Spatial neglect and attention networks

Abstract

Unilateral spatial neglect is a common neurological syndrome following predominantly right hemisphere injuries and is characterized by both spatial and non-spatial deficits. Core spatial deficits involve mechanisms for saliency coding, spatial attention, and short-term memory and occur in conjunction with nonspatial deficits that involve reorienting, target detection, and arousal/vigilance. We argue that neglect is better explained by the dysfunction of distributed cortical networks for the control of attention than by structural damage of specific brain regions. Ventral lesions in right parietal, temporal, and frontal cortex that cause neglect directly impair nonspatial functions partly mediated by a ventral frontoparietal attention network. Structural damage in ventral cortex also induces physiological abnormalities of task-evoked activity and functional connectivity in a dorsal frontoparietal network that controls spatial attention. The anatomy and right hemisphere dominance of neglect follow from the anatomy and laterality of the ventral regions that interact with the dorsal attention network.

Figures

FIGURE 1. Behavioral and lesion analyses of the egocentric spatial bias in neglect patients

A) Effects of stimulus salience on eye movements of neglect patients. Patients were instructed to saccade to a target letter in an array. A sequence of arrays was presented, some containing a target, some containing a distinctive probe stimulus of a higher luminance or different orientation than the distracter squares, and some containing both a target and a probe. The right graph shows that targets (red line) and irrelevant but distinctive luminance probes (green line) produced a similar ipsilesional bias for saccades, indicating a similar gradient for automatic and goal-directed orienting (Bays et al 2010). B) Neglect patients show an ipsilesional gaze bias while searching for a target letter in a letter array (blue traces) and at rest (green traces). Non-neglect patients (bottom) showed no bias (Fruhmann Berger et al 2008). C) Lesions classically associated with neglect involve an ipsilesional bias within an egocentric reference frame, but not biases observed within stimulus-centered or object-centered frames. Each image shows the voxelwise lesion distribution associated with deficits within a particular reference frame (Medina et al 2008).

FIGURE 2. Relationship between anatomical distribution of lesions associated with neglect, attentional networks, and damage to fiber tracts

A) Anatomical regions associated with neglect, as shown by lesion-symptom mapping (left panel)(Karnath et al in press), overlap of lesions in patients diagnosed with neglect (middle panel)(A.R. Carter, G.L. Shulman, and M. Corbetta, unpublished data), and comparisons of groups of patients with severe neglect vs. no neglect (right panel)(Verdon et al 2010). The three cortical distributions are quite similar and emphasize ventral regions in IPL (Angular and Supramarginal Gyri), TPJ, STG, and VFC. B) The dorsal (left panel) and ventral (right panel) attention networks as determined by resting state functional connectivity in 25 healthy controls. The ‘yellow-orange’ voxels indicate regions with strong and significant temporal correlation at rest. The ‘blue’ voxels in the left panel indicate regions negatively correlated with the dorsal network, corresponding to the default network. The anatomical distributions shown in Figure 2A match the distribution of the ventral but not dorsal attention networks. Seed regions for the connectivity analysis were determined by meta-analyses of task activation paradigms (shown in Figures 3A and 6A). Dorsal seed regions were determined from the activations evoked by a central cue to shift attention and from comparisons of activations evoked by peripheral cues to shift vs. maintain attention. Ventral seed regions were determined from comparisons of activations to targets presented at unattended and attended locations. C) Slice representations from the anatomical distributions shown in the left (Karnath et al in press) and middle panels (A.R. Carter, G.L. Shulman, and M. Corbetta unpublished data) of Figure 2A indicate that anatomical damage includes both gray and white matter. White matter tracts corresponding to the arcuate fasciculus (AF) and superior longitudinal fasciculus (SLF) II and III, as determined by diffusion tensor imaging (DTI) in 30 healthy controls, are shown in the right panel (M.Glasser and K.Patel, unpublished data). The regions of maximal damage in neglect patients, shown in the left and middle panels, are outlined in green and blue. Lesion damage overlaps most strongly with the AF, but also with SLF II and III. These tracts connect regions within and across networks.

