Visual Search in Behavioral Variant Frontotemporal Dementia

Abstract

Background:

Changes to visual search have shown specific patterns in a number of dementia subtypes. The cortical regions involved in the control of visual search overlap with the regions affected in behavioral variant frontotemporal dementia (bvFTD). Previous literature has examined visual search in bvFTD with smaller array sizes.

Objective:

To examine the pattern of behavior shown by bvFTD patients while undertaking visual search in the presence of larger numbers of distractors to model increased cognitive load.

Methods:

15 bvFTD and 17 control participants undertook three visual search tasks: color, orientation, and conjunction searches. A wide range of array sizes was used, from 16 to 100 items arranged as a square. Behavior was quantified using accuracy, response time, and eye movements.

Results:

BvFTD participants displayed a reduction in accuracy and an increased response time across all task types. BvFTD participants displayed an increase in number of objects examined and number of fixations made for color and conjunction tasks. Fixation duration was increased for orientation and conjunction (the more difficult tasks) but not color search. Results indicated the increase in time to response to be due to an increased intercept, with no significant difference in slope for the different tasks.

Conclusion:

BvFTD participants display a pattern of visual search behavior consisting of a decrease in accuracy, an increase in response time, and a corresponding increase in the number and length of eye movements made during visual search. The pattern seen corresponds to studies of frontal lobe damage, while differing in pattern from that seen in a range of other cognitive conditions.

Keywords

Eye movements frontotemporal dementia search visual scan-paths

INTRODUCTION

In everyday life, we are constantly presented with search tasks: finding your keys, a familiar person in a crowd, or locating an item on a supermarket shelf. Common visual search tasks often include a significant number of distractors, and the behavior displayed in visual search can be examined in a laboratory or clinical setting, including dissecting the search task into different components. Different search behaviors can be elicited by different instructions or by changing the target or number of distractors. Broadly, search tasks can be broken into feature and conjunction tasks. Feature search tasks are those in which the target is clearly identified by one aspect (e.g., color or orientation), while conjunction search tasks are those in which more than one feature must be combined to identify the target (e.g., color and orientation together). Using different search paradigms, different areas of cognitive processing and behavior can be investigated, and the pattern of eye movements made during visual search has been identified as a useful tool in investigating dementia patients’ behavior [1].

Classically two main types of visual search processes are theorized: parallel processing resulting in pop-out identification of the target (usually occurring with feature searches) and serial processing (usually associated with conjunction search tasks) where multiple items must be attended to and then discarded before identifying the target. Serial search processes typically result in increased reaction times with increasing numbers of distractors [2]. Other theories have described visual search in two parts: the first providing limited parallel processing of the entire scene and the second directing attention to different locations based on the features extracted during the first part [3, 4]. These other theories may account for the behavior observed in more challenging feature search tasks in which serial search behavior is observed. Despite the target being identifiable based on a single element, this signal may not be large enough and the information extracted during part one may not be enough to identify the target, thus a serial guided search may be required.

Once a participant has searched the scene, a response of target-present or target-absent is generally made. The response then requires initiation, planning, and action to indicate which alternative is selected. The behavior during a search task can be examined by extracting objective measures: the number of errors, accuracy, reaction times, and eye movements.

Visual search produces cortical and sub-cortical activation of a frontal-parietal network, the degree of activation of which increases as task difficulty increases [5]. Different quantitative aspects of visual search are linked with different cognitive regions. Examining reaction time efficiency by looking at response time versus array size, parietal activation but not frontal activation is correlated with slope changes with increasing numbers of distractors [6]. Frontal lobe activation correlates with the baseline task difficulty in visual search (indicated by the intercept of the response time versus number of items regression line) [6], a measure which is thought to represent a participant’s ability to react to a stimulus when there are no distractors. Patients with frontal lobe lesions are reported to make more errors in locating targets in the contralateral visual field and exhibit a longer response time for conjunction visual search tasks [7].

Errors in accuracy can occur for a number of reasons, have been linked to multiple brain regions, and can be considered using signal detection theory. Determining whether errors are a missed target or a false alarm gives information about the behavioral style; a high miss rate may indicate apathy with the participant quickly defaulting to responding no, while a high false alarm rate with quick response times may indicate disinhibition. A range of different behaviors can result in the same accuracy and reaction time results [8]. Eye movements have been suggested as an alternative metric during visual search [9], as eye movements give a more detailed, objective measure of the behavior state underlying performance. Such performance depends upon the allocation of attention; the frontal eye fields play a role in directing attention as well as controlling eye movements [10]. We might therefore expect disorders affecting this region to alter visual search performance.

