fMRI localizer technique: Efficient acquisition and functional properties of single retinotopic positions in the human visual cortex

www.elsevier.com/locate/ynimg NeuroImage 28 (2005) 453 – 463

fMRI localizer technique: Efficient acquisition and functional properties of single retinotopic positions in the human visual cortex Antje Kraft,a,b,c Mark M. Schira,a Herbert Hagendorf,c Sein Schmidt,a Manuel Olma,a and Stephan A. Brandt a,* a

Department of Neurology, Charite´, Berlin NeuroImaging Center, Schumannstr. 20/21, 10117 Berlin, Germany Department of Neurology II, Otto-von-Guericke University, Magdeburg, Germany c Cognitive Psychology, Humboldt-University, Berlin, Germany b

Received 25 February 2005; revised 24 May 2005; accepted 27 May 2005 Available online 12 July 2005

Current fMRI retinotopic mapping procedures often use checkerboard stimuli consisting of expanding rings and rotating wedges to measure the topography within human visual areas. Efficient procedures are well described in the literature. For many experimental paradigms, e.g., visuo-spatial attention paradigms, the identification of taskrelevant positions is the only mandatory prerequisite. To define these specific ‘‘regions-of-interest’’ (ROIs), spatially defined localizers are used. A precise evaluation of localizer techniques in regard to efficient scanning time, optimal BOLD (blood oxygenic level dependent) response, as well as quantification of the resulting ROIs within each visual area (size, overlap, surround effects) has not been studied to date. Here, we suggest a mapping procedure designed to quantify spatial and functional properties of single positions at close proximity in multiple human visual areas. During a passive viewing task, various stimuli (e.g., checkerboards or colored objects) subtending 1.4- of visual angle were presented at one out of four positions in a randomized block design. We measured the degree of overlap between positions at different hierarchical levels of the visual system (V1 – V4v) and quantified modulatory effects on a specific position by stimulation at neighboring (1.7- spacing) or distant positions (5.1- or 8.5- spacing). Within each visual area, ‘‘mexican-hat’’ distributions of local signal intensity changes, which describe a particular combination of facilitatory and suppressive effects, were found. Cubic fitting revealed the most localized tuning effect in V1, which gradually decreased throughout the higher visual areas. Colored objects were most efficient in localizing circumscribed retinotopic positions in both early and higher areas. D 2005 Elsevier Inc. All rights reserved. Keywords: Occipital cortex; Retinotopy; Receptive field; Negative BOLD; Mapping

* Corresponding author. Fax: +49 30 450560943. E-mail address: [email protected] (S.A. Brandt). Available online on ScienceDirect (www.sciencedirect.com). 1053-8119/$ - see front matter D 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.neuroimage.2005.05.050

Introduction A hallmark of the functional neuroanatomy of the visual system is its hierarchical organization, in which the visual field is topographically represented in multiple visual areas. In 1974, Hubel and Wiesel showed in non-human primate striate cortex (V1) that both scatter and the size of a receptive field (RF) increases with its distance from the fovea (eccentricity). This principle has been shown to also hold true for higher levels of visual processing (e.g., V2, V3/VP, V3a/V4v); although accompanied by coarser resolution with increase in hierarchical level (V2: Gattas et al., 1981; V3/VP: Burkhalter et al., 1986; V3/V4v: Gattas et al., 1988). In human subjects, functional magnetic resonance imaging (fMRI) is used to map these areas in regard to their individual finegrained topographical organization (retinotopic mapping). To do so, current mapping procedures commonly use a battery of stimuli consisting of expanding rings or rotating wedges (e.g., Engel et al., 1997; Sereno et al., 1995; DeYoe et al., 1996). The spatial and functional properties of the resulting activations within the visual areas are well described (e.g., Dougherty et al., 2003). The particular mapping procedures differ with respect to their stimulus characteristics, e.g., black – white checkerboards (e.g., Sereno et al., 1995), colored checkerboards (e.g., Warnking et al., 2002; Tootell and Hadjkhani, 2001), or video stimuli (e.g., Schira et al., 2004); the flicker frequency of stimuli, e.g., 4 Hz (e.g., Sereno et al., 1995) or 8 Hz (e.g., DeYoe et al., 1996); the block length, e.g., 16 – 20 s (e.g., Tootell et al., 1998a; Martinez et al., 2001), 32 s (e.g., Tootell and Hadjkhani, 2001; Engel et al., 1997), 40 s (e.g., Tootell et al., 1998b; DeYoe et al., 1996); and the task, e.g., passive viewing (e.g., Sereno et al., 1995; Tootell et al., 1998b; Engel et al., 1997) versus central or peripheral task (e.g., Martinez et al., 2001; Tootell et al., 1998a; DeYoe et al., 1996). To optimize the standard mapping technique, some groups have addressed the question how data acquisition can be improved in regards to efficient scanning time and optimal BOLD vascular response (e.g., Warnking et al., 2002; Slotnick and Yantis, 2003; Hagenbeek et al., 2002).

454

A. Kraft et al. / NeuroImage 28 (2005) 453 – 463

It must be pointed out that in many experimental paradigms, only the distinction of task-relevant positions is necessary, e.g., in visuo-spatial attention paradigms. To spatially identify these specific positions, circumscribed localizer stimuli were introduced (e.g., Rees et al., 2000; Somers et al., 1999; Slotnick et al., 2003; Mu¨ller et al., 2003; McMains and Somers, 2004). Unfortunately, little is known about the spatial and functional properties of the resulting activations (ROIs) as the extent and characteristics of local cortical activation (e.g., size, overlap, and surround effects) under specific targeting of a circumscribed position within the visual field. It would also be helpful to know which specific stimulus type produces the strongest local BOLD effect. Bearing in mind that in many studies the main paradigm (e.g., attention task) will be demanding both, with regard to the subject’s performance, as well as with regard to scanning time, it would be advantageous to optimize ROI mapping procedures. Here, we suggest an fMRI mapping procedure designed to functionally separate circumscribed retinotopic positions at close proximity in human test subjects. This technique should be able to quantify the area-dependent local spatial and functional characteristics (size, overlap, surround effects) as well as differentiate effects across visual field quadrants. Moreover, different stimuli will be tested in terms of efficiency criteria.

