|
Subjects
Eleven right-handed adults (five men, six women), ages 22 to 35, were recruited from the local community. Subjects had no history of neurological or psychiatric illness, no history of CNS trauma, and no current medications affecting CNS function. The experimental protocol was approved by the McLean Hospital Institutional Review Board. Informed consent was obtained from each subject, and subjects were paid for their time.
Stimuli and Task
Video was back-projected onto a screen fastened to the front of the head coil, made visible to the subject by a mirror. Red and green coloured squares subtending 1.8° were displayed in the upper hemifield, centred 5° lateral and 3° superior to fixation. Stimuli were flashed at 9s-1 with a 50% duty cycle, for a stimulus duration of 56ms and an inter-stimulus interval of 56ms on each side, giving a total stimulus period of 112ms. Flashes on the two sides were 180° out of phase with each other, that is, the offset of a stimulus on one side coincided with the onset of a stimulus on the other side.
The experiment comprised ten trials, each of which consisted of a 60s period during which subjects received stimulation and performed the behavioural task, flanked by two 30s periods of visual fixation without stimulation, for a total duration of 120s. In order to avoid visual fatigue, subjects were permitted to rest for up to three minutes between trials. Subjects were instructed to maintain fixation at the centre of the display and to begin by attending covertly to one of the two stimulus locations. This starting location was counter-balanced across trials. On detecting the target (red) stimulus in the attended location, subjects had to respond by shifting attention covertly to the newly attended location, and by moving the index finger of the dominant hand in the direction of this shift. (The current position of the finger thus served as a mnemonic for the task.) The rest of the trial then proceeded in the same manner for the newly attended location. Stimuli of the target colour occurred with equal probability both in the attended location and in the unattended location.
The first trial contained four nominally attended targets. We describe these as 'nominally' attended since a missed target or a false alarm affected whether or not specific subsequent targets were attended. Of the remaining trials, six contained three nominally attended targets and three contained two nominally attended targets. Targets were uniformly distributed over the final 42s of each 60s period of stimulation, except in the case of the second trial, whose first target occurred at 8.68s. No attended target followed within 12s of a previous attended target, except in the case of one pair of attended targets in the fourth trial which were separated by 6.28s. These variations in number and timing of attended targets were intended to ensure that subjects kept attending throughout each trial. As our interest lay not in psychophysical measures of ability to direct attention but in the neurophysiological effects associated with successful direction of attention, the task was designed so that most subjects would be at or near ceiling.
Behavioural Recording
Fixation was monitored using an infrared eye tracking system (ISCAN, Inc., Burlington, Massachusetts). Infrared illumination of the eye was supplied by an array of LEDs mounted on the head coil and arranged so as to minimise the external magnetic field due to current flow. The eye image was recorded by an infrared camera mounted at the rear of the magnet bore, and transmitted to a computer outside the scanner room for analysis.
In order to avoid magnetic transients associated with electrical switching, an optical signalling device was constructed to transduce behavioural data. A high-output red LED (Radio Shack #276-086A, 5cd at 660nm peak) was mounted at the top of a wooden enclosure, powered from an external supply via a twisted-pair cable so as to minimise external magnetic field. Two strands of DB-1000 1mm plastic optical fibre were mounted in holes drilled at the bottom of this enclosure. When the subject's finger was positioned on the right side of the enclosure, it blocked the light path between the LED and the fibre ends. When the subject's finger was positioned on the left, the fibre ends were illuminated. The fibres led outside the scanner room through an RF wave guide to a fast-acting Type 7H cadmium selenide photocell (Mouser Electronics #621-CL707H) connected in parallel with a 100MΩ resistor, and the voltage across this photocell was digitised and sampled at 18s-1 through a standard PC game port.
Scanning
During each 120s trial, forty single-shot echo-planar images (TR=3s, effective TE=40ms, flip angle 90°) were collected in coronal slices perpendicular to the midsagittal line between the anterior and posterior commissures. Images were 7mm thick and spaced 1mm apart, with a 20cm square field of view in a 64x64 matrix for an in-plane resolution of 3.125mm, and were acquired on a General Electric Signa 1.5T system with a standard birdcage head coil. Scanning was synchronised to the onset of stimulation by using a trigger pulse from the scanner to cue stimulus delivery. Following the functional scan, T1-weighted (TR=500ms, TE=8ms) and high-resolution echo-planar (spin-echo, TR=3s, TE=80ms) images were acquired in 256x256 matrices, in the same planes and with the same field of view as the functional images.
