Rehabilitation of Unilateral Spatial Neglect: New Insights From Magnetic Resonance Perfusion Imaging

Neuroanatomic Evidence for Heterogeneity in USN 

It is widely agreed that there are various forms of USN, although authors vary considerably in how they divide the syndromes. Investigators have shown dissociations between near versus far neglect, intentional versus attentional neglect, visual versus motor or tactile neglect, and personal versus peripersonal neglect.1, 2, 3, 4, 5, 6 Another distinction has been between different reference frames of USN.7, 8, 9, 10, 11, 12, 13, 14 That is, some patients neglect the contralesional side of their view (as defined by the midsagittal plane of the viewer’s head or trunk; viewer-centered neglect), whereas others neglect the contralesional side of each stimulus in the view (stimulus-centered neglect). Some patients have both viewer-centered (egocentric) and stimulus-centered (allocentric) neglect. However, the percentage of all stroke patients with both types is relatively small. In a study of 100 patients with acute right hemisphere, supratentorial, ischemic stroke, 17 patients showed egocentric visual USN, 4 showed allocentric USN, and only 2 showed both.13 On a test of tactile neglect that distinguishes between these forms of USN, 19 patients had egocentric USN, 1 had allocentric USN, none showed both types. Likewise, in a study of 50 patients with right hemisphere stroke, 11 had egocentric, 4 had allocentric, and only 1 had both types of USN.14 A separate study showed that the distributions of allocentric versus egocentric neglect were opposite in patients with acute stroke in left hemisphere, even though right neglect was as common after left hemisphere stroke as left neglect is after right-hemisphere stroke15 (fig 1).

View Large Image Download to PowerPoint

Fig 1. Distributions (in percentage of total patients in study) of different types of USN after right-hemisphere stroke (left neglect) and after left-hemisphere stroke (right neglect).

Although dissociations between patterns of neglect have been clearly shown, most studies have failed to identify different brain lesions that lead to the distinct types.16 One potential reason for this failure is that patients are often studied long after stroke, so that they may have recovered from 1 or more forms of USN caused by their lesions. Furthermore, when only chronic patients are studied, the infarcts of the subjects are generally large (because patients with small infarcts usually enjoy rapid recovery from USN). It is difficult to determine what part of a large infarct is responsible for the residual form of neglect. Studying patients in the acute stage of stroke allows one to study patients with small infarcts and to identify the form(s) of USN before substantial recovery or rehabilitation.

However, in the acute stage of stroke, deficits depend not only on the infarct, but also areas of dysfunctional tissue surrounding the infarct caused by low blood flow (hypoperfusion). Both the acute infarct and regions of hypoperfusion can be identified with advanced MRI and computed tomography methods. For example, magnetic resonance diffusion-weighted imaging (DWI) shows densely ischemic or infarcted tissue in the first few minutes to hours of stroke onset, and magnetic resonance perfusion-weighted imaging (PWI) shows areas of hypoperfusion that correspond to dysfunctional tissue.17, 18 Several studies have shown that the abnormalities on DWI and PWI together can account for neglect in the acute stage of stroke. For example, the volume of hypoperfusion on PWI is strongly associated with error rate on a battery of tests of USN.19 Furthermore, changes in the volume of hypoperfusion shown by PWI correspond to changes in error rate on line cancellation tests for USN.20 Furthermore, in patients with purely subcortical infarcts on DWI, the presence of cortical hypoperfusion on PWI was strongly associated with the presence of USN in acute stroke. That is, subcortical strokes without cortical hypoperfusion did not cause neglect.21, 22 These results show that identifying sites of hypoperfusion as well as infarct is essential in studying lesion-deficit correlations, at least in the acute stage of stroke.

Using these imaging techniques along with tests of USN that distinguish between neglect in different reference frames, in the first 48 hours of stroke symptoms, we have identified distinct regions of hypoperfusion associated with viewer-centered (egocentric) neglect versus stimulus-centered (allocentric) neglect. In 1 study, a battery of tests of neglect, including reading words in various orientations, reading and spelling words and sentences, line cancellation, line bisection, copying, drawing, and tests of visual, tactile, and motor extinction were administered to 95 patients with right subcortical or cortical infarcts within 48 hours of symptom onset. DWI and PWI scans were obtained the same day. Viewer-centered neglect, characterized by omission of words or other stimuli on the left side of the page or view, omission of lines or figures on the left side of the page, and so on, was strongly associated with hypoperfusion and/or infarct of right angular gyrus ( test=23.2, P<.001), right supramarginal gyrus ( test=12.0, P<.001), and right visual association cortex ( test=43.5, P<.001).23 Viewer-centered neglect was not associated with hypoperfusion and/or infarct of right superior temporal gyrus or posterior, inferior frontal gyrus. Left stimulus-centered USN, characterized by errors on the left sides of words on both sides of sentences, errors copying the left sides of figures (on both sides of the page) was strongly associated with hypoperfusion and/or infarct of superior temporal gyrus ( test=15.2, P<.001) and not with hypoperfusion and/or infarct in right inferior frontal gyrus, angular gyrus, supramarginal gyrus, or visual association cortex (fig 2).

