Specific versus Nonspecific Brain Activity in a Parametric N-Back Task

Share Embed


Descrição do Produto

NeuroImage 12, 688 – 697 (2000) doi:10.1006/nimg.2000.0645, available online at http://www.idealibrary.com on

Specific versus Nonspecific Brain Activity in a Parametric N-Back Task Johan Martijn Jansma,* ,1 Nick F. Ramsey,* Richard Coppola,† and Rene´ S. Kahn* *Department of Psychiatry, HP A01.126, University Medical Center Utrecht, P.O. Box 85500, 3508 GA Utrecht, The Netherlands; and †Clinical Brain Disorders Branch, National Institute of Mental Health/National Institutes of Health, Building 10, 9000 Rockville Pike, Bethesda, Maryland 20892 Received February 1, 2000

INTRODUCTION In this study functional magnetic resonance imaging (fMRI) was used to examine cerebral activity patterns in relation to increasing mental load of a working memory task. Aim of the experiment was to distinguish nonspecific task-related processes from specific workload processes analytically. Twelve healthy volunteers engaged in a spatial n-back task with four levels. FMRI data were acquired with the 3D-PRESTO pulse sequence. Analysis entailed a twostep multiple regression algorithm, which was specifically designed to measure and separate load-sensitive and load-insensitive activity simultaneously, while preserving the original high spatial resolution of the fMRI signal. Load-sensitive and load-insensitive activity was found in both dorsolateral-prefrontal and parietal cortex, predominantly bilaterally, and in the anterior cingulate. As expected, the left primary sensorimotor cortex showed predominantly load-insensitive activity. Load-sensitive activity reflects specific working memory functions, such as temporary retention and manipulation of information, while load-insensitive activity reflects supportive functions, such as visual orientation, perception, encoding, and response selection and execution. Good performance was correlated with a large area of load-sensitive activity in anterior cingulate, and with a small area of load-insensitive activity in the right parietal cortex. The findings indicate that nonspecific and specific working memory processes colocalize and are represented in multiple frontal and parietal regions. Implication of this analytical strategy for application in research on psychiatric disorders is discussed. © 2000 Academic Press Key Words: fMRI; working memory; parametric design; n-back task; prefrontal cortex; parietal cortex; anterior cingulate.

1

To whom correspondence should be addressed. Fax: 31-30-2505443. E-mail: [email protected]. 1053-8119/00 $35.00 Copyright © 2000 by Academic Press All rights of reproduction in any form reserved.

The term working memory, as defined by Baddeley (1992) refers to a brain system that provides temporary storage and manipulation of information necessary to execute complex cognitive tasks. It is a vital function for human beings to interact with the environment in a meaningful way, for instance to store information about what one has done in the recent past (seconds), to temporarily store information about the present situation, or to process this information to make plans for the near future. Abnormal working memory function has been linked to psychiatric disorders such as schizophrenia by neuropsychological measures of cognitive deficits (Aleman et al., 1999; Goldman Rakic, 1994; Goldman Rakic and Selemon, 1997; Park and Holzman, 1992; Weinberger et al., 1992). Several imaging studies have implicated a link between cognitive deficits and regions within prefrontal cortex (Callicott et al., 1998; Kindermann et al., 1997; Weinberger et al., 1996). However, interpretation of imaging results in patient studies is not straightforward as it is complicated by the following two issues. First, performance on various cognitive tasks is generally worse in patients than in healthy controls. Deviant brain activation patterns are often interpreted as a correlate of a specific cognitive deficit, but they may alternatively reflect a general slowing of motor responses which interferes with stimulus processing, if a task requires continuous responding to sequentially presented stimuli. Second, factors that affect the functional imaging measurements, such as artifacts generated by increased head movement in patients compared to healthy controls or generalized effects of medical treatment on brain physiology and on psychophysiological variables, such as vigilance, may contribute to differences in brain activity between groups. These problems can be addressed by incorporating an internal reference in the design of the experiment. For instance, differences in activity related to primary functions such as visual processing and the

688

689

BRAIN ACTIVITY IN A PARAMETRIC N-BACK TASK

generation of finger movements may indicate that functions other than the cognitive function of interest are affected, and this could form an alternative explanation of differences observed in performance and in the associated brain activity maps. An alternative approach makes use of a parametric design. Several functional imaging studies have shown that the fMRI signal in regions in the prefrontal and parietal cortex increases as the demands made on working memory increase (Cohen et al., 1994; Manoach et al., 1997; Rypma et al., 1999; Smith et al., 1998). Parametric manipulation of this demand within an fMRI task-design makes it possible to distinguish cerebral activity that is sensitive for workload, or “workload specific,” from activity that is not sensitive to workload, i.e., “nonspecific.” The main focus of this study is to present a method to simultaneously measure and separate workloadspecific from nonspecific activity. In contrast with other parametric studies on working memory (Braver et al., 1997; Callicott et al., 1999; Smith et al., 1998) the current approach separates workload-specific (also referred to as “load-sensitive”), from nonspecific (or “load-insensitive”) activity while preserving the original high spatial resolution of the fMRI signal. To accomplish this, voxels are analyzed within individual subjects, avoiding confounds associated with image smoothing and analyzing activation maps across subjects. This method is chosen above a typical groupwise analysis because the latter does not take into account the possibility that workload-specific and nonspecific activity may be limited to small regions in close proximity to one another, but with considerable topographical variation across individual subjects. In this study a spatial variant of an n-back task was used to induce four levels of working memory load in a parametrically controlled design. The various components of which a parametric task consists can ideally be divided into two categories: those that are not parametrically varied and those that are. The former elicit brain activity that is not specific for the cognitive function of interest, hence called nonspecific processes, and the latter elicit specific processes. The complexity of cognitive processing makes it difficult to distinguish between the two categories on a theoretical basis. Proper examination of the functional imaging results however allows for empirical distinction between the two types of processes. In this study an analytical approach is presented that allows for such a distinction. The 3D-PRESTO fMRI scan technique (Gelderen et al., 1995; Ramsey et al., 1998) was used for imaging of brain activity. As functional imaging studies have demonstrated involvement of dorsolateral prefrontal cortex (Courtney et al., 1998a; d’Esposito et al., 1998; Smith and Jonides, 1997), parietal cortex (Jonides et