Behavioral variant frontotemporal dementia (bvFTD) is a young onset neurodegenerative condition which causes progressive atrophy of the frontal and temporal lobes, resulting in significant changes to personality and behavior, with apathy the most common behavioral change [11]. BvFTD lies on a spectrum which overlaps with Parkinsonism and motor neuron disease [12, 13]. Impaired areas include the mesial frontal and ventromedial frontal cortex, frontal and posterior orbital gyri, insula, anterior cingulate, and premotor cortex [14, 15], which overlap with those activated in visual search. BvFTD is difficult to diagnose, and its behavioral changes can occur insidiously, with 52% of bvFTD patients receiving a psychiatric diagnosis before their dementia is diagnosed [16].

Behavioral changes in bvFTD can be typically divided into disinhibited or apathetic and include changes to decision making [17]. Atrophic cortical changes are not always easily distinguishable and so differentiating bvFTD from other conditions including Alzheimer’s disease, obsessive compulsive disorder (OCD), and schizophrenia is difficult. Diagnosis is therefore usually made on consensus criteria from a team including neuropsychology, with bvFTD patients displaying an increased latency in decision making, pattern and spatial recognition tasks as well as difficulty with reversal stages of attentional set shifting, and an increase in risk taking [18].

Given the range of cognitive steps required, it is not uncommon for search to be affected in cognitively impaired groups, with each producing a specific pattern of behavioral change. Alzheimer’s disease in conjunction search leads to an increase in missed targets [19], as well as to an increase in reaction time due to both an increased intercept and an increased slope [20]. Eye movements during visual search in Alzheimer’s disease are disorganized [21] and show an increased number of fixations and longer fixation durations than controls [22]. Schizophrenia patients display increased reaction times in conjunction searches, but accuracy in both feature and conjunction tasks is not significantly different to controls [23]. Patients with OCD do not display a longer response time than controls, although an exception has been noted when they have been primed for a different target, in which case they have difficulty changing where their attention is directed [24]. Parkinson’s patients with frontal lobe changes show an increased reaction time due to an increased intercept in visual search, which is thought to represent increased cognitive slowing and decreased vigilance, when compared to Parkinson’s disease patients without frontal changes [25].

Visual search in bvFTD has been briefly explored. Viskontas et al. [26] examined bvFTD participants and compared them with age-matched controls on two feature search tasks (either discriminating by color or discriminating by letters) and on a conjunction task, all presented for 3.5 s, with four array sizes: 1, 5, 15, or 30 items. BvFTD participants performed slower than controls on both feature and conjunction search tasks but only made more errors on the conjunction task. In examining eye movements during these tasks, no significant difference in fixation duration or spread of eye movements was found [26]. The Viskontas study did not examine very large array sizes, or evaluate the individual components of response time (slope and intercept), which are thought to be controlled by different cortical regions.

Our study aimed to follow on from Viskontas et al.’s work [26] by examining both feature and conjunction searches in bvFTD, using larger array sizes to increase the cognitive difficulty and examining the different components of the response, thus further characterizing the effects of bvFTD on this task.

MATERIALS AND METHODS

Three visual search tasks were examined using the Tobii 1750 eye tracker, which presented the stimuli on a 17” monitor and recorded eye movements at 50 Hz. The T1750 has an accuracy reported as 0.5°. The tasks were: 1) feature color (e.g., find the yellow triangle), 2) feature orientation (e.g., find the down-pointing triangle), and 3) a conjunction of color and orientation (e.g., find the red triangle pointing right). These tasks were chosen to represent classic easy feature, difficult feature and conjunction searches. An example of each is presented in Fig. 1. In each task, the total number of elements on screen ranged between 16 and 100, all arranged in a grid. There were ten images at each array size (4×4, 5×5, 6×6, etc.) for both target-present and target-absent trials, for a total of 140 trials in each task.

Fig.1

Examples of the search task presented; color, orientation, and conjunction.

Participants were seated approximately 50 cm from the eye tracker and allowed to move their head freely. Screen resolution was 1024×768.