View Large Image Download to PowerPoint

Fig 2. (A) DWI and PWI sequences showing hypoperfusion of right angular (yellow arrows) but not superior temporal gyrus (orange arrows) in a patient with left viewer-centered USN, indicated by omissions on the left side of the view (right figure). (B) DWI and PWI sequences showing hypoperfusion and infarct of left superior temporal gyrus (orange arrows) but not angular gyrus (yellow arrows) in a patient with left stimulus-centered USN, indicated by omissions on the left side of stimuli, even on the right side of the viewer. In this and other figures, hypoperfused areas are blue on PWI; normally perfused areas are light green.

Similarly, in a separate study, a series of 50 patients with acute ischemic stroke restricted to right subcortical areas was administered a task specifically designed to distinguish left viewer-centered (egocentric) USN from left stimulus-centered (allocentric) USN.14 Patients underwent DWI and PWI and completed a test described by Ota et al12 to distinguish viewer- versus stimulus-centered neglect within 48 hours of onset. Patients were presented with a page of 30 circles, of which 10 had left-sided gaps, 10 had right-sided gaps, and 10 had no gaps. Patients were asked to circle each complete stimulus (with no gap) and to cross out each stimulus with a gap. Viewer-centered neglect was characterized by failure to mark stimuli on the left side of the page but correct marking of stimuli on the right side of the page (including stimuli with left-sided gaps). Stimulus-centered left neglect was characterized by marking all stimuli on the page, but incorrectly circling stimuli with left-sided gaps (indicating failure to detect left gaps on both sides of the viewer). In this study, left viewer-centered neglect was most strongly associated with hypoperfusion of right angular gyrus (Fisher exact test, P<.001). In contrast, stimulus-centered neglect was strongly associated with hypoperfusion of right superior temporal gyrus (P<.001). Viewer-centered left neglect was also strongly associated with right posterior-inferior frontal gyrus and visual association cortex (P<.001), consistent with several studies that have shown a role of prefrontal cortex and visual areas in spatial attention.

A recent reanalysis of data from the study of 50 patients with subcortical stroke showed that nearly all patients with viewer-centered neglect with hypoperfusion of right angular gyrus also showed hypoperfusion of right intraparietal sulcus. This finding was consistent with the proposal of Corbetta and Shulman24, 25 that left (viewer-centered) spatial neglect requires damage or dysfunction of 2 separate attentional mechanisms with different localizations: (1) a top-down, spatially specific attentional mechanism in left or right intraparietal sulcus (dorsal parietal), biased toward attending to space in the opposite visual field or space defined by the midline of the trunk and (2) a bottom-up attentional mechanism devoted to reshifting of attention toward a stimulus on either side of space, located in right ventral parietal (angular gyrus−temporoparietal junction). On this hypothesis, USN is more common after right-hemisphere stroke than left-hemisphere stroke, because a single lesion can affect both mechanisms only in the right hemisphere. There is evidence from both functional imaging studies and lesion studies, as well as from acute lesion studies using PWI, for this account of left viewer-centered (egocentric) USN.

The distinct sites of imaging abnormalities associated with viewer-centered versus stimulus-centered neglect have potential implications for rehabilitation. The early studies indicate that rehabilitation of USN in patients with right parietal lesions might be most effective if it focuses on improving attention to the left sides of space as defined by the midline of the trunk or view. In contrast, rehabilitation of USN in patients with right temporal lesions might be most effective if it focuses on improving attention to the left sides of individual stimuli. However, the recent re-analysis, which provided evidence for the 2-component model of viewer-centered neglect, suggests that therapy for viewer-centered neglect may require treatment of both the spatially specific component of exploratory attention and the nonspatial component of vigilance and reorienting of attention to stimuli within a viewer-centered workspace. Several recently described novel ways to induce shifting of attention to the left, using caloric stimulation, repetitive transcranial magnetic stimulation (TMS), vibrotactile stimulation of the neck muscles, and prism adaptation, are briefly reviewed below. In addition, novel methods for improving vigilance or increasing the viewer-centered attentional window or workspace are described. However, these sorts of treatments would not be likely to benefit patients with left stimulus-centered neglect, because they neglect the left sides of individual stimuli, even on the right side of space or view. Innovative approaches to treatment of left stimulus-centered USN are also briefly described but require further evaluation of their effectiveness.