FIG. 1. Illustration of the n-back task showing five trials of the 0-back and 3-back level, together with the appropriate responses.

al., 1998), and anterior cingulate, (Barch et al., 1997; Paus et al., 1998) in working memory, these are the areas that we have focused on in our analysis. METHOD General Procedure Twelve healthy volunteers participated in the experiment (eight male, four female, right handed: Edinburgh handedness inventory (Oldfield, 1971): 0.98 (SEM: 0.02), with an average age of 27 years (SEM: 5.0)). Subjects were predominantly college students who were recruited from the university. Participants signed an informed consent approved by the ethical committee of the University Medical Center Utrecht. Subjects were trained for 1 h prior to the experiment. A video-projector was placed in the scanner control room to project stimuli from outside the control-room to a screen in front of the volunteer. A mirror enabled the subjects to see the screen. Subjects had to perform a spatial variant of the nback working memory task with increasing levels of difficulty. This task was designed after Gevins (Gevins and Cutillo, 1993) and has been used in several previous fMRI-experiments (Callicott et al., 1998, 1999; Casey et al., 1998). Subjects looked at a screen, which showed four large dots in a square indicating the four possible places where a stimulus (a number) could appear (see Fig. 1). Subjects were instructed to respond to the stimulus by pushing one of four buttons with the right thumb. Layout of the four buttons corresponded spatially to the four possible positions in which the stimulus appeared. Responses had to be made either directly following the stimulus (“0-back”) or with a delay of one (“1-back”), two (“2-back”), or three (“3back”) stimuli. The task-level was shown above the stimuli during the whole task. Responses were re-

690

JANSMA ET AL.

FIG. 2. Task order used in the experiment (R ⫽ resting baseline; 0, 1, 2, 3⫽ 0, 1, 2, 3-back level) and a visual representation of the factors used in the multiple regression analysis. Factors 2 and 3 model machine-related signal drifts (intensity in factors 2 and 3 reflect the load on the factor). Factor 1 is used for selecting voxels of interest in the mapping analysis (dark, rest; light, 3 back; white, not included in analysis) and in the signal analysis (from dark to light, 0, 1, 2, 3 back; white, not included).

corded through a pneumatic response buttonbox. The interstimulus interval was 2.2 s. The design of the n-back task is such that in all levels, subjects have to respond to all stimuli, so the task requires a continuous monitoring and updating of information in working memory. The information that is presented during each trial is the same within all levels, as are the demands made on response selection and execution. In contrast to other variants of the n-back task, in this task only error rate is a measure of workload or performance, and not reaction time. In all levels, except for the 0-back level, subjects know what the next response will be before they are allowed to respond, as it is determined by preceding stimuli. Therefore, subjects were instructed to emphasize accuracy rather than speed, and the task program did not record reaction times. To show the dissociation between reaction time and workload, we conducted a separate experiment with 4 male and 4 female subjects (mean age 29 years) outside the scanner. This study showed no effect of workload level on reaction time (general linear

model, repeated measurements, F 1,7 ⫽ 3.04; P ⫽ 0.13, see Fig. 3a) confirming the dissociation. fMRI The BOLD-sensitive 3D-PRESTO fMRI pulse-sequence was used (Gelderen et al., 1995; Liu et al., 1993; Ramsey et al., 1998), on a 1.5 Tesla Philips Gyroscan ACS-NT standard clinical scanner (parameters: TE/TR 35/24 ms; flip angle 9°; FOV 225*175*70; matrix 64*50*20; slice thickness: 3.51 mm; scan time: 5.8 s). Voxelsize was 3.51 mm isotropic (i.e., 0.043 cc). To compensate for the delay of the vascular response to neuronal activity (Ogawa et al., 1993), execution of the task started 6 s before the first functional scan. fMRI data were acquired in 8 blocks of 50 scans (5.12 min). During each block, each level (rest, 0-back, 1-back, 2-back, 3-back) was performed two times for periods of 29 s (for sequence: see Fig. 2), resulting in five volume scans per task period. During the rest condition no stimuli were presented and subjects simply looked at the background image of the task. The sequence of the

FIG. 3. (a) Reaction times averaged over correct response trials per level for eight subjects recorded separately from scan sessions. Results show that in the used version of the n-back task, reaction time is not correlated with level, indicating that speed of response execution and selection is not affected by workload in the spatial n-back task. Error bars represent standard deviations. (b) Error rate per task level during scans. Calculated within subject as the percentage errors and omissions during the whole scansession. Results are averaged over all subjects. Error bars represent standard deviations.