Participants were given the following instructions on screen: “In this task you need to find the target triangle as quickly as possible. If you find the target, click to the next slide and answer yes. If the target is not there, click to the next slide and answer no. You will need to press the button three times to move on to the next trial, this ensures that if you accidentally make a mistake because you are in a rhythm you can correct it. Once you have answered yes or no look at the central dot until the next triangle page appears.” An example of the target triangle was given before each task. Participants were verbally instructed to press the right arrow button on the keyboard if the target was present, and the left arrow target on the keyboard if it was not. Participants were informed that they would be scored for timing and accuracy. Each search trial was presented on-screen until the participant responded. No feedback as to the accuracy of their decision was given. Participants were given the option to press the button themselves, or to respond verbally with a yes or no and the researcher would press the response button. This mimicked the participants natural response and was thought to decrease the total time required as it eliminated the added complication of having to recall which button was appropriate, and thus more accurately reflected the search process occurring.

All participants were presented with the color search task first as this is the easiest to undertake and represents “pop-out search”. Following this, subjects undertook the serial search tasks: orientation and conjunction (the order of these two later tests was randomized). Within each task type the array sizes were arranged in a pseudo-random order and thus the same stimuli were presented to both bvFTD and control participants. Tasks were not interleaved in order to avoid increasing working memory load.

Participant information

15 bvFTD and 17 control participants were recruited for the study. BvFTD participants were recruited from the Royal Melbourne Hospital Neuropsychiatry Unit after meeting clinical criteria for probable bvFTD [11] from a team including neurology, psychiatry, psychology, and occupational therapy. Participants are part of a larger cohort some of which have been presented elsewhere [27]. There was no significant difference in age or gender distribution between groups. All bvFTD participants underwent an MRI, functional MRI, and cognitive workup as part of their normal diagnostic assessment. The standard measure of general cognition used on the unit is the Neuropsychiatry Cognitive Assessment tool (NUCOG), which assesses cognition in five separate domains and provides a total score out of 100 [28]. Patient participants’ age, NUCOG scores, and medications are presented in Table 1. Participants were excluded if there was any history of stroke, alcoholism or substance abuse, bipolar disorder, major depressive disorder, learning disorder, schizophrenia, acquired brain injury, or any other neurological or psychiatric condition. All participants had contrast sensitivity and corrected visual acuity within normal limits.

Table 1

BvFTD participant age and cognitive scores with abnormal values asterisked. Note scores for the last participant (denoted with a #) are BATCH scores not NUCOG scores

Age	Gender	NUCOG Attention	NUCOG Spatial	NUCOG Memory	NUCOG Executive	NUCOG Language	NUCOG Total	Psychiatric Medications
63	M	16^*	17	12^*	7^*	14.5^*	66.5^*	NIL
57	M	18	18	14^*	12.5^*	15^*	77.5^*	quetiapine XR, diazepam
56	M	11^*	14.5^*	9.5^*	10^*	18	63^*	valproate, lithium carbonate, quetiapine XR
61	F	18	15^*	12.5^*	7^*	19	71.5^*	risperidone
55	F	17	16^*	10.5^*	11^*	20	74.5^*	NIL
51	M	16^*	20	18.5	15.5^*	17^*	87	NIL
61	M	14^*	20^*	16^*	17^*	19^*	86^*	Amitryptiline, clonazepam, valproate
65	M	13^*	14^*	13.5^*	13.5^*	16^*	70^*	Desvenlafaxine, rivastigmine
48	F	10^*	15	8^*	8.5^*	14	55.5	NIL
38	M	12^*	13.5^*	6^*	6^*	16^*	53.5^*	Sodium valproate, clozapine, quetiapine, amisulpride, venlafaxine, benztropine
74	M	17	18	14^*	6^*	18	73^*	Sertraline
53 #	F	17.5	17.9	15^*	15.3^*	16.7	82.4	Aripiprazole, venlafaxine, zopiclone

Analysis

Accuracy, response time, fixation duration, number of fixations before a decision is reached, and number of objects examined were analyzed. In agreement with the decision made by Viskontas et al., as the response data is non-parametric the accuracy measure A’ [29] was calculated using the following formula suggested in Snodgrass and Corwin [30], where H = hit rate and FA = false positive rate: $\begin{matrix} A' = 0.5 + \frac{(H - FA) (1 + H - FA)}{4 H (1 - FA)} \\ (When H is greater than or equal to FA) \end{matrix}$ (1) $\begin{matrix} A' = 0.5 - \frac{(FA - H) (1 + FA - H)}{4 FA (1 - H)} \\ (When FA is greater than H) \end{matrix}$ (2)

Areas of interest (AOI) were drawn adjacent to one another in a grid around each item on display. Results were extracted from the Tobii Clearview (v2.7.1) analysis program. Results were averaged per participant for each array size and then compared between participant groups. Response time was further analyzed by calculating the slope (to examine the effect of extra distractors) and intercept values (to examine baseline task difficulty) across the different array sizes. Scores on the NUCOG were correlated against eye movement and task results.

