the patients might have been implicit, similar to the ‘mere exposure effect’, which refers to the finding that subjects express greater preference for stimuli that they have been exposed to over new stimuli (Kunst-Wilson and Zajonc, 1980). This effect, which is considered to be an affective response and a measure of implicit memory, has been found for faces in Alzheimer’s disease patients to the same extent as that seen in healthy elderly adults despite patients’ impaired explicit memory for faces (Winograd et al., 1999; Hamann et al., 2000). These findings suggest that Alzheimer’s disease patients’ implicit emotional reactions to faces are intact and even explicit emotional processing of faces in Alzheimer’s disease patients may not be impaired (Roudier et al., 1998). In addition, there is evidence that affective responses to stimuli occur prior to cognitive processing of these stimuli and that once cognitive processing occurs, it can modify or remove the influence of the affective response (Murphy and Zajonc, 1993). So it may be that the control subjects utilize a cognitive system to remember the faces (e.g. an anterior prefrontal–hippocampal– extrastriate system), whereas Alzheimer’s disease patients have an impairment of the cognitive processing system so the affective system may assume greater influence (e.g. an inferior prefrontal–amygdala–insula system). In addition, the affective system may partially ameliorate the damage to memory resulting from the altered cognitive system, as indicated by the finding that activity in the left amygdala correlated with better performance on the memory tasks as measured by both accuracy and reaction time. Clearly, although our results are consistent with such a notion, this is a hypothesis that needs direct experimental confirmation in future work. Thus, our results indicate that Alzheimer’s disease patients are impaired on short-term face memory tasks and show differences in the functional networks underlying memory over short delays. Although a difference in performance between the patients and controls leads to an obvious difficulty in interpretation of the brain activity data, there are two aspects of the behavioural data that mitigate these difficulties. First, the patients were able to perform the face memory task at near normal levels during the 1 s delay condition, making it unlikely that brain differences seen in this condition were due to an inability to carry out the task instructions. Secondly, even at the 16 s delay condition, their performance was well above chance (50% for this type of two alternative task), indicating that the overt behaviour of the patients was similar to that of the controls for the majority of the time spent on the task. Thus the differences seen in brain activity between the two groups reflect alterations in the way the task was carried out but not a failure on the part of the patients to understand what was expected of them. We have been able to show that in mildly demented Alzheimer’s disease patients the breakdown in memory for unfamiliar faces is due, at least in part, to a reduction in the
753
functional connectivity of the right prefrontal cortex with other areas such as the right hippocampus and visual cortices. These connections are present in healthy elderly individuals and appear to be critical for maintaining representations of the faces over the short-delay intervals used in this experiment. These results clearly show the usefulness of assessing functional interactions in disease populations, which can reveal the failure to engage the network used by controls and can identify alternate networks that may be involved. They also provide further support for the idea that alternate brain networks may play a compensatory role early in the course of Alzheimer’s disease under some conditions and that in the case of face memory this alternate network may be one that mediates the processing of the affective content in the faces.
Acknowledgements This work was supported by the intramural program of the National Institute on Aging (USA), the Alzheimer’s Society of Canada (#97–03), and the Canadian Institutes of Health Research (MOP14036).
References Adolphs R, Tranel D, Damasio H, Damasio A. Impaired recognition of emotion in facial expressions following bilateral damage to the human amygdala. Nature 1994; 372: 669–72. Amaral DG, Price JL, Pitkanen A, Carmichael ST. Anatomical organization of the primate amygdaloid complex. In: Aggleton JP, editor. The amygdala. Neurobiological aspects of emotion, memory and mental dysfunction. New York: Wiley–Liss; 1992. p. 1–66. Backman L, Andersson JL, Nyberg L, Winblad B, Nordberg A, Almkvist O. Brain regions associated with episodic retrieval in normal aging and Alzheimer’s disease. Neurology 1999; 52: 1861–70. Becker JT, Mintun MA, Aleva K, Wiseman MB, Nichols T, DeKosky ST. Compensatory reallocation of brain resources supporting verbal episodic memory in Alzheimer’s disease. Neurology 1996; 46: 692–700. Braak H, Braak E, Bohl J. Staging of Alzheimer-related cortical destruction. [Review]. Eur Neurol 1993; 33: 403–8. Broks P, Young AW, Maratos EJ, Coffey PJ, Calder AJ, Isaac CL, et al. Face processing impairments after encephalitis: amygdala damage and recognition of fear. Neuropsychologia 1998; 36: 59–70. Cabeza R, Nyberg L. Imaging cognition II: An empirical review of 275 PET and fMRI studies. [Review]. J Cogn Neurosci 2000; 12: 1–47. Cabeza R, Grady CL, Nyberg L, McIntosh AR, Tulving E, Kapur S, et al. Age-related differences in neural activity during memory encoding and retrieval: a positron emission tomography study. J Neurosci 1997; 17: 391–400.