Materials and methods Subjects Eight healthy right-handed subjects (mean age 27 years, range 25 – 30) with normal color vision and sufficient visual acuity participated in the study, which was conducted in conformity with

the Declaration of Helsinki. All subjects signed a consent form and were rewarded for their participation. Experimental paradigm Mapping of visual areas (meridian-mapping) To define the borders separating early visual areas, we used a standard meridian-mapping experiment (DeYoe et al., 1996; Slotnick and Yantis, 2003; Schira et al., 2004). Checkerboard stimuli were sequentially presented at the horizontal and vertical meridians for 20 s. All participants completed two 9-min runs (12 repetitions per condition), each with five 20-s periods of fixation (see Fig. 1A). By simultaneously stimulating either both horizontal (right/left) or both vertical meridians (upper/lower), data acquisition time was reduced (Yantis and Slotnick, 2003). Mapping of single positions (localizer-mapping) (see Figs. 1A and B) During a passive viewing task, various stimuli encompassing 1.4- of visual angle were presented at one out of four positions in the visual field in a block design. The specific positions were carefully selected in order (i) to test whether we can identify separate regions within each visual area (V1 – V4v), (ii) to measure the degree of overlap between positions at different hierarchical levels of the visual system and, (iii) to measure modulatory effects on a specific position when neighboring (1.7spacing) or distant (5.1- or 8.5- spacing) positions were stimulated. A central fixation cross and four squares indicating the four possible stimulus positions (positions 1 – 4) were visible during the entire experiment and subjects were instructed to maintain central fixation. Stimuli were presented in randomized order at the four locations (positions 1 – 4, block length 20 s) with 8 repetitions for each position. Three ‘‘no-stimulus’’ periods of

Fig. 1. Experimental design. (A) Schematic illustration of meridian- and localizer-mapping protocols. Row 1: One run of the meridian-mapping (F = fixation, H = checkerboards oriented along the horizontal meridian, V = checkerboards oriented along the vertical meridian). Row 2: One run of the localizer-mapping (F = no stimulus condition, numbers 1 – 4 indicate stimulus presentation at positions 1 – 4, respectively). (B) Illustration of stimuli. Positions 1 and 2 were 1.2- above or below the horizontal meridian and positions 3 and 4 were 1.2- right or left from the vertical meridian (dashed lines). Distance between target positions and fixation as well as spacing between stimulus positions (in degrees). Subjects passively viewed the stimuli presented randomly at each position sequentially for 8  20 s while maintaining central fixation. Different stimulus types (black – white/blue – yellow checkerboards, white flash stimuli, colored objects) were used in distinct experimental runs.

A. Kraft et al. / NeuroImage 28 (2005) 453 – 463

455

27 s at the beginning, middle, and end of each experimental run defined a baseline condition. Each subject performed eight 12-min experimental runs (2 runs per stimulus condition described below).

white contrast for accurate segmentation and reconstruction of individual surface structures.

Stimuli (see Fig. 1B) Four different types of stimuli were used in different experimental runs. Black – white and blue – yellow checkerboard stimuli that were presented at a rate of 4 Hz. Different types of colored object stimuli (circle, rotated ‘‘T’’ and triangle; black, white, green) and white flash stimuli were sequentially presented at a rate of 7 Hz. Each participant performed two runs per stimulus condition. The order of runs was balanced across subjects.

fMRI data were analyzed using BrainVoyager 4.9 software package (BrainInnovation, Maastricht, Netherlands). Functional data preprocessing contained motion correction, high pass filtering of frequencies above 3 cycles per time course and temporal smoothing (5 s FWHM). Only the meridian-mappings were spatially smoothed using Gaussian 3D kernel with 3 mm FWHM. All anatomic and functional data were transformed into stereotactic space (Talairach and Tournoux, 1988).

Fixation control It was required that subjects maintain central fixation during the whole mapping procedure. During the mapping of single positions, fixation behavior was not controlled. But all subjects participated in two further experiments where eye movements were recorded inside the scanner with an infrared video eye tracker system (MEye-Track, SMI, Teltow, Germany). In these experiments, all subjects (N = 8) were able to maintain fixation within 2- of the center in more than 95% of the trials. fMRI procedure Visual stimuli were presented by ERTS (BeriSoft Cooperation, Frankfurt, Germany) and displayed by an NEC LCD-projector and a custom-made lens on a small back-projection screen mounted in front of the head coil. Subjects viewed the screen via a mirror mounted at a distance of 24 cm from the head. MRI data were acquired with a 1.5-T Magnetom Vision (Siemens, Erlangen, Germany). Subjects’ heads were stabilized with a vacuum pillow in a standard head coil. Occipital functional images were collected using single-shot EPI sequences (TR = 3.3 s, TE = 60 ms, FA = 90-, 18 slices, 128  128 matrix, FOV 256, voxel size: 2  2  3.3 mm). Each fMRI session contained three preliminary saturation scans to allow for T1 equilibration effects. Meridian-mapping During each experimental run (2 per subject), 174 volumes were collected (2 conditions: checkerboards at horizontal/vertical meridians, 6 volumes per condition, 12 repetitions; 6 volumes fixation, 5 repetitions; see Fig. 1A). Localizer-mapping During each experimental run (2 per stimulus condition), 216 volumes were collected (4 positions: P1 – 4, 6 volumes per position, 8 repetitions; 8 volumes fixation, 3 repetitions; see Fig. 1A). Anatomical images Structural three-dimensional data sets were acquired, using a T1-weighted sagittal MP-Rage-sequence (TR/TE = 10/4 ms, FA = 12-, TI = 100 ms, voxel size = 1 mm3). High-quality threedimensional data sets for each subject were recorded using a T1weighted sagittal FLASH-sequence (TR/TE = 38/5 ms, FA = 30-, voxel size = 1 mm3) with two acquisitions for excellent gray –