BRAIN ACTIVITY IN A PARAMETRIC N-BACK TASK

task-levels was partly randomized: only levels 2- and 3-back were randomized but not the other levels. This was done to avoid large variability of intervals between rest and level 3, for the benefit of an originally planned subtraction analysis (which is replaced by a more powerful multiple regression analysis, see below). A high contrast functional image (flip angle 30°) was acquired for registration. Each session was concluded with an anatomical scan, which was later used as background to determine the location of activity (parameters: 3D-FFE; TE/TR 4.6/30 ms; flip angle 30; FOV 256*256, matrix 256*256*110; slice-thickness: 1.2; scan time: 7 min). Data Analysis Performance on the four task levels is expressed as percentage, of required responses, that was incorrect or omitted. The 16 repetitions of each task level were averaged within subjects, to obtain one score on every task. fMRI data were processed off-line on a HP-workstation using PV-wave processing software. All volumeimages were registered to one high contrast functional scan to correct for rotation and translation between scans (Thevenaz et al., 1998). Scans were normalized for mean volume signal-intensity. For the fMRI dataanalysis a multiple regression algorithm (Worsley and Friston, 1995) with three factors was used, i.e., one for task-level, and two to model machine-related signal drifts (see Fig. 2). Multicollinearity of the multiple regression factor matrix was well within acceptable limits, (variance inflation factor (VIF) was lower then 1.5 for each regression factor, see for instance Montgomery and Peck, 1982). The algorithm differs from that of Worsley and Friston in that no temporal or spatial smoothing was applied. Data-sets of each individual were analyzed separately to avoid loss of spatial detail in detecting load-sensitive and load insensitive activity. First, a “map-analysis” was performed to identify all areas that were activated during execution of the task. This step is analogous to a standard subtraction method. To maximize power of detecting activity, the map-analysis included only the scans made during the rest condition and the scans made during the level with the highest mental load, i.e., the 3-back level (a total of 80 scans). A Bonferroni-correction was used to correct for the total number of comparisons made, i.e., the total number of brain-voxels in the functional scan. A second analysis served to select voxels that were specifically related to working memory. In this “signalanalysis,” the voxels selected in the map-analysis were tested for a positive linear correlation with task-level. All scans were included that were made during the execution of the n-back task, thus excluding the scans made during the rest-condition. A linear input func-

691

tion, correlating with task-level (0, 1, 2, 3 for the 0-back, 1-back, 2-back, and the 3-back level, respectively, see Fig. 2) was used to select voxels showing a significant effect of level. A Bonferroni-correction was used to correct for the total number of comparisons made (i.e., the set of voxels that was selected with the map-analysis). The following volumes of interest (VOIs) were determined for each subject, using the brain atlas of Duvernoy et al. (1991), guided by the classification system of Brodmann: left and right dorsolateral prefrontal-cortex (superior- and middle-frontal gyrus, Brodmann area BA 46, 9, 8), the anterior cingulate (BA 24, 32), the left and right parietal cortex (area within post central sulcus, parieto-occipital sulcus and temporooccipital sulcus, BA 7a, 7b, 39, 40), and the left primary sensorimotor-cortex (precentral and postcentral gyrus, BA 1–5). Brodmann area indications are only approximations, as his classification system is based on cytoarchitecture rather than sulcal landmarks, and is not three-dimensional. The VOI sizes are relatively large, but this is not uncommon in studies using individual activity maps (see for instance Zarahn et al., 2000). Large VOIs are chosen to take functional topographical variability across subjects into account. Individual unsmoothed activation maps typically show large differences between subjects. Smaller VOIs would become less meaningful, as the number of subjects showing activation in a VOI would become much smaller. Voxels with significant load on the task-level factor in the signal analysis are referred to as load-sensitive voxels and the other voxels from the map-analysis selection as load-insensitive voxels. Within each voxel, the percentage signal-change was determined for each level, relative to resting baseline. Presented results are averaged within subjects across active voxels in each VOI, and over subjects that showed active voxels in that VOI. To verify effective distinction between loadsensitive and load-insensitive activity, within each VOI the signal increase in the load-sensitive voxels was compared to that in the load-insensitive voxels with a general linear model for repeated measurements (linear contrast, interaction between level of task (0- through 3-back), and signal in each of the task levels). In this post-hoc analysis, only subjects with both load-sensitive and load-insensitive activity within a VOI were included, as both variables are required for statistical comparison. Because averaging of the signals is solely done with the purpose of testing the assumption that the load sensitive voxels show a generally different signal than the load-insensitive voxels, we only included signals from the selected voxels. Relationships between performance on the task and various measures of regional brain activity were investigated in an exploratory manner. To this purpose, performance decrement was compared to numbers of load-sensitive and load-insensitive voxels in the vari-

692

JANSMA ET AL.