FIGURE 3. Physiology of spatial attention in healthy adults

A) Dorsal fronto-parietal regions are activated following a central cue to shift attention. The statistical map shows the z-map from a meta-analysis of four experiments (n=58) in which BOLD activity was measured following a central cue to shift attention to a peripheral location (Astafiev et al 2003, Corbetta et al 2000, Kincade et al 2005). The time course of the response to the cue shows bilateral activity from right IPS with a contralateral preponderance indicating spatial selectivity. B) Occipital and dorsal fronto-parietal regions show spatially selective attentional modulations following a stimulus-driven shift of attention (from a meta-analysis of two experiments (n=47), (Shulman et al 2009) and (A. Tosoni, G.L. Shulman, D.L.W. Pope, McAvoy, M.P., and M. Corbetta, unpublished data). Subjects were cued to attend left or right to detect targets in a rapid-serial-visual-presentation (RSVP) stream presented among distracter streams. The z-map indicates voxels showing contralateral activity > ipsilateral BOLD activity following a shift of attention to the peripheral cue (red square). Note the strong spatially selective response in right IPS and visual cortex for shifting and attending to contralateral rather than ipsilateral stimulus streams. Also, maps for purely endogenous (Panel A) vs. stimulus-driven (Panel B) shifts of attention are very similar. C) Contralateral topographic maps in dorsal parietal cortex. The left image shows five contralateral polar angle maps along R IPS. The right image shows the activations in these maps during a VSTM task in which subjects remembered the orientation and location of target lines presented among distracters. The bottom graph shows a comparison of the magnitude of contralateral and ipsilateral activations in left and right IPS maps as a function of VSTM load. While left and right IPS contains contralateral polar angle maps (left IPS not shown), right IPS was equally activated by VSTM load in the contralateral and ipsilateral hemifieds but left IPS was only modulated by load in the contralateral hemifield (Sheremata et al 2010). This pattern of activity matches that postulated by the ‘standard’ model for neglect (Mesulam 1981). D) Inter-hemispheric coding of spatial attention. BOLD activity was measured following an auditory cue to attend to a peripheral location. The top right graph shows the magnitude of activity in L and R FEF on a trial-to-trial basis following leftward (blue dots) and rightward (red dots) cues. Activity in L and R FEF is highly correlated across trials, but a contralateral signal is superimposed on the positively correlated ‘noise’ (i.e. blue dots plot above red dots). This correlated ‘noise’ is partly explained by the presence of strong correlations at rest (bottom graph) between homologous regions (e.g. left-right FEF) or parts of maps (e.g. left-right fovea in V1). The locus of attention is only weakly predicted (AUC value=~.60, chance=.50) by ‘reading out’ activity only from the portion of the map in visual cortex or area (e.g. FEF) coding for the attended location. The prediction increases significantly (AUC=~.80) when subtracting activity from the attended-minus-unattended homologous portion of map or area in the two hemispheres (Sylvester et al 2007).

FIGURE 4. Physiology of egocentric spatial bias in neglect patients

A) Statistical map of BOLD activations during a Posner spatial attention task (same as in Figure 3A), in which subjects are cued to a peripheral location and detect a subsequent target. Right hemisphere acute neglect patients show hypo-activation of both hemispheres (right>left) that partly recovers at the chronic stage. The dark shading in the anatomical image indicates the distribution of structural damage. (Corbetta et al 2005) B) As a result of right hemisphere hypo-activation, acute patients show a large imbalance of BOLD activity in IPS/SPL, with much greater left than right hemisphere activity. This imbalance normalizes at the chronic stage. The left columns show statistical maps of activity in parietal cortex, the right column shows the averaged time course of activity time-locked to the presentation of the cue over a trial of the Posner task in left (blue line) and right (red line) parietal cortex. (Corbetta et al 2005). C) Acute neglect patients (top graph) show low correlations in BOLD spontaneous activity between homologous regions of left (red lines) and right (blue lines) parietal cortex, but the correlation recovers at the chronic stage (bottom graph)(He et al 2007). D) Abnormal physiological signals in the dorsal attention network of acute neglect patients are functionally significant. Left graph: Left parietal activity was stronger in subjects with more severe neglect, as indexed by the difference in response times to contralesional vs. ipsilesional visual targets. Right graph: reduced inter-hemispheric correlation within fronto-parietal regions of the dorsal attention network correlates with the severity of neglect of the left visual field, as indexed by longer RTs to contralesional vs. ipsilesional visual targets (Carter et al 2010).