Data analysis

Single-subject level We segmented and reconstructed the surface of the white matter and produced flat maps from the high-resolution structural MRI images of each subject. Based on the results of the meridianmapping experiment, the borders of the visual areas (V1d, V1v, V2d, V2v, V3, VP, V3a, V4v) were defined. For definition of the ROIs (localizer-mapping), multipleregression models were fitted to compute statistical maps for the effect stimulus position. The predictors (one for each position) were generated by convolving a square-wave function representing the time course of the experimental conditions with a gamma function (delta = 2.5; tau = 1.25) modeling the hemodynamic response (Boynton et al., 1996). ROIs for the positions 1 – 4 (each position relative to the ‘‘no-stimulus’’ baseline) were defined using a consistent significance level ( P < 0.0001, uncorrected). Significant meridians and localizer activity were projected onto the flattened cortical surface and color-coded; dark green and gray corresponding to the horizontal and vertical meridians, respectively. Blue, yellow, red, and green correspond to positions 1 through 4, respectively (Figs. 2, 3A, and 4). Each of the early visual areas (V1 to V4v) contained position-based ROIs for positions 1 – 4, resulting in 16 ROIs (4 positions  V1 – V4v) for each subject. The size of each ROI (number of voxels, 2  3  3.3 mm3) and the overlap between ROIs (percent overlap) within each visual area were acquired (Figs. 3B and C). Group level To address the level of activation within each ROI (positions 1 – 4), t test versus baseline were performed in order to test whether significant increases or decreases occurred when stimuli were presented at the exact (0-), neighboring (1.7-), or distant (5.1 or 8.5-) positions. It is known that measurable positive hemodynamic activity can last up to 30 s after stimulus onset for a stimulus duration of 24 s (e.g., Boynton et al., 1996). To ensure that the baseline is not confounded by preceding events, we only used the mean activity from second 13 to second 27 of the ‘‘no-stimulus’’ condition (33 to 47 s after the last stimulus onset) for each subject. This allowed us to compare the increases and the decreases in activity in relation to this specific baseline interval. To test differences in overlap between ROIs within the visual areas against spacing, a two-way analysis of variance (ANOVA) with factors ‘‘visual area’’ (V1, V2, V3/VP, V3a/V4v) and ‘‘distance’’ (near 1.7-, distant 5.1- or 8.5-) was calculated. In order to measure ROI size differences across visual areas, a two-

456

A. Kraft et al. / NeuroImage 28 (2005) 453 – 463

Fig. 2. ROIs of stimulus positions. Reconstructed flat patch of left and right occipital cortices of five subjects (s1 – s5); results of the ROI mapping are depicted in blue (position 1), yellow (position 2), red (position 3), and green (position 4). Borders of previously mapped retinotopic visual areas (V1 – V4v) are indicated by white lines, representations of meridians are marked in light gray (horizontal) and dark green (vertical).

way ANOVA with factors ‘‘position’’ (positions 1 – 4) and visual area (V1 – V4v) was computed. Additionally, a two-way (ANOVA) with factors ‘‘stimulus type’’ (checkerboards black – white, checkerboards blue – yellow, white flash, colored objects) and ‘‘visual area’’ (V1, V2, V3/VP, V3a/V4v) was performed in

order to test whether different stimulus types revealed significant differences in percent signal change in distinct visual areas. For all tests on the group level, multiple comparisons within each ROI or within each area, respectively, were Bonferroni-corrected ( P < 0.05).

A. Kraft et al. / NeuroImage 28 (2005) 453 – 463

457

the second derivative of the curve about the local maxima, i.e., the ‘‘peak’’ BOLD signal. This analysis derived the tangent’s gradient rate of change within V1, V2, V3/VP to V3a/V4v, i.e., the ‘‘steepness’’ of the curves.

Results and discussion Our analysis revealed differential patterns of activation regarding size and overlap of ROIs as well as differential patterns of facilitatory and suppressive activations within the ROIs for each visual area V1 – V4v. Also, the analysis revealed differential activations for distinct stimulus types within the visual areas. ROIs in V1 – V4v: overlap and size

Fig. 3. Overlap and size of ROIs of the stimulus positions. (A) Reconstructed flat patch of a left and right occipital cortex from one subject (s1); the color convention is the same as in Fig. 2. (B) Percent of overlapping voxels between neighboring ROIs (1.7-: P1 – P2, P2 – P3, P3 – P4) and distant ROIs (5.1-: P1 – P3, P2 – P4; 8.5-: P1 – P4) for each visual area (V1 white, V2 light gray, V3/VP dark gray, V3a/V4v black). P2 – P3 represents overlap within one quadrant; P1 – P2 and P1 – P3 reflect overlap between the upper and lower quadrant, respectively; P3 – P4, P2 – P4, and P1 – P4 show overlap between the left and right hemispheres. (C) Number of voxels (2  2  3.3 mm3, P < 0.0001, uncorrected) in each ROI (P1 – P4) within each visual area (V1 – V4v) averaged across subjects (N = 8).