TABLE 1

s1 s2 s3 s4 s5 s6 s7 s8 s9 s10 s11 s12

Whole brain

DLPFC left

Perf

LI

LS

LI

LS

LI

LS

LI

LS

LI

LS

LI

LS

LI

LS

⫺12,3 ⫺16,5 ⫺4,7 ⫺0,3 ⫺1,2 0,8 ⫺0,8 ⫺0,9 ⫺17,3 ⫺14,2 ⫺3,8 ⫺2,8

132 42 7 104 23 161 110 22 121 40 159 78

19 39 58 83 17 196 25 73 9 12 99 78

7 0 0 8 0 10 4 0 5 1 1 2

4 0 0 9 0 20 0 8 0 1 7 0

4 4 0 20 1 11 1 0 1 0 0 10

0 0 8 3 5 15 0 21 0 1 0 5

3 2 0 0 4 9 1 7 2 3 3 19

1 5 0 6 1 16 0 4 0 0 2 4

48 4 0 6 9 25 17 0 30 7 26 5

9 21 4 33 4 45 11 5 2 4 31 19

13 8 2 3 2 16 3 0 22 8 6 20

3 6 30 22 0 71 7 10 6 3 15 36

33 5 0 24 2 47 58 12 25 4 67 6

1 1 1 2 0 8 1 4 0 0 7 0

83.3 15.9

59 15.4

3.17 1.03

4.08 1.76

4.33 1.81

4.83 1.96

4.42 1.53

3.25 1.46

14.8 4.26

15.7 4.09

8.58 2.16

17.4 5.85

23.6 6.69

2.08 0.80

Mean SEM

DLPFC right

AC

PC left

PC right

PSMC left

Note. Performance decrement (perf, percentage of correct responses in 3-back minus that in 0-back); number of load-sensitive (LS) and load-insensitive (LI) voxels, together with the mean (all subjects included) per subject for whole brain, left and right dorsolateral prefrontal cortex (DLPFC), left and right parietal cortex (PC), anterior cingulate (AC), and left primary sensorimotor cortex (PSMC).

ous VOIs. To reduce large between-subject variations in these numbers, they were corrected for total number of active voxels in the brain (obtained in the first multiple regression analysis) prior to correlation analyses. Performance decrement was calculated by subtracting the percentage of correct responses (relative to the required responses) made in the 0-back level from that in the 3-back level. Given the exploratory nature, these analyses were not corrected for multiple comparisons. RESULTS General Results All 12 subjects activated the dorsolateral prefrontal cortex in the map-analysis (i.e., 3-back versus rest). Three subjects showed left lateralized and two subjects showed right-lateralized dorsal-prefrontal activity (see Table 1). All 12 subjects showed parietal activity in the map-analysis. Eleven subjects activated the parietal cortex bilaterally and one subject showed left-lateralized parietal activity. The anterior cingulate activated in eleven subjects in the map-analysis, as did the left primary sensorimotor cortex. With the signal-analysis, load-sensitive voxels were found in all subjects (see Table 1). In the dorsolateral prefrontal cortex, load-sensitive activity was left lateralized in three subjects, right lateralized in three subjects, and bilateral in four subjects. Two subjects did not show load-sensitive voxels in this region. In the parietal cortex, load-sensitive activity was bilateral in 11 subjects and left lateralized in one subject. The anterior cingulate showed load-sensitive activity in eight subjects. In Fig. 4a comparison of the numbers of load-sensi-

tive and load-insensitive voxels in each VOI is presented. The largest number of load-sensitive voxels was found in the parietal cortices. The only region where the difference between numbers of load-sensitive and load-insensitive voxels was significant, was the left primary sensorimotor cortex (Wilcoxon matched pairs, Z ⫽ 2,67; P ⬍ 0.01). An example of the distribution of load-sensitive and load-insensitive activity in a single subject is shown in Fig. 7. Relation between Signal and Task-Level Concurrent load-sensitive and load-insensitive activity occurred in most subjects, in left and right parietal cortex (10 of 12 subjects). Concurrence of load-sensitive and load-insensitive activity was observed in seven subjects in the anterior cingulate, in five subjects in the

FIG. 4. Number of load-sensitive and load-insensitive voxels for each VOI, averaged across all subjects (plus standard error of the mean) (*P ⬍ 0.01, difference between number of load-sensitive and load-insensitive voxels).

BRAIN ACTIVITY IN A PARAMETRIC N-BACK TASK

693

FIG. 5. Percentage signal increase relative to resting-baseline for the six VOIs and the four levels of the task, averaged across all subjects that showed at least one significant voxel in the VOI (straight line, load-sensitive voxels; dotted line, load-insensitive voxels. PC, parietal cortex; DLPFC, dorsolateral prefrontal cortex; AC, anterior cingulate; PSMC, primary sensorimotor cortex).

left prefrontal cortex and in four subjects in the right prefrontal cortex (see Table 1). The linear increase in signal in load-sensitive voxels was significantly higher than that in the load-insensitive voxels, in all examined VOIs. The difference in slope was highest in the right parietal cortex F 1,11 ⫽ 155.9; P ⬍ 0.001), followed by the left parietal cortex (F 1,11 ⫽ 48.2; P ⬍ 0.001), the anterior cingulate (F 1,5 ⫽ 28.9; P ⬍ 0.005), the left prefrontal (F 1,3 ⫽ 16.6; P ⬍ 0.05), and the right prefrontal cortex (F 1,3 ⫽ 12.0; P ⬍ 0.05). The initial signal increase in the load-sensitive voxels, i.e., that at the 0-back level, also differed from that in the load-insensitive voxels, which is illustrated in Fig. 5. The load-insensitive voxels show a larger signal increase from resting baseline to 0-back than loadsensitive voxels. Relation between Task Performance and Activity All subjects performed all levels of the task with low error rates. The highest individual error rate in the 3-back level remained under 20%. Group averaged performance data reveal a main effect of level on error rate from 0-back to 3-back (1.1, 1.9, 5.0, and 7.8%; MANOVA: F 1,11 ⫽ 10.48, P ⬍ 0.01), reflecting a higher