754
C. L. Grady et al.
Carson RE, Berg GW, Finn RD, Patlak CS, Daube-Witherspoon ME, Stein SD, et al. Tomographic measurement of LCBF with high-resolution PET and H215O: comparison of methods. J Cereb Blood Flow Metab 1987; 7 Suppl 1: S578. Chase TN, Burrows GH, Mohr E. Cortical glucose utilization patterns in primary degenerative dementias of the anterior and posterior types. Arch Gerontol Geriatr 1987; 6: 289–97. Convit A, De Leon MJ, Tarshish C, De Santi S, Tsui W, Rusinek H, et al. Specific hippocampal volume reductions in individuals at risk for Alzheimer’s disease. Neurobiol Aging 1997; 18: 131–8. Courtney SM, Ungerleider LG, Keil K, Haxby JV. Transient and sustained activity in a distributed neural system for human working memory. Nature 1997; 386: 608–11. Duara R, Grady C, Haxby J, Ingvar D, Sokoloff L, Margolin RA, et al. Human brain glucose utilization and cognitive function in relation to age. Ann Neurol 1984; 16: 702–13. Dubois S, Rossion B, Schiltz C, Bodart JM, Michel C, Bruyer R, et al. Effect of familiarity on the processing of human faces. Neuroimage 1999; 9: 278–89. Edgington ES. Randomization tests. New York: Marcel Dekker; 1980. Efron B, Tibshirani R. Bootstrap methods for standard errors, confidence intervals and other measures of statistical accuracy. Stat Sci 1986; 1: 54–77.
Grady CL. Neuroimaging and activation of the frontal lobes. In: Miller BL, Cummings JL, editors. The human frontal lobes: function and disorders. New York: Guilford Press, 1999. p. 196–230. Grady CL, Haxby JV, Horwitz B, Sundaram M, Berg G, Schapiro MB, et al. Longitudinal study of the early neuropsychological and cerebral metabolic changes in dementia of the Alzheimer type. J Clin Exp Neuropsychol 1988; 10: 576–96. Grady CL, Haxby JV, Schapiro MB, Gonzalez-Aviles A, Kumar A, Ball MJ, et al. Subgroups in dementia of the Alzheimer type identified using positron emission tomography. J Neuropsychiatry Clin Neurosci 1990; 2: 373–84. Grady CL, McIntosh AR, Bookstein F, Horwitz B, Rapoport SI, Haxby JV. Age-related changes in regional cerebral blood flow during working memory for faces. Neuroimage 1998; 8: 409–25. Hamann SB, Monarch ES, Goldstein FC. Memory enhancement for emotional stimuli is impaired in early Alzheimer’s disease. Neuropsychology 2000; 14: 82–92. Hart RP, Kwentus JA, Harkins SW, Taylor JR. Rate of forgetting in mild Alzheimer’s-type dementia. Brain Cogn 1988; 7: 31–8. Haxby JV, Duara R, Grady CL, Cutler NR, Rapoport SI. Relations between neuropsychological and cerebral metabolic asymmetries in early Alzheimer’s disease. J Cereb Blood Flow Metab 1985; 5: 193–200.
Foster NL, Chase TN, Mansi L, Brooks R, Fedio P, Patronas NJ, et al. Cortical abnormalities in Alzheimer’s disease. Ann Neurol 1984; 16: 649–54.
Haxby JV, Horwitz B, Ungerleider LG, Maisog JM, Pietrini P, Grady CL. The functional organization of human extrastriate cortex: a PET-rCBF study of selective attention to faces and locations. J Neurosci 1994; 14: 6336–53.
Frackowiak RS, Friston KJ. Functional neuroanatomy of the human brain: positron emission tomography—a new neuroanatomical technique. [Review]. J Anat 1994; 184: 211–25.
Haxby JV, Ungerleider LG, Horwitz B, Rapoport SI, Grady CL. Hemispheric differences in neural systems for face working memory: a PET-rCBF study. Hum Brain Mapp 1995; 3: 68–82.
Friedland RP, Budinger TF, Koss E, Ober BA. Alzheimer’s disease: anterior-posterior and lateral hemispheric alterations in cortical glucose utilization. Neurosci Lett 1985; 53: 235–40.
Holden JE, Gatley SJ, Hichwa RD, Ip WR, Shaughnessy WJ, Nickles RJ, et al. Cerebral blood flow using PET measurements of fluoromethane kinetics. J Nucl Med 1981; 22: 1084–88.
Friston KJ, Frith CD, Liddle PF, Frackowiak RS. Functional connectivity: the principal-component analysis of large (PET) data sets. J Cereb Blood Flow Metab. 1993; 13: 5–14.