FIGURE 5. Behavioral analyses of ‘non-spatial’ deficits in neglect patients and healthy adults

A) Reorienting deficits in neglect patients with VFC and TPJ lesions (Rengachary et al in press). Patients detected a visual target (the asterisk) that occurred in a validly cued location (indicated by the dotted circle) or at an invalidly cued location (shown in the figure). Both TPJ and VFC patients showed large contralesional deficits in reorienting, as indexed by longer RTs to unattended (invalid) than attended (valid) targets. The VFC group additionally showed reorienting deficits in the ipsilesional field and larger overall detection deficits. Similar results were observed for accuracy (not shown), but the TPJ group showed evidence of a small reorienting deficit in the ipsilesional field. B) Detection deficits in neglect patients. Neglect patients show abnormally slow simple RTs to an ipsilesional auditory stimulus (Samuelsson et al 1998). Controls were healthy age- and gender-matched subjects. The mild and severe groups consisted of non-neglect patients with minor and major right hemisphere strokes. C) Arousal deficits in neglect patients. Parietal neglect patients show a vigilance decrement (red curve) in a task that involved detection of letter targets in two locations (indicated by arrows) within a central column. No deficit is observed in right hemisphere patients without neglect (green curve). The anatomical images show the association of damaged voxels in right TPJ with the vigilance decrement, with darker areas indicating a weaker association (Malhotra et al 2009).

FIGURE 6. Physiological analyses of ‘non-spatial’ deficits in neglect patients and healthy adults

A) Physiology of reorienting in healthy adults. BOLD activity was compared for visual targets that occurred in invalidly cued (left panel) vs. validly cued (right panel) locations. The statistical map shows the z-map from a meta-analysis (n=58) of four experiments (Astafiev et al 2003, Corbetta et al 2000, Kincade et al 2005). Strong activations are observed in right TPJ, extending from posterior temporal cortex (STG) to ventral IPL (SMG), and in right IFG/insula. These ventral regions are co-activated with dorsal attention regions in IPS and FEF. B) Physiology of detection in healthy adults. A spatial cue directed subjects to attend to a rapid-serial-visual-presentation (RSVP) stream in the left or right visual fields that might contain a target. The z-map shows all voxels with greater right than left hemisphere BOLD activity to detected targets. The map includes ventral voxels that showed significant activations from reorienting in panel D, but also additional regions in prefrontal cortex. The dorsal attention network, however, largely does not show an asymmetry. C) Physiology of arousal in healthy adults. A meta-analysis of foci from 5 experiments reporting activations that index arousal/vigilance, visual (red spheres) or auditory (green spheres)(Coull et al 1998, Foucher et al 2004, Paus et al 1997, Sturm et al 1999, Sturm et al 2004). The right hemisphere foci are superimposed on the average of z-maps for reorienting and detection-related hemispheric asymmetries from panels A and B. The dark shading shows the distribution of neglect lesions from a recent neglect study (Rengachary et al 2009). Many more foci are observed in the right than left hemispheres. Foci are clustered around the TPJ and insula/frontal operculum regions that are activated by reorienting and target detection, but many prefrontal foci are anterior to both distributions. Few foci occur in dorsal fronto-parietal regions, indicating that arousal-related activations overlap more the ventral than dorsal attention network.

FIGURE 7. Pathophysiology of spatial neglect

A) In the healthy brain, activity during visual search is symmetric, and inter-hemispheric interactions between left and right dorsal attention and visual occipital areas are balanced. Each side of the dorsal attention network direct shifts of attention and eye movements contralaterally, and the locus of spatial attention is coded by a differencing mechanism that takes into account activity from both hemispheres, as described in Figure 3D and in (Sylvester et al 2007). Balanced interhemispheric activity results in a normal eye movement search pattern, shifts of attention, and coding of stimulus saliency. The ventral network is lateralized to the right hemisphere due to a slight asymmetric (right>left) arousal input from the brainstem LC/NE system, and interacts with the dorsal network (right>left). Accordingly, decreases in arousal shifts spatial attention rightward because of greater left than right activity in the dorsal attention network, while under normal conditions spatial attention shows a slight leftward bias due to slightly greater right than left dorsal activity. B) In a patient with a ventral stroke, direct damage of ventral regions causes a reduction of arousal, target detection, and reorienting that leads to a bilateral visual field impairment. Abnormal ventral-to-dorsal interactions cause a dorsal imbalance with a relative over-activation of left dorsal spatial maps, leading to tonic and task-dependent rightward spatial biases in attention, eye movements, and stimulus salience.