Eye movements in visual search indicate impaired saliency processing in Parkinson's disease

Abstract

Previous studies have produced contradictory evidence on the nature of the visual search impairment in patients with Parkinson's disease (PD). Eye movements were measured during multi-target search in nine individuals with mild-to-moderate PD. Subjects were asked to click on a response button whenever they judged they were fixating a target for the first time. Compared to age-matched healthy volunteers, PD patients were impaired at efficient search (detecting “+”s amongst “L”s) but not inefficient search (“T”s amongst “L”s). However, these patients had normal memory for locations as indexed by their rate of re-clicking on previously inspected locations. We suggest that the pattern of gaze for efficient search may reflect impaired saliency processing in PD.

Introduction

Idiopathic Parkinson's disease (PD) is a neurodegenerative disorder caused by a loss of dopaminergic innervation of the basal ganglia (BG). Although PD is commonly assumed to be disorder of motor control, recent studies have implicated a range of cognitive and visual deficits in PD, including deficits in visual search performance. The nature of the visuo-cognitive deficit in visual search in PD, however, remains unclear (Horowitz et al., 2006).

In visual search, participants look for a target item embedded amongst non-target items (distractors). Search for a “+” amongst rotated “L”s, or a search for a “T” amongst rotated “L”s, are considered classic search tasks, the former termed “efficient,” and the latter “inefficient.” In efficient search, where response times (RT) are independent of the number of distractors, search is commonly assumed to involve stimulus-driven parallel processing in early cortical visual areas. By contrast, inefficient search — where RT increases with distractor number — is often hypothesized to necessitate focal attention and working memory resources for a variety of cognitive operations, e.g., including keeping track of locations already attended.

Several studies have found impairment in efficient, but not inefficient, search in patients with PD using manual RT to index performance (Horowitz et al., 2006). Others have found no such pattern, but that only a subgroup of PD patients who were classified as “frontally impaired” exhibited impairments of both efficient and inefficient search. Finally, Horowitz et al. (2006) found no impairment on efficient search, but impaired performance in inefficient search when targets were unknown. However, to-date there has been no reports concerning the nature of eye movements during search in patients with PD. Here we measured eye position during efficient and inefficient multi-target search, in order to determine whether differences in gaze pattern between patients and controls resulted in impaired visual search. We also asked the novel question: Do patients with PD fail to remember the locations of items already visited during search? In order to assess this we also asked participants to “click” manually on a response button only when they judged that they were looking at a “new” target, but not to “re-click” on targets they recognized they had found earlier. Clicking does not alter the search display and thus the frequency of re-clicking on a target can index a failure to remember the location of a previously found target across saccades (Husain et al., 2001; Mannan et al., 2005).

Methods

Nine patients with mild-to-moderate, clinically defined idiopathic PD (eight patients at Hoehn and Yahr stage of 1 or 2, and one at stage 3, M=1.8, SD=0.6), together with 10 age-matched healthy volunteers. Mean disease duration was 5 years (range 4–7), and mean patient age was 63 years (range 55–70). Control mean age was 66 years (range 61–70). Participants were screened for dementia using a version of the Mini Mental State Examination (MMSE) (all participants scored above 28, with no significant difference between groups). All participants were corrected to 6/12 vision in both eyes. Eight out of nine patients were on levodopa; five were also taking benzhexol while one patient was taking pergolide. Patients were on medication during testing. Eye movements were measured using an infrared video-based eye tracker, the Eyelink I (SR Research Ltd.), whilst patients searched for either (i) 19 “+” targets embedded amongst 44 L's, (ii) 19 “T” targets amongst 44 L's. All participants completed a minimum of nine of each type of search screen (for details see Husain et al., 2001; Mannan et al., 2005). All data was entered into a two-way mixed ANOVA with Group (patients vs. controls) and Search Type (efficient vs. inefficient) as factors. Significance was set at p<0.05.

Results

First, although there was no difference in the time taken to complete search between groups, PD patients crucially missed significantly more targets than controls in efficient but not inefficient search (Table 1A–B). Thus, in terms of our two outcome measures (time taken to complete search and number of targets found), patients are impaired in efficient — but not inefficient — search performance.

Table 1.