Analysis was undertaken using R (v1.0.136) and IBM SPSS (v25). ANOVAs, with diagnosis as the between-subjects and array size as the within-subjects independent variables, were calculated with response time, number of fixations, number of objects examined, and errors as dependent variables for each task. Outliers were not removed. Effect sizes are reported as Cohen’s f values.

RESULTS

Technical difficulties in the eye movement recordings and patient participant distractibility resulted in inadequate data from a few patient participants. This resulted in 12 bvFTD and 17 controls for color and conjunction tasks, and 9 bvFTD and 17 controls for orientation searches. NUCOG scores, age, gender, and medications for patient participants with analyzable results are presented in Table 1. A global NUCOG score of less than 80 is indicative of dementia, and a score of 16 or less in any one domain is abnormal. One patient participant did not have NUCOG results but instead had Behavioral Assessment Tool for Cognition and Higher function (BATCH) results which, are correlated with NUCOG scores [31]; however, these results were not included in NUCOG correlation analyses. Response time values failed the assumption of equality of error variances and so a robust ANOVA was performed with a HC3 adjustment. Slope and intercept values were found not to be normally distributed and so the non-parametric Wilcoxon rank sum test was utilized (with r calculated for effect size). Both target-present and target-absent trials produced the same outcomes as all trials combined and so only results for all trials combined are presented here.

Accuracy

For color, ANOVA revealed a significant main effect of diagnosis (F(1,189) = 23.065, p < 0.0001, f = 0.349). Array size was not significant (F(6,189) = 0.767 p = 0.596, f = 0.156), and there was no interaction between array size and diagnosis (F(6,189) = 1.241 p = 0.287 f = 0.198). Post hoc analysis revealed bvFTD participants displayed significantly worse accuracy (bvFTD: 0.983, control: 0.997).

When examining the conjunction task, there was a significant effect of diagnosis (F(1,189) = 20.314, p < 0.0001, f = 0.328) and array size (F(6,189) = 2.526, p = 0.0224, f = 0.283). No significant interaction was observed between diagnosis and array size (F(6,189) = 0.119, p = 0.994 f = 0.062). With post hoc examination, bvFTD participants displayed significantly worse accuracy for conjunction search (bvFTD: 0.925, control: 0.966). Array size was significantly different only between 10×10 and 4×4 array sizes (p = 0.035).

For the orientation task, ANOVA revealed a significant main effect of diagnosis (F(1,168) = 9.706, p = 0.002, f = 0.240), but not of array size (F(6,168) = 1.641, p = 0.139, f = 0.242) and no significant interaction between diagnosis and array size (F(6,168) = 1.393, p = 0.220, f = 0.223). Post hoc examination revealed bvFTD participants displayed significantly worse accuracy for orientation (bvFTD: 0.899 control: 0.955).

Response time

Two way ANOVA examining diagnosis and array size revealed statistically longer response times in the bvFTD group for color (bvFTD: 2505.8 ms control: 959.3 ms F(1,189) = 150.873, p < 0.0001, f = 0.894), orientation (bvFTD: 14213.3 ms, control: 8989.7 ms F(1,168) = 16.461, p < 0.0001, f = 0.313), and conjunction searches (bvFTD: 9976.5 ms, control: 4590.1 ms F(1,189) = 63.331 9, p < 0.0001, f = 0.579). Array size was not statistically significant for color searches (F(6,189) = 0.328, p = 0.922 f = 0.101); however, it did have a significant effect for orientation (F(6,168) = 6.631, p < 0.0001, f = 0.486) and conjunction (F(6,189) = 6.717, p < 0.0001, f = 0.462) searches. Post hoc examination revealed significant differences between 10×10 versus 4×4, 5×5, and 6×6 as well as between 9×9 versus 4×4 and 5×5 arrays for both conjunction and orientation. Significant differences were also seen for 9×9 versus 6×6 as well as 8×8 versus 4×4 arrays for orientation. There was no significant interaction between array size and diagnosis for color (F(6,189) = 0.139, p = 0.991, f = 0.063), orientation (F(6,168) = 0.284, p = 0.944, f = 0.101), or conjunction (F(6,189) = 0.631, p = 0.705, f = 0.143).