Treatment of Viewer-Centered USN 

As mentioned, viewer-centered neglect may have 2 components—abnormal spatial attention biased to the contralateral side of the viewer, and abnormal nonspatial attention, with impaired reorienting of attention or an abnormally reduced window of attention. The second component of neglect has been described by Redding and Wallace26 as a pathologically reduced size of the work space for a task (what they call “calibration dysfunction”). Rehabilitation of USN is likely to be most effective if both are treated.

An abnormal shift of attention to the right in patients with left neglect has recently been found to improve after prism adaptation. This intervention entails practice in pointing to toward visual targets presented at midline that are optically displaced to the right by 10° or 15° with prisms. This practice has resulted in reduced symptoms of USN for several hours or even days after practice with prisms.26, 27, 28, 29, 30 Even neglect in the tactile domain has been found to improve after prism adaptation.31 Pointing further toward the left—closer to midline targets—as an aftereffect of prism adaptation is sometimes markedly exaggerated, relative to the aftereffects in normal subjects.

Another recent technique for correcting the abnormal shift to the right of the viewer-centered attention (or coordinates of an egocentric workspace, in Redding and Wallace’s terms) is through caloric stimulation. This intervention involves irrigating the left external auditory canal with cold water (or warm water in the right ear), which results in tonic eye movements to the left side.32, 33 Such change in eye movements should not improve left stimulus-centered neglect, although it might improve left viewer-centered neglect. One study34 showed more improvement in viewer-centered attentional neglect than viewer-centered intentional neglect. However, this type of vestibular stimulation has also been reported to improve neglect of the left side of a mental representation of a view and viewer-centered tactile exploration.35, 36

Another approach to ameliorating the abnormal shift of viewer-centered spatial attention to the right in USN is repetitive TMS (rTMS). This intervention is based on the hypothesis that the spatial attention mechanism in each dorsal parietal cortex (intraparietal sulcus) is biased to the contralateral side. Therefore, a lesion in the right parietal cortex leaves only the left parietal spatial attention mechanism to attend to the right side of the viewer. It is thought that the left hemisphere spatial attention mechanisms compete with right hemisphere spatial attention mechanisms, and that each inhibits the other.37 Repetitive TMS has been used to suppress the left parietal cortex transcollosal suppression of the already compromised right parietal cortex to reduce viewer-centered neglect. Small studies have provided evidence that rTMS over the unaffected left parietal cortex was associated with reduced viewer-centered USN for hours38 or even days to weeks.39

Vibratory stimulation of the posterior neck muscles has also been shown to improve USN.40, 41, 42 One study showed that both caloric vestibular stimulation and neck vibration resulted in a shift of egocentric space to the right, consistent with the hypothesis that this intervention would ameliorate egocentric, rather than allocentric, neglect.42 A positron-emission tomography study immediately after caloric stimulation or neck vibration showed activation of the right insula and parietal cortex with both interventions. The right parietal activation is also consistent with the proposed effects on egocentric neglect.

Other interventions might be required to treat the nonspatial attention component of neglect. This nonspatial attention deficit has been characterized as impaired vigilance or reduced bottom-up reorienting of attention to stimuli on the viewer’s left,24 reduced maintenance of attention during exploration of space,43 or reduced attentional workspace.26 One recent study evaluated the effect of pharmacologic intervention on maintenance of attention during exploration of space in patients with viewer-centered USN. Based on the assumption that noradrenergic pathways are important for maintenance of attention, Malhotra et al43 evaluated the effects of guanficine, a noradrenergic agonist known to improve vigilance, on spatial attention performance in a very small double-blind trial with 3 patients with viewer-centered USN. Two patients with damage involving right temporoparietal cortex, sparing dorsolateral prefrontal cortex, showed increased detection of targets on the contralesional side of their view after guanficine administration but not after placebo administration. The preliminary results indicate that amelioration of nonspatially specific deficits in vigilance (sustained attention) might reduce viewer-centered USN, consistent with the 2-component account of viewer-centered neglect.