Hornak J, Rolls ET, Wade D. Face and voice expression identification in patients with emotional and behavioural changes following ventral frontal lobe damage. Neuropsychologia 1996; 34: 247–61.
Furey ML, Pietrini P, Haxby JV, Alexander GE, Lee HC, Van Meter J, et al. Cholinergic stimulation alters performance and task-specific regional cerebral blood flow during working memory. Proc Natl Acad Sci USA 1997; 94: 6512–16.
Horwitz B. Data analysis paradigms for metabolic-flow data: combining neural modeling and functional neuroimaging. Hum Brain Mapp 1994; 2: 112–22.
Fuster JM, Bauer RH, Jervey JP. Functional interactions between inferotemporal and prefrontal cortex in a cognitive task. Brain Res 1985; 330: 299–307. Goldman-Rakic PS. Cellular and circuit basis of working memory in prefrontal cortex of nonhuman primates. In: Uylings HBM, Van Eden CG, De Bruin JPC, Corner MA, Feenstra MGP, editors. The prefrontal cortex. Progress in brain research; vol. 85. Amsterdam: Elsevier; 1990. p. 325–36. Golomb J, Kluger A, de Leon MJ, Ferris SH, Convit A, Mittelman MS, et al. Hippocampal formation size in normal human aging: a correlate of delayed secondary memory performance. Learn Mem 1994; 1: 45–54.
Horwitz B, McIntosh AR, Haxby JV, Furey M, Salerno J, Schapiro MB, et al. Network analysis of PET-mapped visual pathways in Alzheimer type dementia. Neuroreport 1995; 6: 2287–92. Jacobs DM, Sano M, Dooneief G, Marder K, Bell KL, Stern Y. Neuropsychological detection and characterization of preclinical Alzheimer’s disease. Neurology 1995; 45: 957–62. Jagust WJ, Friedland RP, Budinger TF, Koss E, Ober B. Longitudinal studies of regional cerebral metabolism in Alzheimer’s disease. Neurology 1988; 38: 909–12. Kanwisher N, McDermott J, Chun MM. The fusiform face area: a module in human extrastriate cortex specialized for face perception. J Neurosci 1997; 17: 4302–11.
Brain activity during face memory Kemper TL. Neuroanatomical and neuropathological changes during aging and in dementia. In: Albert ML, Knoefel JE, editors. Clinical neurology of aging. 2nd ed. New York: Oxford University Press; 1994. p. 3–67. Kohler S, Black SE, Sinden M, Szekely C, Kidron D, Parker JL, et al. Memory impairments associated with hippocampal versus parahippocampal-gyrus atrophy: an MR volumetry study in Alzheimer’s disease. Neuropsychologia 1998; 36: 901–14. Kunst-Wilson WR, Zajonc RB. Affective discrimination of stimuli that cannot be recognized. Science 1980; 207: 557–8. Madden DJ, Turkington TG, Provenzale JM, Denny LL, Hawk TC, Gottlob LR, et al. Adult age differences in the functional neuroanatomy of verbal recognition memory. Hum Brain Mapp 1999; 7: 115–35.
755
verbal and spatial working memory revealed by PET. J Cogn Neurosci 2000; 12: 174–87. Roudier M, Marcie P, Grancher A-S, Tzortzis C, Starkstein S, Boller F. Discrimination of facial identity and of emotions in Alzheimer’s disease. J Neurol Sci 1998; 154: 151–8. Rypma B, D’Esposito M. Isolating the neural mechanisms of agerelated changes in human working memory. Nat Neurosci 2000; 3: 509–15. Sahakian BJ, Morris RG, Evenden JL, Heald A, Levy R, Philpot M, et al. A comparative study of visuospatial memory and learning in Alzheimer-type dementia and Parkinson’s disease. Brain 1988; 111: 695–718.
Mayberg HS. Limbic-cortical dysregulation: a proposed model of depression. J Neuropsychiatry Clin Neurosci 1997; 9: 471–81.
Sahakian BJ, Downes JJ, Eagger S, Evenden JL, Levy R, Philpot MP, et al. Sparing of attentional relative to mnemonic function in a subgroup of patients with dementia of the Alzheimer type. Neuropsychologia 1990; 28: 1197–213.
McGeer EG, Peppard RP, McGeer PL, Tuokko H, Crockett D, Parks R, et al. 18-Fluorodeoxyglucose positron emission tomography studies in presumed Alzheimer cases, including 13 serial scans. Can J Neurol Sci 1990; 17: 1–11.
Sampson PD, Streissguth AP, Barr HM, Bookstein FL. Neurobehavioral effects of prenatal alchohol: Part II. Partial least squares analysis. Neurotoxicol Teratol 1989; 11: 477–91.