Examination of the response time slopes revealed no significant difference between bvFTD and control participants for color (W = 146, p = 0.053, r = 0.362), orientation (W = 68, p = 0.672, r = 0.090), or conjunction (W = 114, p = 0.616, r = 0.099). For the color task, the average slope was not significantly different from zero, while a positive slope value was found for both orientation and conjunction tasks (Fig. 2).

Fig.2

Response time regression line slopes.

Intercepts from the linear regression data were significantly higher for the bvFTD participants for color (median bvFTD = 1757.602, control = 938.556 W = 201, p < 0.0001, r = 0.814), orientation (median bvFTD = 5508.822, control = 2670.275 W = 134, p = 0.001, r = 0.608), and conjunction (median bvFTD = 4585.823, control = 1639.75 W = 202, p < 0.0001, r = 0.822) (Fig. 3). There was no significant correlation between the NUCOG and slope and intercept values for any of the tasks.

Fig.3

Response time regression line intercepts.

Eye movement results

There was a significant increase in the number of objects examined by bvFTD participants relative to controls for color (F(1,189) = 33.820, p < 0.0001, f = 0.423), conjunction (F(1,189) = 4.505, p = 0.035, f = 0.153), but not orientation (F(1,168) = 2.096, p = 0.150 f = 0.110) searches. Array size had no significant effect on the number of objects examined for color (F(6,189) = 1.523, p = 0.173, f = 0.220); however, the number of objects examined did show an increase with increasing array size for orientation (F(6,168) = 12.874, p < 0.0001, f = 0.678) and conjunction (F(6,189) = 14.122, p < 0.0001, f = 0.670).

Examination of the number of fixations made before making a decision revealed a significant difference in the number of fixations made by bvFTD versus control participants for color (F(1,189) = 56.616, p < 0.0001, f = 0.548) and conjunction (F(1,189) = 21.312, p < 0.0001, f = 0.335), but not orientation (F(1,168) = 1.873, p = 0.173, f = 0.106) searches. Array size did not produce a significant difference in the number of fixations for color (F(6,189) = 0.488, p = 0.817, f = 0.123) but did result in an increase in the number of fixations with increasing array size for orientation (F(6,168) = 5.466, p < 0.0001, f = 0.441) and conjunction F(6,195) = 5.883, p < 0.0001, f = 0.432) tasks.

Fixation duration was significantly different for diagnosis for both orientation (F(1,168) = 23.825, p < 0.0001, f = 0.376) and conjunction (F(1,189) = 12.775, p < 0.0001 f = 0.259), but not the color task (F(1,189) = 0.717, p = 0.398, f = 0.063). There was no significant difference in fixation duration for array size in any of the three tasks, and no significant interaction between array size and diagnosis for any of the eye movement parameters (p > 0.05).

DISCUSSION

This study aimed to examine task performance and fixation patterns during visual search in bvFTD across a range of larger array sizes in both feature and conjunction search tasks. The color task was successful as a pop-out feature search, with search slopes not significantly different to zero, while both the orientation feature task and the conjunction task elicited serial searches with slopes significantly greater than zero, making these tasks adequate probes into participants’ serial search ability but with different levels of difficulty. Feature searches not producing a pop-out effect have been reported in the literature, and represent a more difficult feature search task [32, 33].

Accuracy was significantly reduced in bvFTD participants across all three tasks. This contrasts to Viskontas et al., who found reduced accuracy only on the conjunction task. This may be due to our increased array sizes producing increased cognitive difficulty or to differences in the degree of frontal atrophy between our cohorts, as visual search errors have been associated with frontal eye field integrity [34]. The preferred response by both bvFTD and control participants appears to be responding that a target is not present, despite targets actually being present in fifty percent of trials. Accuracy was not significantly correlated with NUCOG values or cognitive domain scores for any of the target conditions, indicating that the visual search changes may reflect a different skill set to those measured by the NUCOG. Array size was not a significant factor for accuracy in the feature search tasks but was significant for the conjunction task; this may reflect the additional integration required to detect a target defined by two separate properties. As array sizes increase, an increase in mistakes can result for tasks where the distractor is similar to the target, as the signal is weaker [35]. Consequently, it is reasonable to expect an increase in mistakes from both groups as the difficulty caused by increasing distractors numbers increases.