Curve fitting (see Fig. 4) We performed a 3- polynomial fit on the data obtained from four areas V1, V2, V3/VP, and V3a/V4v for each ROI (positions 1 – 4). To characterize the properties of these curves, we looked into

Despite differences between single subjects, for each subject, circumscribed visual field representations (ROIs) for each position (P1 – P4) could be reliably mapped within the visual areas V1 – V4v. Fig. 2 shows examples of five different subjects — that are representative for all eight subjects. These retinotopic representations were well separated for each position in V1, V2, and V3/VP. In higher visual areas (V3a, V4v) this separation degraded. Fig. 3A depicts the overlap between ROIs of one representative subject (s1). Fig. 3B shows the overlap between ROIs for the whole group (N = 8). Despite a close proximity of the stimulus positions to the horizontal or vertical meridian, the overlap of representations— across upper and lower quadrants (P1 – P2, P1 – P3; Fig. 3B) or between hemispheres (P3 – P4, P2 – P4, P1 – P4; Fig. 3B)—was minor (<5%) in areas V1 to V3/VP. In higher visual areas, the overlap was stronger (up to 10 – 15%), which is in line with the increase in RF size (e.g., Smith et al., 2001). Overlap was larger in higher areas for near positions (1.7-) as compared to distant (5.1and 8.5-) positions (Fig. 3B). The results were statistically confirmed by a two-way ANOVA with factors ‘‘visual area’’ and ‘‘distance’’, revealing a significant main effect for the factor ‘‘visual area’’ [ F(3,21) = 4.36, P = 0.016] and a marginal, not significant, interaction between both factors [ F(3,21) = 3.04, P = 0.052]. Also the ROI sizes are comparable in areas V1 – VP/V3, while the size of ROIs increases significantly throughout the higher visual areas V3a/V4v (Fig. 3C). Performing a two-way ANOVA with factor ‘‘position’’ and ‘‘visual area’’, a significant main effect for the factor ‘‘visual area’’ [ F(3,15) = 3.67, P = 0.037] was obtained.1 To discuss these results, it is important to note that the V1 activation induced by stimulation of position 2 (yellow, Fig. 2) is either represented on the upper, or the lower or both calcarine lips of V1. There are two explanations contributing to this variability. First, the anatomical cut through the calcarine fissure is artificial and does not mark the exact border between V1 ventral and V1 dorsal. Thus, incorrect separations of the upper and lower representation of visual field quadrants in V1 will occur in some instances. Second, some variance in size and overlap of retinotopic representations may be introduced by inaccurate fixation behavior.

1 Note that P1/P4 and P2/P3 were at different eccentricities. Due to the cortical magnification factor, stimuli at 3.4- (P2/P3) should activate more cortex compared to the equally large stimuli presented at 6.1- (P1/P4). This becomes apparent in higher areas V3a/V4.

458

A. Kraft et al. / NeuroImage 28 (2005) 453 – 463

Small saccadic eye movements could lead to some smearing and displacement of retinotopic representations. Size differences in the range and location of visual field representations between subjects have been reported by Dougherty et al. (2003) using fMRI and Amunts et al. (2000) creating cytoarchitectonic maps of Brodmann’s areas 17 and 18. The exact definition of the size of visual field representations as described in physiological studies is not feasible with fMRI (e.g., Smith et al., 2001, Kastner et al., 2001). The size of activation always depends on the statistical threshold level. However, with consistent intrasubject significance levels, one can make suggestions about the relative differences between areas within the single subject. The size of visual field representations within the visual areas is a function of receptive field size (e.g., Smith et al., 2001). Thus, one would expect that an ROI of a defined visual field position is larger in higher visual areas compared to the early visual areas. On the other hand, it is also known that early visual areas span a larger region of visual cortex than the higher visual areas. For instance, Dougherty et al. (2003) measured the mean size of human V1 with 1470 mm2 compared to 819 mm2 for V3, representing 2 – 12- of the visual eccentricities. Thus, smaller regions in the visual field are represented larger in early visual areas as compared to higher visual areas. Taken together, both the receptive field size and the size of the cortical visual area need to be considered to estimate the size of an activated region (ROI) in distinct visual areas. Thus, for a specific stimulus representation in the cortex, one might not necessarily expect an increase in ROI size along the visual processing stream. Further, one would expect less overlap between visual representations in early visual areas as compared to the higher visual areas. Our results coincide well with this notion.2 We found overlap between ipsi- and contralateral visual field representations and furthermore overlap between the upper and lower fields confirming previous fMRI studies (Slotnick and Yantis, 2003; Tootell et al., 1998c). This was the case especially within the higher areas V3a and V4v. In comparison of unifield and bifield stimulation, Slotnick and Yantis (2003) proposed bifield stimulation to map the visual meridians more efficiently. They also pointed out that ipsilateral and contralateral field overlap cannot be measured using bifield stimulation. Instead, they proposed to use bifield stimulation when only the mapping of early visual areas (V1 – V3) is required, while unifield stimulation would be appropriate for mapping of higher areas. For the localizer technique, one should not simultaneously stimulate at multiple positions—especially for locations near the meridians—as overlapping activations may occur both in early and higher visual areas. In primates, it was suggested that ipsilateral input in early visual areas is negligible and primarily occurs around the vertical meridian (e.g., Tootell et al., 1998c). It was proposed that these specific regions are connected via transcallosal fibers (e.g., Desimone et al., 1993). In higher areas, receptive fields are larger and extend into the ipsilateral visual field and consequently, ipsilateral input is larger. Indeed, in the current study, there was

2

Dougherty et al. (2003) found that V1 in the left hemisphere tended to be larger than V1 of the right hemisphere. For areas V2 and V3, no differences between the hemispheres were found. In the current study (Fig. 2B), the ROI of V1 in the left hemisphere (P4, 6.1-) also tended to be larger compared to V1 in the right hemisphere (P1, 6.1-). Size differences of ROIs in V2 and V3/VP between the left and right hemispheres were minor.