Analysis
Functional images were corrected for head motion using Decoupled Automated Rotational and Translational motion correction [Maas & al. 1997], a method that uses a k-space representation of the images to separate rotational and translational (k-space phase) components. Linear trend and baseline offset were removed from each trial separately. A thresholding method was applied to classify voxels as brain or non-brain, and these automatically generated classifications were then manually examined and retouched. In order to avoid any assumptions about hæmodynamic response, and since the task and fixation periods were relatively long in comparison to hæmodynamic response latencies, the ideal response function was defined as a simple square wave, 1 for time points within each of the 60s task periods and 0 for time points within each of the flanking 30s fixation periods. The first point within each of these periods was excluded, as were all points during breaks in fixation. This task-versus-fixation waveform was applied using the permutation test [Belmonte & Yurgelun-Todd 2001a] component of the AFNI software package [Cox 1996] to generate tail probabilities for task-related activation of each brain voxel.
Probability maps generated by this task-versus-fixation comparison were used in combination with each subject's anatomical images to draw regions of interest for a left-hemifield-versus-right-hemifield attention comparison, constructed so as to be independent of the task-versus-fixation comparison. The three regions of interest in each hemisphere consisted of five voxels in ventral occipital cortex centred on middle occipitotemporal gyrus, four voxels just superior to the fundus of the intraparietal sulcus, and four voxels in superior parietal cortex. (Because the study focussed a priori only on these three regions, voxels that were activated in the task-versus-fixation comparison but were not located within any of these anatomically defined regions (e.g., voxels in striate cortex) were not considered in the attention comparison.) These regions were drawn on a slice whose anterior-posterior Talairach coordinate was approximately -70mm. Regions of interest for all subjects are illustrated in Figure 1.
|
For the attention comparison, behavioural data from each subject were used to define a second ideal waveform that described the direction of attention as a function of time. The first point in each task period was excluded from this waveform, as was the first point following each shift of attention. In addition, all points during breaks in fixation were excluded. In cases of missed targets or false alarms, points were excluded backwards in time to the previous correct response or to the beginning of the task period, and forward in time to the next correct response or to the end of the task period. The points remaining after these exclusions were assigned a value of 1 for leftward attention, or 0 for rightward attention. This behaviour-based procedure thus produced an ideal waveform defining short (~15s) attention blocks, embedded in the longer (60s) task blocks and within the optimal range of block length for an alternating block design [Skudlarski & al. 1999]. Because it considered only the task periods, during which the ideal waveform for task-versus-fixation was always 1, the attention comparison was independent of the task-versus-fixation comparison.
The data from the attention comparison were processed in two methods in parallel, one using a conventional measure of percent signal change, and the other using z-scores derived from a statistical parametric map. For the analysis of percent signal change, for each of the two attention conditions separately, BOLD signal within each region of interest was summed and then expressed as a percentage above the mean baseline signal acquired during the fixation periods. The percent increase for the attend-right condition was then subtracted from the percent increase for the attend-left condition. The resulting attend-left - attend-right differences in percent signal change were subjected to a 2x2x3 (sex by hemisphere by region-of-interest) analysis of variance. Post hoc t-tests were applied as indicated by F values from this analysis. The strategy of computing within-subjects percent signal change and using it as the independent variable in a between-subjects analysis of variance thus implemented a mixed-effects (often referred to as `random-effects') model [Holmes & Friston 1998].
For the map-based analysis, the regression coefficient between the attention waveform and the fMRI time series was transformed to a z-score at each voxel, forming an SPM{Z} [Friston & al. 1995]. z-scores from this map were then averaged within each region of interest, to form a z-score reflecting the degree of attention-related activity in the region as a whole. Since the ideal waveform was arbitrarily chosen to be positive for leftward attention and zero for rightward attention, a positive regional z-score denotes correlation with leftward attention while a negative regional z-score denotes correlation with rightward attention. These values were then subjected to the same analysis of variance and post hoc tests as specified above for the analysis of percent signal change. Here again, the two-stage analysis (within-subjects z-score followed by between-subjects analysis of variance) implemented a mixed-effects model.