Response time, linear regression slope, and linear regression intercept values were selected for inclusion in this study as global measures of efficiency. BvFTD patients have shown increased latencies in decision making across a range of tasks [18]. Consistently seen across all three search tasks, and in agreement with Viskontas et al.’s data, was an increased response time by bvFTD participants, regardless of search task. The increase in response time was not due to increasing difficulty from distractors in the task, as evidenced by there being no significant difference in the slope values seen in the bvFTD participant group but was instead due to the presence of an increase in the intercept values for response time.

Increased intercept values (reflecting increased response time) are thought to be due to generalized cognitive slowing and have been associated with frontal lobe changes in Parkinson’s disease [25]. That is, the task is understood and is executed correctly but more slowly. When the task is executed not only more slowly but less efficiently as the number of distractors increases, an increase in the slope of the response time regression is seen. Increased slopes of response time regressions have been associated with increased processing difficulty during a task as the number of distractors increases, and have been linked with parietal lobe changes [19]. An increased slope value was not seen with bvFTD participants, indicating that their execution of the task as the number of distractors increases is not affected by the degeneration. In contrast to this, their regression intercepts were consistently significantly increased relative to controls. Reaction time regression intercept values have been linked with baseline difficulty in visual search, and correlated with frontal lobe activation [6]. This is consistent with the cortical changes seen in bvFTD and our results display bvFTD visual search performance similar to that of Parkinson’s patients with frontal changes [25]. Visual search timing from bvFTD participants, while increased overall, appears to be different to the pattern of increased response timing seen in AD patients, who display not only an increased regression intercept for feature search but also both an increased intercept and an increased slope for conjunction tasks [19, 20].

Examination of the eye movements made during visual search in bvFTD patients versus controls revealed a small increase in the number of objects examined and of fixations made before reaching a decision. This was more prominent in smaller array sizes, suggesting that the increase is due to the eye movements made during the delay before making a decision. An increase in fixation duration was also seen for the more difficult tasks, suggesting a link with cognitive slowing. Our results suggest that the difficulty is in responding to the task rather than being hindered by the number of distractors. It would be of interest to see how bvFTD participants perform on time-limited search tasks to separate the search process itself from the increased latency of responding.

This study has further characterized bvFTD behavior in visual search tasks. BvFTD patients present with a characteristic visual search behavior: reduced accuracy, increased number of objects fixated, increased number of fixations made, increased fixation duration, and increased response time due to an increase in the response regression intercept. The pattern of deficits seen in bvFTD is different to that reported in Alzheimer’s disease, but closer to that produced by Parkinson’s patients with frontal lobe changes [25]; bvFTD patients display reduced accuracy and increased response time similar to AD, but this is not due to increased processing time for individual distractors. The pattern is also different to schizophrenia patients who do not display a significant difference to controls for accuracy [23], and to OCD patients, who do not display an increased latency [24]. Characterization of visual search tasks in different neurodegenerative conditions is useful as it produces objective measures of behaviors which underlie a daily-used skill, and can indicate where in the processing pathways deficits may be occurring.

Limitations

The above results indicate some significant and easily elicited differences in both behavior and eye movements between bvFTD and control participants; however, there are some limitations to the study methodology. The Tobii remote view tracker was selected for ease of use with dementia participants as it allows a degree of free head movement while still tracking gaze. Unfortunately, in our participants this resulted in periods of time during the tasks where participants were no longer attending to the task, and instead were seeking validation from the researcher for their behavior, or were gazing at other objects around them. The resultant head movement resulted in varying visual angles for different participants and the distractions may have contributed to the increased response times seen. While we have attributed the above increase in response time to cognitive slowing, we cannot exclude the increase being due to reduced attention, but do not suspect this to be the case as the increases in response time are consistent across trial array sizes, while distraction from the task was prominent on more difficult tasks. Color visual search tasks were performed first with every participant, as this is the simplest task to comprehend since the target is highly salient. The color task was clearly the quickest and likely to display the highest attention levels as it was performed first, yet this task still displays the increased intercept values seen across all the search tasks.

Our response time values may have been more accurate with the aid of a button box synchronized to the Tobii tracker, and the researchers did press the button for a number of patient participants, which has added some noise to the response time data. However, we believe the difference seen between the bvFTD and control group is significantly larger than the noise added from this methodology, which was included to ensure mental spatial effort was directed at the task. We do not, however, believe that altering this would have changed the above results as the magnitude of the difference between bvFTD and control participants is in the order of seconds, and is consistent with the result expected for frontal atrophy changes.