mental load with increasing task-level (see Fig. 3b). Individual performance rates differed from a small decrease in error rate (⫺0.8%) to large increases in error rate (17.3%) from 0-back to 3-back (see Table 1). When examining the signal change in each VOI, no significant correlations were found with performance across subjects. When the numbers of load-sensitive and load-insensitive voxels in each VOI were analyzed, two effects were found. The first effect was a correlation between performance and relative number of loadsensitive voxels in the anterior cingulate cortex (rho ⫽ 0.70, P ⬍ 0.01). The second effect involved the right parietal cortex, where the relative number of loadinsensitive voxels, correlated negatively with performance (rho ⫽ ⫺0.68; P ⬍ 0.01). A better performance was correlated with a larger area of load-sensitive activity in anterior cingulate and with a smaller area of load-insensitive activity in the right parietal cortex (see Fig. 6). DISCUSSION This study examined cerebral activity in response to graded demands on working memory, in terms of loadsensitivity and load-insensitivity, using an n-back work-

694

JANSMA ET AL.

FIG. 6. Correlation of load sensitive activity in anterior cingulate (left graph) and load insensitive activity with performance in right parietal cortex (right graph). Load-sensitive and load-insensitive activity are presented as percentage of active voxels in map-analysis. Performance is presented as percent correct responses in 3 back minus that in 0 back.

ing memory task with four levels. Analyses were performed on individual subjects, allowing for voxel-by-voxel tests of load-sensitive and load-insensitive activity. An extended network of working memory related activity was identified in all subjects. This network consisted of both load-sensitive and load-insensitive activity bilaterally in dorsolateral prefrontal cortices, parietal cortices, and in anterior cingulate. The analytically obtained distinction between load-sensitive and load-insensitive brain activity was confirmed with post-hoc comparisons of signal increase patterns, in all examined VOIs.

The distribution of load-sensitive activity agrees with previous publications, in which bilateral loadcontingent activity was reported in superior middle and frontal gyrus, and inferior or superior parietal cortex (Braver et al., 1997; Callicott et al., 1999; Cohen et al., 1994; Rypma et al., 1999; Smith et al., 1998). In contrast, those studies did not report load-sensitive activity in anterior cingulate although this area has been implicated in task difficulty (Paus et al., 1998). A possible explanation may be that overall performance in our study was at a high level and, as our results indicate, load-sensitive activity is correlated to perfor-

FIG. 7. Illustration of the distribution of load-sensitive (red squares) and load-insensitive (yellow squares) activity in a single subject (transformed to Talairach orientation for display purpose only). The outline of the functional scan is shown (green line) on top of the anatomical background scan. (a) Transaxial slices, with the orientation of sagittal and coronal slices shown as blue lines. (b) Sagittal slices; (c) Coronal slices.

BRAIN ACTIVITY IN A PARAMETRIC N-BACK TASK

mance. The presence of both load-insensitive and loadsensitive voxels, in all of the examined VOIs, suggests that specific working memory processes and nonspecific processes take place in these areas, a finding which has been previously reported for prefrontal cortex (d’Esposito et al., 1998). None of the VOIs examined, except the left primary sensorimotor cortex, were associated exclusively with load-insensitive activity. This supports the notion that both supportive and working memory-specific functions are colocalized in various parts of the brain. As mentioned in the method section, the VOIs in the current study were relatively large, particularly when considering reports on functional heterogeneity in prefrontal and parietal cortex (Corbetta et al., 2000; Courtney et al., 1998b; d’Esposito et al., 1998; Gitelman et al., 1999; Hopfinger et al., 2000; Ungerleider et al., 1998). Most of these reports however have made use of group-averaging methods, and as such disregard heterogeneity of functional topography across subjects. Such heterogeneity has been reported, even for the primary motor cortex (Penfield and Boldrey, 1937), and should therefore be considered when a within-subject voxel-by-voxel analysis is performed, as in the current study. Individual activity patterns in this study did not show a topographic consistency of activation across subjects that would have made smaller VOIs meaningful. One could argue that not all load-sensitive voxels exhibit the same b-level (slope of signal increase), within our chosen VOIs. While this is possible, and might be looked into, this was not the scope of this study. The main goal of this study was to separate load sensitive from load-insensitive activity on a voxel by voxel basis. Increased demands on working memory can be expected to lead to increased cerebral activity, as long as a certain level of performance can be maintained. If this is not the case, and performance drops dramatically, then one would expect to find a commensurate loss of activity in areas that mediate working memory function either because maximum processing capacity has been exceeded, or because some brain regions that are critical for this function fail to contribute. In the current study, performance was maintained at a high level, indicating that working memory was engaged successfully at all task levels in all subjects. Across subjects, good performance was associated with the extent of load-sensitive activity in the anterior cingulate (Fig. 6). The correlation indicates that the ability to maintain high performance on this task across levels is linked to increased engagement of anterior cingulate cortex. An extensive review (Paus et al., 1998) has shown that activity in the anterior cingulate is strongly related to task difficulty. Reportedly, increased activity is not directly associated with increased demands on working memory, but to a more general arousal effect. We did not find significant cor-