large interhemispheric activation in higher areas V3a/V4v, but there was also some overlap in the early visual areas. As an example, our results showed small overlapping representations in the region near the vertical meridian (P3 – P4, Fig. 3C). This could be due to a bilateral retinotopic representation of this visual region as measured by Dow et al. (1981) in non-human primates and Tootell et al. (1998c) in a human fMRI study. Activation patterns within ROIs in V1 – V4v We measured both the hemodynamic signal intensity changes within each ROI when stimuli were presented at the corresponding position, as well as changes from baseline activity (for baseline computation, see Materials and methods) within each ROI when surrounding positions (spacing 1.7-, 5.1-, or 8.5-) were stimulated. We found a significant increase of the signal intensity in the targeted cortical region, and also a significant decrease in the same region when stimuli were presented to surrounding positions. For example, the BOLD signal extracted from the ROIs representing position 1 (Fig. 4, first column) was enhanced when position 1 was stimulated and was reduced in areas V1 through VP when position 3 (5.1- spacing) or position 4 (8.5- spacing) was stimulated. In contradistinction, in V4v, the BOLD signal was enhanced also when position 2 was stimulated (1.7- spacing). The modulations were related to the visual areas’ eccentricity-based approximate RF sizes (Fig. 4, gray rows) as estimated in monkeys and humans (Smith et al., 2001). Thus, stimulation within an RF showed a signal increase. Stimulation of—as well as beyond—the border of an RF led to either no modulation or to local signal suppression. These ‘‘mexican-hat’’ distributions were steepest in V1/V2 (Figs. 4 and 5). Cubic fitting performed on the averaged data revealed the localized (strong) tuning effect in V1, and gradual decrease over the higher areas (Fig. 4). To quantify the properties of these curves, we looked into the second derivative of the curve about the local maxima (see Materials and methods). For example, for position 2, the values of the second derivative about these maxima are: (V1) f1_dd(1.75748) = 2.85456 (V2) f2_dd(1.86066) = 1.85428 (V3) f3_dd(1.86426) = 1.82028 (V3a) f4_dd(1.93088) = 0.793342

From this analysis, we could derive that the rate of change in tangent gradient, i.e., the ‘‘steepness’’ of the curve, about the maxima is highest at V1 (=2.9), and gradually decreases [linear function F(1,2) = 20.1; R 2 = 0.91; P = 0.046] with higher areas (V3a = 0.8). Note that the scale of the abscissa in the plots of Fig. 4 is not linear. A linear scale is used in Fig. 5. Although the spacing of stimuli is not fine enough to characterize the spatial profile as accurately as in single cell studies, again it is obvious that the localization of the peak response and surrounding suppressive effects (i.e., ‘‘mexican-hats’’) are more salient in V1 compared to V3a/V4v. This can be observed for all retinotopic representations (P1 – P4). The origin of a negative BOLD signal in the visual system has been discussed in several fMRI studies (e.g., Williams et al., 2003; Chen et al., 2005). In the current study, the decrease of BOLD vascular response in the unstimulated regions follows the onset of the stimulus and has a similar duration to that of the positive

A. Kraft et al. / NeuroImage 28 (2005) 453 – 463

459

Fig. 4. ROIs BOLD signal modulation. First row: Reconstructed flat patch of a right (positions 1 – 3) and a left (position 4) occipital cortex (one subject); see also Fig. 2 legend. Columns and rows: Averaged (N = 8) hemodynamic response in different ROIs (positions 1 – 4) in V1 – V4v (rows 1 – 4). Cubic fitting was done for each position within each visual area. Column 1: BOLD signal modulation in position 1 ROIs (V1 – V4v: rows 1 – 4) when stimuli were presented at positions 1 – 4 (at 0-, 1.7-, 5.1-, and 8.5- spacing, see x coordinate). Columns 2 – 4: BOLD signal modulation in position 2, 3, or 4 ROIs, respectively, when stimuli were presented at positions 1 – 4 (spacing, see x coordinate). *P  0.05, **P  0.001 significant difference to baseline. (¨) Estimated RF sizes derived from macaque monkeys (Smith et al., 2001) are given as a function of eccentricity (3.4-, 6.1-) and visual area (V1 – V4v).

BOLD vascular response in the stimulated region. For this reason, the results cannot be equated with the small ‘‘initial-dip’’ which is sometimes observed during the first seconds of stimulus presentation (e.g., Menon, 2001) or with the post-stimulation undershoot

after stimulus offset (e.g., Kwong et al., 1992). One can ascribe the negative BOLD as a reduction of neuronal activity (Chen et al., 2005), but we also have to consider a possible reduction of cerebral blood flow (CBF) due to spatial redistribution of CBF (Tootell et

460

A. Kraft et al. / NeuroImage 28 (2005) 453 – 463

Fig. 5. Summary of Fig. 4. Graphs show the averaged response of 8 subjects (black circles). Hemodynamic response in different ROIs (positions 1 – 4) in V1 (above) and V4v/V3a (below) when the position (0-) or a surrounding position was stimulated (1.7- to 8.5- spacing, linear spacing of abscissa). Data points for each position were bonded (in blue position 1, in yellow position 2, in red position 3, in green position 4).

al., 1998c; Smith et al., 2004; Shmuel et al., 2002). Such a ‘‘blood stealing’’ as a main factor for a negative BOLD signal appears unlikely, as some ROIs were separated across hemispheres in our experiment. Reduced local blood pressure arising from nearby capillary dilation cannot explain negative BOLD in this setting (Tootell et al., 1998c; Smith et al., 2004). Even within one hemisphere, a negative BOLD signal is probably mainly based on suppression of neuronal activity below spontaneous firing, as it is associated with a reduced rate of oxygen consumption (CMRO2) during visual stimulation of nearby positions (Shmuel et al., 2002). Suppression outside the RF, induced by stimulation of a neighboring position, can well be explained by non-classical RF effects, which are mediated either by long range lateral (e.g., Lamme, 1995; Desimone et al., 1993; Stettler et al., 2002; Chen et al., 2005) or divergent feedback connections (e.g., Angelucci and Bullier, 2003; Kastner et al., 2001). In contrast, in higher visual areas, neighboring positions are covered by the same RF. Here, it can be expected that a neighboring position is being co-activated (e.g., Fig. 4, lower row) and that the overall signal intensity is weaker as a result of competing effects within an RF (Reynolds et al., 1999; Moran and Desimone, 1985). On the one hand, the spatial resolution of fMRI (‘‘macroscopic’’ system level) is not comparable with the resolution of neurophysiological techniques acquiring data on a ‘‘microscopic’’ level of single neurons. Consequently, one cannot directly compare the results of fMRI studies with neurophysiological studies, which describe spatial, orientation, and frequency tuning characteristics for single neurons. Moreover,