The search tasks used in our study are not identical to those used when examining other dementia types. We do not believe this has altered the results but recognize that this work would benefit from a future study applying a standardized test to bvFTD participants and across a range of different dementia groups.

Molecular and postmortem diagnosis was not available in our bvFTD cohort and thus we were unable to determine if the results seen were due to a particular mutation or molecular aggregate and cannot exclude the possibility of cross-contamination of the results from another pathology. This problem applies to other dementing illnesses where accurate antemortem diagnostics are lacking.

Conclusions

BvFTD participants display both behavioral and eye movement changes in a range of visual search tasks. Behaviorally, bvFTD participants displayed a significant increase in response time which, when examined across array sizes, mimicked the pattern of visual search seen in Parkinson’s disease patients with frontal lobe changes. Visual search is a tool that can be objectively quantified, and this study reports on the specific pattern of behavior seen in visual search in bvFTD with larger array sizes. It may be contrasted to search task behavior in other conditions such as Alzheimer’s disease, schizophrenia, and OCD, with which it is often confused diagnostically. Further work is required to produce a standardized visual search task with which to examine a range of neurocognitive conditions. On its own, search is unlikely to completely differentiate between different conditions, but may be useful as part of a larger battery which may be used to build evidence in the diagnostic process.

Footnotes

ACKNOWLEDGMENTS

This research was supported by an Australian Postgraduate Award.

Authors’ disclosures available online ().

References

Ramzaoui

, Faure

, Spotorno

(2018) Alzheimer’s disease, visual search, and instrumental activities of daily living: A review and a new perspective on attention and eye movements. J Alzheimers Dis 66, 901–925.

Treisman

, Gelade

(1980) A feature-integration theory of attention. Cogn Psychol 12, 97–136.

Wolfe

, Cave

, Franzel

(1989) Guided search: An alternative to the feature integration model for visual search. J Exp Psychol Hum Percept Perform 15, 419–433.

Wolfe

(1994) Guided Search 2.0 A revised model of visual search. Psychon Bull Rev 1, 202–238.

Maximo

, Neupane

, Saxena

, Joseph

, Kana

(2016) Task-dependent changes in frontal-parietal activation and connectivity during visual search. Brain Connect 6, 335–344.

Nobre

, Coull

, Walsh

, Frith

(2003) Brain activations during visual search: Contributions of search efficiency versus feature binding. Neuroimage 18, 91–103.

Greenlee

, Berg

, Stuhr

, Mergner

(2000) Visual search and visual working memory in patients with chronic focal cortical lesions. Vision Res 40, 3759–3773.

Binello

, Mannan

, Ruddock

(1995) The characteristics of eye movements made during visual search with multi-element stimuli. Spat Vis 9, 343–362.

Zelinsky

, Sheinberg

(1997) Eye movements during parallel-serial visual search. J Exp Psychol Hum Percept Perform 23, 244–262.

10.

Muggleton

, Juan

, Cowey

, Walsh

(2003) Human frontal eye fields and visual search. J Neurophysiol 89, 3340–3343.

11.

Rascovsky

, Hodges

, Knopman

, Mendez

, Kramer

, Neuhaus

, van Swieten

, Seelaar

, Dopper

, Onyike

, Hillis

, Josephs

, Boeve

, Kertesz

, Seeley

, Rankin

, Johnson

, Gorno-Tempini

, Rosen

, Prioleau-Latham

, Lee

, Kipps

, Lillo

, Piguet

, Rohrer

, Rossor

, Warren

, Fox

, Galasko

, Salmon

, Black

, Mesulam

, Weintraub

, Dickerson

, Diehl-Schmid

, Pasquier

, Deramecourt

, Lebert

, Pijnenburg

, Chow

, Manes

, Grafman

, Cappa

, Freedman

, Grossman

, Miller

(2011) Sensitivity of revised diagnostic criteria for the behavioural variant of frontotemporal dementia. Brain 134, 2456–2477.

12.

Lillo

, Savage

, Mioshi

, Kiernan

, Hodges

(2012) Amyotrophic lateral sclerosis and frontotemporal dementia: A behavioural and cognitive continuum. Amyotroph Lat Scler 13, 102–109.

13.

Siuda

, Fujioka

, Wszolek

(2014) Parkinsonian syndrome in familial frontotemporal dementia. Parkinsonism Relat Disord 20, 957–964.

14.

Rosen

, Gorno-Tempini

, Goldman

, Perry

, Schuff

, Weiner

, Feiwell

, Kramer

, Miller

(2002) Patterns of brain atrophy in frontotemporal dementia and semantic dementia. Neurology 58, 198–208.