695

relations of performance with number of active loadsensitive voxels in other VOIs. In the right parietal cortex, good performance was associated with a reduced extent of load-insensitive activity (Fig. 6). This may be associated with application of an efficient or effective task execution strategy. It may be postulated that an efficient strategy is associated with adequate dynamic engagement of anterior cingulate, while an inefficient strategy imposes increased claim on parietal support systems with concomitant reduced engagement of anterior cingulate. Hemispheric dominance was not observed, neither for load-sensitive nor for load-insensitive activity. Individual differences in lateralisation were large, but none of the subjects showed completely lateralized activity. In literature, evidence for hemispheric dominance in working memory is not convincing, as both left dorsal-prefrontal dominance (Braver et al., 1997; Demb et al., 1995) and right hemispheric dominance (Buckner et al., 1996; McCarthy et al., 1994) have been reported. In fact, most studies report bilateral prefrontal and parietal activity (Barch et al., 1997; Braver et al., 1997; Cohen et al., 1997; Courtney et al., 1997; Jonides et al., 1998), and this is confirmed by the present findings. Load-sensitive activity is related to the increasing demands made on working memory when the task becomes more difficult. Conceptually, the working memory load in the n-back task affects three different processes, i.e., active temporary retention of information, rehearsal of information held in temporary retention (Awh et al., 1996), and retention of the temporal characteristics of the stimulus (Fuster, 1980). The demand for temporary retention of information in the n-back task is moderate (i.e., three items at most), and activity in working memory regions is reportedly more sensitive to manipulation of information than to temporary retention of information (Smith et al., 1998). From this point of view, load-sensitive activity in the n-back task may reflect increased manipulation of stimuli, rather than an increased load on temporary retention of stimuli. One could argue that load-insensitive activity might simply be the consequence of a partial volume effect (i.e., averaging with signal from adjacent nonactive tissue) on load-sensitive activity, instead of reflecting two different types of processing. However, signal increase in the 0-back level, relative to rest, in loadinsensitive voxels is higher than that in load-sensitive voxels. As this finding is not compatible with the partial volume notion, it is unlikely that such effects can explain the presence of load-insensitive voxels. Another possible explanation of load-sensitive signal change is that with increasing levels in cognitive tasks, reaction times generally increase, causing an increase in activity in areas that subserve motor-response programs (due to response delays). As responses are de-

696

JANSMA ET AL.

termined by preceding stimuli, and not by the latest stimulus, reaction time was not affected by workload (see Fig. 3a). Accordingly, activity associated with motor program does not increase with workload and therefore cannot explain load-sensitive signal increase in the current study. The present experiment does not allow for differentiation among processes or functions that are loadsensitive, such as holding stimuli on-line, and rehearsal of these stimuli. Topographical distinction between functions has been postulated in other studies where associations were reported between parietal activity and perceptive memory and between prefrontal activity and verbal rehearsal of the material that is held on-line (Awh et al., 1996; Smith et al., 1998). The current finding of load-sensitive activity in prefrontal cortex suggests that it is additionally involved in the process of sustaining or manipulating spatial representations of stimuli that are stored in the parietal cortex, for subsequent (delayed) use. The present design also does not address the issue of separating the various non-specific processes that contribute to load-insensitive activity, such as visual orienting, stimulus perception, encoding and processing, and response selection or execution. This would require multiple dedicated task designs with appropriate control tasks. The focus of the present study was not so much on identification of neuroanatomical substrates of the various functions, but on the separation of specific and nonspecific workload processes, to provide a means of further investigating the specificity of the types of cognitive processes that are associated with cognitive deficits in psychiatric disorders. This study demonstrates that activity of some brain structures in frontal and parietal cortex correlate with mental load, while activity in other structures remains constant, irrespective of mental load. It also demonstrates benefits of preserving high spatial resolution, and of emphasizing individual signal-analysis, as the process of smoothing images to obtain group averages would inevitably decrease differences in load-sensitivity between voxels. Especially for higher order cognitive functions, the loci of activity may not be as consistent over subjects as for lower order functions such as motor functions. Although it is not possible to specify the processes reflected by load-sensitive or load-insensitive voxels, the concurrence of both types of activity in all examined VOIs, does indicate that nonspecific, or supportive processes and specific working memory processes co-occur in multiple regions. This interpretation of our results is in line with earlier studies, in which evidence for a functional anatomical separation of working memory “slave” systems, and executive control-centers was not found (Cohen et al., 1997; d’Esposito et al., 1995). The presented approach allows one to deal with a dilemma in patient studies, where both performance

and cerebral activity are affected in unison. A simple mapping analysis showing deviant cerebral activity in, for instance, a schizophrenic patient, reflects the overall task-execution, but does not allow for a distinction between deficits in specific processes (e.g., working memory) and deficits in nonspecific processes (e.g., stimulus perception, motor response, etc.). By using a multiple level-task, one can screen for activity at a well-performed level, and then use the other levels to examine the relationship between the signal in involved voxels on the one hand and performance on the other hand. Examination of load-sensitive activity allows one to examine if a poor performance with higher demands is directly related to a capacity problem, such as an ineffective use of limited working memory resources. Load-insensitive activity provides a “control” measure, addressing problems in execution of nonspecific task-related processes, such as perception, encoding of information, selecting and executing responses, as well as general effects of medical treatment or increased movement. ACKNOWLEDGMENTS This research was supported in part by Solvay Duphar BV. and by the University Medical Center Utrecht.