Although caloric stimulation activates diffuse areas of the left and right hemispheres, vibratory stimulation of the neck muscles has been shown to selectively activate right insula and parietal cortex in normal subjects.42 Nevertheless, activation of these regions might result in improvement of both viewer- and stimulus-centered neglect. However, improvement in stimulus-centered neglect has not been shown by either of these methods. Likewise, the effect of guanficine might not be specific to viewer-centered neglect. However, viewer-centered neglect seems to result only from a combination of impaired sustained attention (potentially ameliorated by guanficine) and impaired spatial attention to egocentric space.24 There is no evidence that stimulus-centered neglect requires the additional deficit in sustained attention (which may explain why it is at least as common after left hemisphere as right hemisphere dysfunction). Therefore, there is no theoretic reason to believe that guanficine would improve stimulus-centered neglect. Dopamine agonists (eg, bromocriptine) have also been used with variable effects in patients with USN44 and would likely be most effective in augmenting sustained attention to ameliorate viewer-centered neglect. However, in the absence of a randomized trial, the effects of dopamine agonists are unclear. Because there are theoretic reasons for believing that dopamine agonists may improve performance on attention tasks or augment therapy, further studies are required.

Treatment of Stimulus-Centered USN 

Prism adaptation, rTMS, vibrotactile stimulation of the neck muscles, and caloric stimulation would be unlikely to ameliorate stimulus-centered or object-centered neglect, because these interventions should only correct the rightward shift of viewer-centered coordinates of the task workspace (the spatially specific attentional component of USN caused by right intraparietal sulcus damage or dysfunction24). In fact, in 1 study prism adaptation was associated with improved pointing to targets at midline (defined by viewer-centered coordinates) but no change in left USN in selecting the happier/smiling side of chimeric faces.45 The latter task is likely to depend on stimulus-centered attention. Instead, treatment of stimulus-centered USN may require correction of a rightward shift in a stimulus-centered reference frame caused by right temporal cortex damage or dysfunction.

A few studies have shown that therapy focused on increasing the size of the stimulus-centered task workspace or attentional window can reduce symptoms of USN.46 In 1 study, the size of the “attentional window” for stimulus recognition was increased by presenting large circles in the same block of trials with small circles with left-sided or right-sided targets (gaps). Improved detection of left-sided targets in the small circles (but not in the large circles) was achieved when large circles were presented in the same block of trials. It was hypothesized that this improvement was due to increasing the size of the attentional window with the larger circles.47 Other studies have increased the size of the stimulus-centered task work-space by embedding a stimulus in a larger stimulus48 or by shifting the stimulus-centered task workspace to the right of the total visual stimulus by adding irrelevant visual information on the left side of the stimulus.7 This irrelevant information shifted the target portion of the stimulus toward midline (further to the right) of the stimulus and improved detection of the left target on the stimulus itself. These preliminary results suggest that rehabilitation aimed at improving USN by the reduced stimulus-centered attentional window or by correcting the abnormal rightward shift of the stimulus-centered reference frame can reduce symptoms of stimulus-centered neglect. Whether or not these are dissociable components, or if there is another, nonspatial, component of stimulus-centered USN is unclear. Damage to a single area, right superior temporal gyrus, appears to be sufficient to cause stimulus-centered neglect.

The duration of effects of these various rehabilitation techniques is unknown. Prism adaptation has been shown to be associated with improved performance on neglect tasks for up to several days,26, 27, 28, 29, 30 and the effects of rTMS may last at least as long as a few weeks.39 In 1 crossover trial, visual exploration therapy combined with vibration of the posterior neck muscles resulted in more improvement on tests of neglect and daily activities than visual exploration therapy alone for a period of 2 months.41 In a multiple baseline study, 6 patients showed no improvement in untrained behaviors, but neck vibration resulted in improvement of neglect after treatment that was sustained for at least 1.5 years.40 However, in most cases, the long-term effects are less clear. In rehabilitation, long-term effects are generally brought about by continual reinforcement of short-term effects, and this is likely to be true for rehabilitation of neglect as well. Clearly, more research is required to investigate the long-term effects and generalization to untrained tasks and activities of daily living.

This brief review has illustrated how consideration of the type of neglect in a person might influence the selection of a treatment strategy. It was not meant to review all of the approaches to rehabilitation of USN or to critique the approaches with respect to effectiveness, duration, or generalization. For such a review, see Pierce and Buxbaum.44

Fluctuation in Neglect Severity After Stroke 

One barrier to evaluating the effectiveness of rehabilitation of neglect is that the severity seems to vary widely, at least in the first few weeks after stroke, when most rehabilitation occurs. This variation in severity is a nuisance in trying to establish a stable baseline with which performance after treatment can be compared. However, understanding the mechanisms of this fluctuation may also provide insights into achieving the best possible outcomes.