15.

Salmon

, Garraux

, Delbeuck

, Collette

, Kalbe

, Zuendorf

, Perani

, Fazio

, Herholz

(2003) Predominant ventromedial frontopolar metabolic impairment in frontotemporal dementia. Neuroimage 20, 435–440.

16.

Woolley

, Khan

, Murthy

, Miller

, Rankin

(2011) The diagnostic challenge of psychiatric symptoms in neurodegenerative disease: Rates of and risk factors for prior psychiatric diagnosis in patients with early neurodegenerative disease. J Clin Psychiatry 72, 126–133.

17.

Huey

, Goveia

, Paviol

, Pardini

, Krueger

, Zamboni

, Tierney

, Wassermann

, Grafman

(2009) Executive dysfunction in frontotemporal dementia and corticobasal syndrome. Neurology 72, 453–459.

18.

Rahman

, Sahakian

, Hodges

, Rogers

, Robbins

(1999) Specific cognitive deficits in mild frontal variant frontotemporal dementia. Brain 122, 1469–1493.

19.

Landy

, Salmon

, Filoteo

, Heindel

, Galasko

, Hamilton

(2015) Visual search in dementia with Lewy Bodies and Alzheimer’s disease. Cortex 73, 228–239.

20.

Foster

, Behrmann

, Stuss

(1999) Visual attention deficits in Alzheimer’s disease: Simple versus conjoined feature search. Neuropsychology 13, 223–245.

21.

Moser

, Kompf

, Olschinka

(1995) Eye movement dysfunction in dementia of the Alzheimer type. Dementia 6, 264–268.

22.

Rösler

, Mapstone

, Hays

, Mesulam

, Rademaker

, Gitelman

, Weintraub

(2000) Alterations of visual search strategy in Alzheimer’s disease and aging. Neuropsychology 14, 398–408.

23.

Gold

, Fuller

, Robinson

, Braun

, Luck

(2007) Impaired top-down control of visual search in schizophrenia. Schizophr Res 94, 148–155.

24.

Kaplan

, Dar

, Rosenthal

, Hermesh

, Fux

, Lubow

(2006) Obsessive-compulsive disorder patients display enhanced latent inhibition on a visual search task. Behav Res Ther 44, 1137–1145.

25.

Berry

, Nicolson

, Foster

, Behrmann

, Sagar

(1999) Slowing of reaction time in Parkinson’s disease: The involvement of the frontal lobes. Neuropsychologia 37, 787–795.

26.

Viskontas

, Boxer

, Fesenko

, Matlin

, Heuer

, Mirsky

, Miller

(2011) Visual search patterns in semantic dementia show paradoxical facilitation of binding processes. Neuropsychologia 49, 468–478.

27.

Douglass

, Walterfang

, Velakoulis

, Abel

(2018) Behavioral variant frontotemporal dementia performance on a range of saccadic tasks. J Alzheimers Dis 65, 231–242.

28.

Walterfang

, Siu

, Velakoulis

(2006) The NUCOG: Validity and reliability of a brief cognitive screening tool in neuropsychiatric patients. Aust N Z J Psychiatry 40, 995–1002.

29.

Pollack

, Norman

(1964) Non-parametric analysis of recognition experiments. Psychon Sci 1, 125–126.

30.

Snodgrass

, Corwin

(1988) Pragmatics of measuring recognition memory: Applications to dementia and amnesia. J Exp Psychol Gen 117, 34–50.

31.

Miller

, Walterfang

, Randhawa

, Scholes

, Mocellin

, Velakoulis

(2007) Validity and reliability of the Behavioural Assessment Tool for Cognition and Higher Function (BATCH) in neuropsychiatric patients. Aust N Z J Psychiatry 41, 697–704.

32.

Verghese

, Nakayama

(1994) Stimulus discriminability in visual search. Vision Res 34, 2453–2467.

33.

Wolfe

, Friedman-Hill

, Stewart

, O’Connell

(1992) The role of categorization in visual search for orientation. J Exp Psychol Hum Percept Perform 18, 34–49.

34.

Heitz

, Cohen

, Woodman

, Schall

(2010) Neural correlates of correct and errant attentional selection revealed through N2pc and frontal eye field activity. J Neurophysiol 104, 2433–2441.

35.

Santhi

, Reeves

(2004) The roles of distractor noise and target certainty in search: A signal detection model. Vision Res 44, 1235–1256.