REFERENCES Aleman, A., Hijman, R., De Haan, E. H., and Kahn, R. S. 1999. Memory impairment in schizophrenia: A meta-analysis. Am. J. Psychiatry 156: 1358 –1366. Awh, E., Jonides, J., Smith, E. E., Schumacher, E. H., Koeppe, R. A., and Katz, S. 1996. Dissociation of storage and rehearsal in verbal working memory: Evidence from positron emission tomography. Psychol. Sci. 7: 25–31. Baddeley, A. 1992. Working memory. Science 255: 556 –559. Barch, D. M., Braver, T. S., Nystrom, L. E., Forman, S. D., Noll, D. C., and Cohen, J. D. 1997. Dissociating working memory from task difficulty in human prefrontal cortex. Neuropsychologia 35: 1373–1380. Braver, T. S., Cohen, J. D., Nystrom, L. E., Jonides, J., Smith, E. E., and Noll, D. C. 1997. A parametric study of prefrontal cortex involvement in human working memory. NeuroImage 5: 49 – 62. Buckner, R. L., Raichle, M. E., Miezin, F. M., and Petersen, S. E. 1996. Functional anatomic studies of memory retrieval for auditory words and visual pictures. J. Neurosci. 16: 6219 – 6235. Callicott, J. H., Mattay, V. S., Bertolino, A., Finn, K., Coppola, R., Frank, J. A., Goldberg, T. E., and Weinberger, D. R. 1999. Physiological characteristics of capacity constraints in working memory as revealed by functional MRI. Cereb. Cortex 9: 20 –26. Callicott, J. H., Ramsey, N. F., Tallent, K., Bertolino, A., Knable, M. B., Coppola, R., Goldberg, T., and Weinberger, D. R. 1998. Functional magnetic resonance imaging brain mapping in psychiatry: Methodological issues illustrated in a study of working memory in schizophrenia. Neuropsychopharmacology 18: 186 –196. Casey, B. J., Cohen, J. D., O’Craven, K., Davidson, R. J., Irwin, W., Nelson, C. A., Noll, D. C., Hu, X., Lowe, M. J., Rosen, B. R., Truwitt, C. L., and Turski, P. A. 1998. Reproducibility of fMRI results across four institutions using a spatial working memory task. NeuroImage 8: 249 –261.

BRAIN ACTIVITY IN A PARAMETRIC N-BACK TASK Cohen, J. D., Forman, S. D., Braver, T. S., Casey, B. J., Servan Schreiber, D., and Noll, D. C. 1994. Activation of the prefrontal cortex in a nonspatial working memory task with functional mri. Hum. Brain Mapp. 293–304. Cohen, J. D., Perlstein, W. M., Braver, T. S., Nystrom, L. E., Noll, D. C., Jonides, J., and Smith, E. E. 1997. Temporal dynamics of brain activation during a working memory task [see Comments]. Nature 386: 604 – 608. Corbetta, M., Kincade, J. M., Ollinger, J. M., McAvoy, M. P., and Shulman, G. L. 2000. Voluntary orienting is dissociated from target detection in human posterior parietal cortex. Nature Neurosci. 3: 292–297. Courtney, S. M., Petit, L., Haxby, J. V., and Ungerleider, L. G. 1998a. The role of prefrontal cortex in working memory: Examining the contents of consciousness. Phil. Trans. R. Soc. Lond. B Biol. Sci. 353: 1819 –1828. Courtney, S. M., Petit, L., Maisog, J. M., Ungerleider, L. G., and Haxby, J. V. 1998b. An area specialized for spatial working memory in human frontal cortex. Science 279: 1347–1351. Courtney, S. M., Ungerleider, L. G., Keil, K., and Haxby, J. V. 1997. Transient and sustained activity in a distributed neural system for human working memory [see Comments]. Nature 386: 608 – 611. d’Esposito, M., Aguirre, G. K., Zarahn, E., Ballard, D., Shin, R. K., and Lease, J. 1998. Functional MRI studies of spatial and nonspatial working memory. Cogn. Brain Res. 1–13. d’Esposito, M., Detre, J. A., Alsop, D. C., Shin, R. K., Atlas, S., and Grossman, M. 1995. The neural basis of the central executive system of working memory. Nature 378: 279 –281. Demb, J. B., Desmond, J. E., Wagner, A. D., Waidya, C. J., Glover, G. H., and Gabrieli, J. D. E. 1995. Semantic encoding and retrieval in the left inferior prefrontal cortex: A functional MRI study of task difficulty and process specificity. J. Neurosci. 15: 5870 –5878. Duvernoy, H. M., Canabis, E. A., Iba-Zizen, M. T., Tamraz, J., and Guyot, J. 1991. The Human Brain: Surface, Three-Dimensional Sectional Anatomy and MRI. Springer-Verlag, New York. Fuster, J. M. 1980. Outline of a theory: The prefrontal cortex and the temporal integration of behavior. In The prefrontal cortex; anatomy, physiology, and neuropsychology of the frontal lobe pp. 125– 142. Raven Press, New York. Gelderen, P. V., Ramsey, N. F., Liu, G., Duyn, J. H., Frank, J. A., Weinberger, D. R., and Moonen, C. T. 1995. Three-dimensional functional magnetic resonance imaging of human brain on a clinical 1.5-T scanner. Proc. Natl. Acad. Sci. USA 92: 6906 – 6910. Gevins, A., and Cutillo, B. 1993. Spatiotemporal dynamics of component processes in human working memory. Electroencephalogr. Clin. Neurophysiol. 87: 128 –143. Gitelman, D. R., Nobre, A. C., Parrish, T. B., LaBar, K. S., Kim, Y., Meyer, J. R., and Mesulam, M. 1999. A large-scale distributed network for covert spatial attention. Brain 122: 1093–1106. Goldman Rakic, P. S. 1994. Working memory dysfunction in schizophrenia. J. Neuropsychiatry Clin. Neurosci. 6: 348 –357. Goldman Rakic, P. S., and Selemon, L. D. 1997. Functional and anatomical aspects of prefrontal pathology in schizophrenia [see Comments]. Schizophren. Bull. 23: 437– 458. Hopfinger, J. B., Buonocore, M. H., and Mangun, G. R. 2000. The neural mechanisms of top-down attentional control. Nature Neurosci. 3: 284 –291. Jonides, J., Schumacher, E. H., Smith, E. E., Koeppe, R. A., Awh, E., Reuter Lorenz, P. A., Marshuetz, C., Willis, C. R. 1998. The role of parietal cortex in verbal working memory. J. Neurosci. 18: 5026 – 5034. Kindermann, S. S., Karimi, A., Symonds, L., Brown, G. G., and Jeste, D. V. 1997. Review of functional magnetic resonance imaging in schizophrenia. Schizophren. Res. 27: 143–156.