extracellular single unit ‘‘inhibition’’ could also possibly lead to an increase in BOLD signal due to higher activity in presynaptic inputs (Logothetis, 2001; Arthurs and Boniface, 2002; Attwell and Iadecola, 2002). On the other hand, despite the coarser resolution and difficulties in BOLD signal interpretation, our results coincide well with results from the invasive microelectrode recordings in non-human primates (RF properties: e.g., Hubel and Wiesel, 1974; Gattas et al., 1981, 1988; Burkhalter et al., 1986; Suppression: e.g., Angelucci et al., 2002; Walker et al., 1999; Desimone et al., 1993). However, two further explanations for suppressive effects at surrounding positions must be considered. First, suppression could be due to attentional modulations. Several fMRI studies investigating specific retinotopic attention effects (Slotnick et al., 2003; Kastner et al., 2001; Somers et al., 1999), not only showed increased activations at the representation of the ‘‘attended’’ stimuli, but also profound deactivations in the surrounding cortex. The results of the present study could be explained in relation to these findings. As no central task controlled attention, stimulus presentation most likely captured the subject’s attention automatically. Second, the negative BOLD could be an epiphenomenon of the analysis procedure. It is known that the positive hemodynamic activity can last up to 12 s after stimulus onset for a stimulus duration of up to 3 s (Boynton et al., 1996). However, recent studies showed that physiological effects can also last considerably longer (e.g., Donaldson and Buckner, 2000; Fransson et al., 1999; Miezin et al., 2000). Boynton et al. (1996) showed that longer event durations (e.g., 24 s) can lead to prolonged positive hemodynamic activity (e.g., 30 s after stimulus onset). In contrast, others reported a ‘‘post-stimulus undershoot’’ (e.g., Buxton et al., 1998; Fransson et al., 1999), i.e., negative hemodynamic activity. Calculating a baseline in periods of ongoing positive activity or in periods including an ‘‘undershoot’’ could over- or underestimate the baseline level, respectively. In the present study, the activity of the ‘‘no-stimulus condition’’ from 0 to 12 s (20 to 32 s after the last stimulus onset) was omitted from analysis. The mean activity from 13 to 27 s (33 to 47 s after the last stimulus onset) was defined as baseline. However, due to a stimulus duration of 20 s, residual stimulus activity may possibly still confound the baseline period. In other words, at a position not activated by the stimulus, the subtraction of a mean baseline confounded by positive residual activity would result in overly negative approximation. If the residual activity were negative (i.e., undershoot), the result would be overly positive. To summarize, neither ‘‘initial-dip’’ nor ‘‘blood stealing’’ phenomena are likely to explain the suppressive effects found in our study. Possibly, suppression could be ascribed to a confounded baseline. Physiologically, lateral-mediated suppression as well as top – down-driven attentional feedback could explain our results. The study cannot differentiate between these mechanisms. Stimulus efficiency The results of differential mapping stimuli (checkerboards, white flash stimuli, colored objects) were compared for areas V1 – V4v (average of two runs for each subject). Fig. 6A shows the differential results for position 4 of one representative subject. With a consistent significance level ( P < 0.0001, uncorrected), the activation of the 8 experimental runs is shown in green, the activation of the two runs (either checkerboards, white flashes, or object stimuli) is marked in red – yellow. Checkerboards revealed strong activations in early visual areas, but less activation in higher areas. White flash stimuli

A. Kraft et al. / NeuroImage 28 (2005) 453 – 463

461

visual areas, checkerboards revealed less activations as compared to the complex object stimuli. The reliability of these differences was confirmed by a two-way ANOVA. Significant main effects for ‘‘visual area’’, F(3,21) = 6.21, P < 0.01, and ‘‘stimulus type’’, F(3,21) = 31.11, P = 0.003, were obtained. This confirms that percent signal changes differ across the visual areas, with less modulations in higher visual areas (V3a/V4v) and differences between distinct stimulus types. A significant interaction between both factors, F(9,63) = 2.22, P = 0.032, indicates that the difference between stimulus types varies across areas. The strongest differences between the complex stimuli and all other stimulus type conditions were found in higher areas V3a and V4v. Previous studies using different kinds of stimulus types led to the suggestion that the choice of stimulus is not important for creating retinotopic maps. This is because many stimuli produced satisfactory maps (see Introduction). However, others (e.g., Slotnick and Yantis, 2003, Warnking et al., 2002; Hagenbeek et al., 2002) described criteria to design mapping procedures relative to scanning time reduction efficiency. In regard to such efficiency criteria, it could be useful to choose complex stimuli with a high flicker frequency.3 In our study, two short experimental runs (with 16 repetitions for each condition, a total of 24 min per subject) were sufficient to define ROIs in early and higher visual areas.

Conclusion The goal of this fMRI study was to analyze the localizermapping procedure with respect to spatial precision and stimulus efficiency. The described approach was well suited for efficient definition of specific regions-of-interest (ROIs) in early visual areas (e.g., for visuo-spatial attention paradigms). In both early and higher visual areas, colored stimuli were most efficient. Moreover, functional and spatial properties of the individual ROIs were quantified: Size and overlap as well as particular patterns of facilitatory and suppressive modulations (‘‘mexican-hats’’) were described within the areas V1 – V4v. The overlap between either ipsi- and contralateral or upper and lower visual fields could be mapped precisely. All of the abovementioned results corresponded well with the present spatial and functional understanding of a hierarchical visual system.

Fig. 6. Stimulus efficiency. (A) Reconstructed flat patch of a left occipital cortex from one subject (see Fig. 3). Activation (consistent significance level; P < 0.0001, uncorrected) due to all 8 experimental runs is marked in green; activation of the individually comparable two runs (first line checkerboards, second line white flash or object stimuli) is marked in red – yellow. (B) Percent signal changes (average of 8 subjects) in different areas (V1 – V4v) when different stimulus types were used.

revealed the smallest increase in BOLD signal. Complex stimuli (colored objects) showed the strongest increase in BOLD signal, both in early areas (V1/V2) and in higher areas. The same pattern can be shown in the averaged percent signal changes for the whole group (N = 8) (Fig. 6B). White flash stimuli showed the smallest increase in percent signal change within all visual areas. Checkerboard and complex objects revealed comparably strong modulations in early visual areas. But in higher

Acknowledgments Antje Kraft was supported by International Leibniz Fellowships in Cognitive and Clinical Neurosciences (Otto-von-Guericke University of Magdeburg and Humboldt-University of Berlin). Mark M. Schira and Sein Schmidt were supported by the NeuroImaging Center Berlin sponsored by the BMBF.