Recent studies of DWI and PWI in acute and subacute stroke have shown that at least one of the mechanisms underlying fluctuation in neglect is change in blood flow. In 1 study, patients with acute, right-hemisphere ischemic stroke were administered tests of neglect (including drawing, copying, reading, line cancellation, line bisection, and gap detection) within 24 hours of stroke onset, and underwent DWI and PWI at the same approximate time. The rate of errors on the neglect battery was significantly correlated with the volume of hypoperfusion on PWI but not with the volume of infarct or dense ischemia on DWI.16 These results indicate that the severity of neglect varies with regional cerebral blood flow. Further studies confirmed this finding and showed that changes in blood flow are strongly associated with changes in neglect. In one recent study of 26 patients with acute right hemisphere ischemic stroke, of whom 10 underwent some sort of therapy to improve perfusion, volume of hypoperfusion correlated with errors on neglect tests (r=.71, P<.002) and changes in the volume of hypoperfused tissue correlated strongly with changes in the rate of errors on a line cancellation test (r=.83, P=.003).49 That is, improved perfusion resulted in improved performance on line cancellation.

A series of longitudinal studies of patients with persistent hypoperfusion up to 10 days poststroke showed that pharmacologic manipulation of systemic blood pressure to improve cerebral perfusion also resulted in changes in performance on neglect tests.50 That is, temporary increases in mean arterial pressure (MAP) in patients with large vessel stenosis and persistent hypoperfusion (without infarct in at least some of the hypoperfused areas) resulted in improved perfusion of right parietal and temporal cortex and improved performance on line cancellation and other tests of neglect. To illustrate, we have recently studied a woman with left stimulus- and viewer-centered USN caused by left common carotid occlusion. She sustained a small right frontal infarct and had a large area of poorly perfused tissue surrounding the infarct that included right superior temporal and inferior parietal cortex (fig 3). She was not a candidate for carotid endarterectomy, stenting, or thrombolysis, because her common carotid artery was completely occluded. Her neglect was mild on initial presentation to the hospital but rapidly worsened by the time the MRI in figure 3 was done a few hours later. We increased her blood pressure with intravenous fluids and later with medications. When her blood pressure was elevated, she showed improved perfusion of right temporal and inferior parietal cortex, and her performance on tests of USN improved. She showed resolution of both viewer- and stimulus-centered neglect with reperfusion. Gradually over the next 2 weeks, we were able to normalize her blood pressure, as collateral circulation provided improved blood flow to right temporal and parietal cortex. She eventually maintained her gains in USN tests with normal blood pressure.

View Large Image Download to PowerPoint

Fig 3. (A) DWI and PWI of a patient with stimulus- and viewer-centered USN before (day 2, left 2 scans) and after (day 3, right 2 scans) intervention with temporary blood pressure elevation. PWI after intervention shows reperfusion of right inferior frontal, superior temporal, and supramarginal gyri when USN had recovered substantially. (B) MAP is shown in the black dotted line. Percentage error on detection of left gaps is shown by the green line; percentage error (of total possible scores) in copying a scene is shown by the blue line. Performance was closely linked to changes in blood pressure (MAP).

Cases such as this one indicate that at least some of the fluctuation in USN observed in patients with acute ischemic stroke is due to changes in regional cerebral blood flow. Therefore, deterioration in spatial attention should prompt a search for anything that might have reduced cerebral blood flow, such as lowering of blood pressure (caused by dehydration, bleeding, antihypertensive medications, some eye drops, nitroglycerine, anxiolytics, and a variety of other medications or substances) or worsening cerebral stenosis. Even positioning can affect blood flow, as most dramatically observed when patients show improved cortical function when supine (or even with heads lowered) and deterioration in cortical function when standing.51 Awareness of the potential causes of worsening perfusion can lead to correction of the cause, improvement in USN, and better capacity to respond to rehabilitation efforts.

This type of reperfusion therapy is likely to have an effect only in the acute stage of stroke, hours to days after symptom onset, when there is still poorly functioning but salvageable brain tissue. The other treatments we have mentioned, such as prism adaptation, rTMS, caloric stimulation, and vibrotactile stimulation may be effective in more subacute or chronic stages of stroke.