Comparison of visual search performance in patients with Parkinson's disease and age-matched controls

		Target		Patients	Healthy volunteers
A	Response time (s)	+		19.2 (1.4)	18.8 (2.4)
			T	23.5 (1.3)	23.8 (2.5)
B	Targets missed	+		2.4 (0.5)	1.9 (0.3)
			T	1.9 (0.5)	1.9 (0.5)
C	Re-fixation rate on targets	+		51.2 (6.5)	24.2 (3.4)
			T	65.3 (6.3)	34.0 (5.3)
D	Re-click rate on targets	+		5.9 (1.3)	6.3 (0.7)
			T	8.6 (1.8)	8.4 (0.9)
E	Median interval between fixation and re-fixation	+		2.8 (0.6)	2.1 (0.5)
			T	2.2 (1.0)	4.4 (1.0)
F	Distractors not fixated	+		19.4 (0.8)	31.6 (1.5)
			T	17.1 (0.8)	21.7 (1.4)
G	Re-fixation rate on distractors	+		12.1 (1.2)	2.1 (0.4)
			T	22.1 (1.7)	9.6 (1.6)
H	Dwell time on targets (ms)	+		319.5 (23)	395.3 (30)
			T	323.2 (22)	356.9 (30)
H	Dwell time on distractors (ms)	+		228.2 (17.7)	200.0 (32)
			T	232.0 (15.0)	225.0 (19)

Values represent means and standard deviations are given in parenthesis. Values for efficient search denoted by “+” symbol and for inefficient search by “T” symbol.

Second, the pattern of eye movements for patients differed significantly from that of controls for both efficient and inefficient search in several parameters. For example, the refixation rate, defined as the [total number of refixations on targets/total number of targets fixated], was significantly higher for patients in both efficient and inefficient search (Table 1C). The median interval between a fixation on a target, and a subsequent refixation, in terms of the number of intervening fixations, also varied significantly between the two groups (Table 1E). Patients demonstrated significantly longer intervals than control subjects for efficient search — but significantly shorter intervals for inefficient search. It appears that patients adopt a different oculomotor strategy to controls during inefficient search, making many, short interval, refixations on a target — despite the fact that overall search performance remains unimpaired. Importantly this high target refixation rate for these patients with mild to moderate PD appears not to be due to a spatial working memory deficit for previously fixated targets, as indexed by normal re-click rates (Table 1D). Note also that the distractor refixation rate was also significantly higher for patients on both types of search (Table 1G).

Finally, patients also exhibited abnormal gaze patterns selectively for efficient — but not inefficient — search in a number of ways. They fixated more distractors than controls on efficient search (Table 1F). Furthermore, while dwell time on targets (total time spent fixating an item) was significantly shorter for efficient search, dwell times on distractors were longer for patients with PD (Table 1H). Thus, these patients appear to be captured more by distractors while spending less time processing targets than healthy controls.

Discussion

The present study investigated visual search in patients with mild to moderate PD, using a multi-target paradigm. Our results show that patients were impaired on the efficient search task (search for +'s amongst L's), finding fewer targets than age-matched control subjects. By contrast, they were unimpaired searching for T's amongst Ls (inefficient search).

We found no evidence to support the proposal that there might be a spatial working memory deficit for locations visited during search in this group of mildly affected patients. Patient re-click rates did not differ significantly from that of age-matched controls, indicating good memory for locations of targets already identified. Such a finding is consistent with studies suggesting that a spatial working memory deficit in PD does not occur in the early stages of the disease (Kemps et al., 2005).

We interpret our results as follows. Target detection relies both on top-down processing (knowledge of the target properties) and bottom-up (stimulus defined) processing. When the target is visually highly salient from distractors, bottom-up processing pre-dominates, allowing the target to “pop-out” from the background of distractors. A reduction in target saliency will therefore affect efficient, compared to inefficient search, to a greater extent. Thus, we find that our patients, during efficient search, miss more targets, fixate more distractors, and make more refixations on both targets and distractors, exhibiting abnormally short dwell times on targets. Although the pattern of gaze was also abnormal in inefficient search, particularly in terms of refixation rates, patients appear able to compensate for the (putative) saliency deficit in inefficient search. A saliency deficit in PD is consistent with a number of other studies of visual search (see Horowitz et al., 2006, for a review). Dopamine is known to mediate saliency processing within the BG and associated cortico-striatal pathways. The findings presented here would be consistent with a moderate deficit in the representation of visual salience in individuals with PD.