3 Previous studies revealed that BOLD response increases with higher frequency, with maximum response around 8 – 12 Hz (Hagenbeek et al., 2002). Thus, stimulus efficiency is maybe a function of stimulus complexity (flash, checkerboards, objects) but also a function of frequency (4 Hz vs. 8 Hz). In the current study, checkerboards were presented at a rate of 4 Hz and flash and complex stimuli were presented at a rate of 7 Hz. An increase of frequency cannot be the only factor as flash stimuli show the smallest increase in BOLD signal. The exact part of each factor (complexity vs. frequency) cannot be disentangled in the current study.

462

A. Kraft et al. / NeuroImage 28 (2005) 453 – 463

References Amunts, K., Malikovic, A., Mohlberg, H., Schormann, T., Zilles, K., 2000. Brodmann’s areas 17 and 18 brought into stereotaxic space—Where and how variable? NeuroImage 11 (1), 66 – 84. Angelucci, A., Bullier, J.J., 2003. Reaching beyond the classical receptive fieldof V1 neurons: horizontal or feedback axons? J. Physiol. (Paris) 97, 141 – 154. Angelucci, A., Levitt, J.B., Walton, E.J.S., Hupe´, J.M., Bullier, J., Lund, J.S., 2002. Circuits for local and global signal integration in primary visual cortex. J. Neurosci. 22, 8633 – 8646. Arthurs, O.J., Boniface, S., 2002. How well do we understand the neural origins of the fMRI BOLD signal? Trends Neurosci. 25 (1), 27 – 31. Attwell, D., Iadecola, C., 2002. The neural basis of functional brain imaging signals. Trends Neurosci. 25 (12), 621 – 625. Boynton, G.M., Engel, S.A., Glover, G.H., Heeger, D.J., 1996. Linear system analysis of functional magnetic resonance imaging in human V1. J. Neurosci. 16, 4207 – 4221. Burkhalter, A., Felleman, D.J., Newsome, W.T., Van Essen, D.C., 1986. Anatomical and physiological asymmetries related to visual areas V3 and VP in macaque extrastriate cortex. Vision Res. 26, 63 – 80. Buxton, R.B., Wong, E.C., Frank, L.R., 1998. Dynamics of blood flow and oxygenation changes during brain activation: the ballon model. Magn. Reson. Med. 39, 855 – 864. Chen, C.C., Tyler, C.W., Liu, C.L., Wang, Y.H., 2005. Lateral modulation of BOLD activation in unstimulated regions of the human visual cortex. NeuroImage 24, 802 – 809. Desimone, R., Moran, J., Schein, S.J., Mishkin, M., 1993. A role for the corpus callosum in visual area V4 of the macaque. Vis. Neurosci. 10 (1), 159 – 171. DeYoe, E.A., Carman, G.J., Bandettini, P., Glickman, S., Wieser, J., Cox, R., Miller, D., Neitz, J., 1996. Mapping striate and extrastriate visual areas in human cerebral cortex. Proc. Natl. Acad. Sci. U. S. A. 93, 2382 – 2386. Donaldson, D.I., Buckner, R.L., 2000. Effective paradigm design. In: Matthews, P.M., Jezzard, P., Evans, A.C. (Eds.), Functional Magnetic Resonance Imaging of the Brain: Methods for Neuroscience. Oxford Univ. Press, Oxford. Dougherty, R.F., Koch, V.M., Brewer, A.A., Fischer, B., Modersitzki, J., Wandell, B.A., 2003. Visual field representations and locations of visual areas V1/2/3 in human visual cortex. J. Vis. 3, 586 – 598. Dow, B.M., Synder, A.Z., Vautin, R.G., Bauer, R., 1981. Magnification factor and receptive field size in foveal striate cortex of the monkey. Exp. Brain Res. 44, 213 – 228. Engel, S.A., Glover, G.H., Wandell, B.A., 1997. Retinotopic organization in human visual cortex and the spatial precision of functional MRI. Cereb. Cortex 7, 181 – 192. Fransson, P., Kruger, G., Merboldt, K.D., Frahm, J., 1999. Temporal and spatial MRI responses to subsecond visual activation. Magn. Reson. Imaging 17, 1 – 7. Gattas, R., Gross, C.G., Sandell, J.H., 1981. Visual topography of V2 in the macaque. J. Comp. Neurol. 201, 519 – 539. Gattas, R., Sousa, A., Gross, C.G., 1988. Visuotopic organization and extent of V3 and V4 of macaque. J. Neurosci. 8, 1831 – 1845. Hagenbeek, R.E., Rombouts, S.A.R.B., van Dijk, B.W., Barkhof, F., 2002. Determination of individual stimulus – response curves in the visual cortex. Hum. Brain Mapp. 17, 244 – 250. Hubel, D.H., Wiesel, T.N., 1974. Uniformity of monkey striate cortex: a parallel relationship between field size, scatter, and magnification. J. Comp. Neurol. 158, 295 – 306. Kastner, S., De Weerd, P., Pinsk, M.A., Elizondo, M.I., Desimone, R., Ungerleider, L.G., 2001. Modulation of sensory suppression: implications for receptive field sizes in the human visual cortex. J. Neurophysiol. 86, 1398 – 1411. Kwong, K.K., Belliveau, J.W., Chesler, D.A., Goldberg, I.E., Weisskoff, R.M., Poncelet, B.P., Kennedy, D.N., Hoppel, B.E., Cohen, M.S., Turner, R., Cheng, H.-M., Brady, T.J., Rosen, B.R., 1992. Dynamic