McIntosh AR. Mapping cognition to the brain through neural interactions. [Review]. Memory 1999; 7: 523–48. McIntosh AR, Gonzalez-Lima F. Structural equation modeling and its application to network analysis in functional brain imaging. Hum Brain Mapp 1994; 2: 2–22. McIntosh AR, Bookstein FL, Haxby JV, Grady CL. Spatial pattern analysis of functional brain images using partial least squares. Neuroimage 1996; 3: 143–57. Miller EK, Li L, Desimone R. A neural mechanism for working and recognition memory in inferior temporal cortex. Science 1991; 254: 1377–9.
Small SA, Perera GM, DeLaPaz R, Mayeux R, Stern Y. Differential regional dysfunction of the hippocampal formation among elderly with memory decline and Alzheimer’s disease. Ann Neurol 1999; 45: 466–72. Squire LR. Memory and the hippocampus: a synthesis from findings with rats, monkeys, and humans. Psychol Rev 1992; 99: 195–231. Talairach J, Tournoux P. Co-planar stereotaxic atlas of the human brain. Stuttgart: Thieme; 1988. Taylor SF, Liberzon I, Fig LM, Decker LR, Minoshima S, Koeppe RA. The effect of emotional content on visual recognition memory: a PET activation study. Neuroimage 1998; 8: 188–97.
Morris JS, Friston KJ, Buchel C, Frith CD, Young AW, Calder AJ, et al. A neuromodulatory role for the human amygdala in processing emotional facial expressions. Brain 1998; 121: 47–57.
Wechsler DA. Wechsler Adult Intelligence Scale. New York: Psychological Corporation; 1955.
Morris JS, Ohman A, Dolan RJ. A subcortical pathway to the right amygdala mediating ‘unseen’ fear. Proc Natl Acad Sci USA 1999; 96: 1680–5.
Welsh K, Butters N, Hughes J, Mohs R, Heyman A. Detection of abnormal memory decline in mild cases of Alzheimer’s disease using CERAD neuropsychological measures. Arch Neurol 1991; 48: 278–81.
Moss MB, Albert MS, Butters N, Payne M. Differential patterns of memory loss among patients with Alzheimer’s disease, Huntington’s disease, and alcoholic Korsakoff’s syndrome. Arch Neurol 1986; 43: 239–46. Murphy ST, Zajonc RB. Affect, cognition, and awareness: affective priming with optimal and suboptimal stimulus exposures. J Pers Soc Psychol 1993; 64: 723–39. Phillips ML, Young AW, Senior C, Brammer M, Andrew C, Calder AJ, et al. A specific neural substrate for perceiving facial expressions of disgust. Nature 1997; 389: 495–8. Price BH, Gurvit H, Weintraub S, Geula C, Leimkuhler E, Mesulam M. Neuropsychological patterns and language deficits in 20 consecutive cases of autopsy-confirmed Alzheimer’s disease. Arch Neurol 1993; 50: 931–7. Reuter-Lorenz PA, Jonides J, Smith EE, Hartley A, Miller A, Marshuetz C, et al. Age differences in the frontal lateralization of
Whalen PJ, Rauch SL, Etcoff NL, McInerney SC, Lee MB, Jenike MA. Masked presentations of emotional facial expressions modulate amygdala activity without explicit knowledge. J Neurosci 1998a; 18: 411–18. Whalen PJ, Bush G, McNally RJ, Wilhelm S, McInerney SC, Jenike MA, et al. The emotional counting Stroop paradigm: a functional magnetic resonance imaging probe of the anterior cingulate affective division. Biol Psychiatr 1998b; 44: 1219–28. Winograd E, Goldstein FC, Monarch ES, Peluso JP, Goldman WP. The mere exposure effect in patients with Alzheimer’s disease. Neuropsychology 1999; 13: 41–6. Woodard JL, Grafton ST, Votaw JR, Green RC, Dobraski ME, Hoffman JM. Compensatory recruitment of neural resources during overt rehearsal of word lists in Alzheimer’s disease. Neuropsychology 1998; 12: 491–504.
756
C. L. Grady et al.
Woods RP, Cherry SR, Mazziotta JC. Rapid automated algorithm for aligning and reslicing PET images. J Comput Assist Tomogr 1992; 16: 620–33. Young AW, Aggleton JP, Hellawell DJ, Johnson M, Broks P, Hanley JR. Face processing impairments after amygdalotomy. Brain 1995; 118: 15–24. Zec RF. Neuropsychological functioning in Alzheimer’s disease.
In: Parks RW, Zec RF, Wilson RS, editors. Neuropsychology of Alzheimer’s disease and other dementias. New York: Oxford University Press; 1993. p. 3–80.
Received July 24, 2000. Revised October 10, 2000. Accepted December 4, 2000