697

Liu, G., Sobering, G., Duyn, J., and Moonen, C. T. 1993. A functional MRI technique combining principles of echo-shifting with a train of observations (PRESTO). Magn. Reson. Med. 30: 764 –768. Manoach, D. S., Schlaug, G., Siewert, B., Darby, D. G., Bly, B. M., Benfield, A., Edelman, R. R., and Warach, S. 1997. Prefrontal cortex fMRI signal changes are correlated with working memory load. Neuroreport 8: 545–549. McCarthy, G., Blamire, A. M., Puce, A., Nobre, A. C., Bloch, G., Hyder, F., Goldman Rakic, P., and Shulman, R. G. 1994. Functional magnetic resonance imaging of human prefrontal cortex activation during a spatial working memory task. Proc. Natl. Acad. Sci. USA 91: 8690 – 8694. Montgomery, D., and Peck, E. 1982. Multicollinearity. In Introduction to Linear Regression Analysis, pp. 299 –300. Wiley, New York. Ogawa, S., Menon, R. S., Tank, D. W., Kim, S. G., Merkle, H., Ellermann, J. M., and Ugurbil, K. 1993. Functional brain mapping by blood oxygenation level-dependent contrast magnetic resonance imaging. A comparison of signal characteristics with a biophysical model. Biophys. J. 64: 803– 812. Oldfield, R. C. 1971. The assessment and analysis of handedness: The Edinburgh inventory. Neuropsychologia 2: 97–113. Park, S., and Holzman, P. S. 1992. Schizophrenics show spatial working memory deficits. Arch. Gen. Psychiatry 49: 975–982. Paus, T., Koski, L., Caramanos, Z., and Westbury, C. 1998. Regional differences in the effects of task difficulty and motor output on blood flow response in the human anterior cingulate cortex: A review of 107 PET activation studies. Neuroreport 9: 37– 47. Penfield, W., and Boldrey, E. 1937. Somatic motor and sensory representation in the cerebral cortex of man as studied by electrical stimulation. Brain 60: 389 – 443. Ramsey, N. F., van den Brink, J. S., van Muiswinkel, A. M., Folkers, P. J., Moonen, C. T., Jansma, J. M., and Kahn, R. S. 1998. Phase navigator correction in 3D fMRI improves detection of brain activation: Quantitative assessment with a graded motor activation procedure. NeuroImage 8: 240 –248. Rypma, B., Prabhakaran, V., Desmond, J. E., Glover, G. H., and Gabrieli, J. D. 1999. Load-dependent roles of frontal brain regions in the maintenance of working memory. NeuroImage 9: 216 –226. Smith, E. E., and Jonides, J. 1997. Working memory: A view from neuroimaging. Cogn. Psychol. 33: 5– 42. Smith, E. E., Jonides, J., Marshuetz, C., and Koeppe, R. A. 1998. Components of verbal working memory: Evidence from neuroimaging. Proc. Natl. Acad. Sci. USA 95: 876 – 882. Thevenaz, P., Ruttimann, U. E., and Unser, M. 1998. A pyramid approach to subpixel registration based on intensity. IEEE Trans. Med. Imag. 7: 27– 41. Ungerleider, L. G., Courtney, S. M., and Haxby, J. V. 1998. A neural system for human visual working memory. Proc. Natl. Acad. Sci. USA 95: 883– 890. Weinberger, D. R., Berman, K. F., Suddath, R., and Torrey, E. F. 1992. Evidence of dysfunction of a prefrontal-limbic network in schizophrenia: A magnetic resonance imaging and regional cerebral blood flow study of discordant monozygotic twins. Am. J. Psychiatry 149: 890 – 897. Weinberger, D. R., Mattay, V., Callicott, J., Kotrla, K., Santha, A., van Gelderen, P., Duyn, J., Moonen, C., and Frank, J. 1996. fMRI applications in schizophrenia research. NeuroImage 4: S118 – S126. Worsley, K. J., and Friston, K. J. 1995. Analysis of fMRI time-series revisited—Again. NeuroImage 2: 173–181. Zarahn, E., Aguirre, G. K., and d’Esposito, M. 2000. Replication and further studies of neural mechanisms of spatial mnemonic processing in humans. Cogn. Brain Res. 9: 1–17.

Lihat lebih banyak...

Comentários

Copyright © 2017 DADOSPDF Inc.