magnetic resonance imaging of human brain activity during sensory stimulation. Proc. Natl. Acad. Sci. U. S. A. 89, 5675 – 5679. Lamme, V.A.F., 1995. The neurophysiology of figure-ground segregation in primary visual cortex. J. Neurosci. 15, 1605 – 1615. Logothetis, N.K., 2001. Neurophysiological investigation of the basis of the fMRI signal. Nature 412, 150 – 157. Martinez, A., DiRusso, F., Anllo-Vento, L., Sereno, M.I., Buxton, R.B., Hillyard, S.A., 2001. Putting spatial attention on the map: timing and localization of stimulus selection processes in striate and extrastriate visual areas. Vision Res. 41, 1437 – 1457. McMains, S.A., Somers, D.C., 2004. Multiple spotlights of attentional selection in human visual cortex. Neuron 42 (4), 677 – 686. Menon, R.S., 2001. Imaging function in the working brain with fMRI. Curr. Opin. Neurobiol. 11, 630 – 636. Miezin, F.M., Maccotta, L., Ollinger, J.M., Petersen, S.E., Buckner, R.L., 2000. Characterizing the hemodynamic response: effects of presentation rate, sampling procedure, and the possibility of ordering brain activity based on relative timing. NeuroImage 11 (6), 735 – 759. Moran, J., Desimone, R., 1985. Selective attention gates visual processing in the extrastriate cortex. Science 229, 782 – 784. Mu¨ller, N.G., Bartelt, O.A., Donner, T.H., Villringer, A., Brandt, S.A., 2003. A physiological correlate of the Fzoom lens_ of visual attention. J. Neurosci. 23 (9), 3561 – 3565. Rees, D., Backus, B.T., Heeger, D.J., 2000. Activity in primary visual cortex predicts performance in a visual detection task. Nat. Neurosci. 3 (9), 940 – 945. Reynolds, J.H., Chelazzi, L., Desimone, R., 1999. Competitive mechanisms subserve attention in macaque areas V2 and V4. J. Neurosci. 19, 1736 – 1753. Schira, M.M., Fahle, M., Donner, T.H., Kraft, A., Brandt, S.A., 2004. Differential contribution of early visual areas to the perceptual process of contour processing. J. Neurophysiol. 91 (4), 1716 – 1721. Sereno, M.I., Dale, A.M., Reppas, J.B., Kwong, K.K., Belliveau, J.W., Brady, T.J., Rosen, B.R., Tootell, R.B.H., 1995. Borders of multiple visual areas in human revealed by functional magnetic resonance imaging. Science 268, 889 – 893. Shmuel, A., Yacoub, E., Pfeuffer, J., Van de Moortele, P.F., Adriany, G., Hu, X., Ugurbil, K., 2002. Sustained negative BOLD, blood flow and oxygen consumption response and its coupling to the positive response in the human brain. Neuron 36, 1195 – 1210. Slotnick, S.D., Yantis, S., 2003. Efficient acquisition of human retinotopic maps. Hum. Brain Mapp. 18, 22 – 29. Slotnick, S.D., Schwarzbach, J., Yantis, S., 2003. Attentional inhibition in human striate and extrastriate cortex. NeuroImage 19 (4), 1602 – 1611. Smith, A.T., Singh, K.D., Williams, A.L., Greenlee, M.W., 2001. Estimating receptive field size from fMRI data in human striate and extrastriate visual cortex. Cereb. Cortex 11, 1182 – 1190. Smith, A.T., Williams, A.L., Singh, K.D., 2004. Negative BOLD in the visual cortex: evidence against blood stealing. Hum. Brain Mapp. 21, 213 – 220. Somers, D.C., Dale, M.A., Seiffert, A.E., Tootell, R.B.H., 1999. Functional MRI reveals spatially specific attentional modulation in human primary visual cortex. Proc. Natl. Acad. Sci. U. S. A. 96, 1663 – 1668. Stettler, D.D., Das, A., Bennett, J., Gilbert, C.D., 2002. Lateral connectivity and contextual interactions in macaque primary visual cortex. Neuron 36, 739 – 750. Talairach, J., Tournoux, P., 1988. Co-planar Stereotaxic Atlas of the Human Brain. Thieme, Stuttgart. Tootell, R.B.H., Hadjkhani, N., 2001. Where is Fdorsal V4_ in human visual cortex? Retinotopic, topographic and functional evidence. Cereb. Cortex 11, 298 – 311. Tootell, R.B.H., Hadjikhani, N., Vanduffel, W., Liu, A.K., Mendola, J.D., Sereno, M.I., Dale, A.M., 1998a. Functional analysis of primary visual cortex (V1) in humans. Proc. Natl. Acad. Sci. U. S. A. 95, 811 – 817. Tootell, R.B.H., Hadjikhani, N., Hall, E.K., Marett, S., Vanduffel, W., Vaughan, J.T., Dale, A.M., 1998b. The retinotopy of visual spatial attention. Neuron 21, 1409 – 1422.

A. Kraft et al. / NeuroImage 28 (2005) 453 – 463 Tootell, R.B., Mendola, J.D., Hadjikhani, N., Liu, A.K., Dale, A., 1998c. The representation of the ipsilateral field in human cerebral cortex. Proc. Natl. Acad. Sci. U. S. A. 95, 818 – 824. Walker, G.A., Ohzawa, I., Freeman, R.D., 1999. Asymmetric suppression outside the classical receptive field of the visual cortex. J. Neurosci. 19, 10536 – 10553.

463

Williams, A.L., Singh, K.D., Smith, A.T., 2003. Surround modulation measured with functional MRI in the human visual cortex. J. Neurophysiol. 89, 525 – 533. Warnking, J., Dojat, M., Gue´rin-Dugue´, A., Delon-Martin, C., Olympieff, S., Richard, N., Che´hikian, A., Segebarth, C., 2002. fMRI retinotopic mapping—Step by step. NeuroImage 17, 1665 – 1683.