Rapid Automatized Picture Naming in an Outpatient Concussion Center: Quantitative Eye Movements during the Mobile Universal Lexicon Evaluation System (MULES) Test

Abstract

Number and picture rapid automatized naming (RAN) tests are useful sideline diagnostic tools. The main outcome measure of these RAN tests is the completion time, which is prolonged with a concussion, yet yields no information about eye movement behavior. We investigated eye movements during a digitized Mobile Universal Lexicon Evaluation System (MULES) test of rapid picture naming. A total of 23 participants with a history of concussion and 50 control participants performed MULES testing with simultaneous eye tracking. The test times were longer in participants with a concussion (32.4 s [95% CI 30.4, 35.8] vs. 26.9 s [95% CI 25.9, 28.0], $t=6.1$). The participants with a concussion made more saccades per picture than the controls (3.6 [95% CI 3.3, 4.1] vs. 2.7 [95% CI 2.5, 3.0]), and this increase was correlated with longer MULES times (r = 0.46, p = 0.026). The inter-saccadic intervals (ISI) did not differ between the groups, nor did they correlate with the test times. Following a concussion, eye movement behavior differs during number versus picture RAN performance. Prior studies have shown that ISI prolongation is the key finding for a number-based RAN test, whereas this study shows a primary finding of an increased saccade number per picture with a picture-based RAN test. Number-based and picture-based RAN tests may be complimentary in concussion detection, as they may detect different injury effects or compensatory strategies.

1. Introduction

Concussion is a common form of mild traumatic brain injury for which sensitive, reliable, and objective diagnostic tests are important in areas such as athletic sidelines and military and clinical settings. Tools that are routinely implemented on the athletic sidelines, which are also increasingly utilized in outpatient clinical evaluations, include symptom checklists, the Standardized Assessment of Concussion (SAC), the Balance Error Scoring System (BESS), and the Vestibulo-Ocular Motor Screening (VOMS) test [1,2,3,4,5,6,7,8].

Rapid automatized naming (RAN) tests, utilized in neuropsychological studies since the early 20th century, are performance measures that consist of timed and scored letter, number, color, or picture/object identification [9,10,11,12,13,14]. As visually-based performance measures, RAN tests not only assess vision and eye movement behavior, but they also assess attention, cognition, and language. The results are impacted by dysfunction in any of these domains, as well as by factors such as fatigue, cooperation, and motivation. In concussion, the King–Devick (KD) test, a rapid number naming test, has been validated as a sensitive sideline measure for acute concussion detection. Post-concussion total test times increase (i.e., worsen) relative to pre-season baseline scores [15,16,17,18,19]. More recently, the Mobile Universal Lexicon Evaluation System (MULES), a rapid picture naming test, has been developed [20], studied in multiple neurologic disorders [21,22,23,24], and also shown to yield prolonged test times in individuals with a concussion [21,22]. The completion of the MULES test requires the naming of 54 color photographs of fruits, common objects, and animals in context (Figure 1) as rapidly and accurately as possible. The addition of vision-based rapid automatized naming (RAN) tests may improve the sensitivity of concussion detection [25].

Although standard sideline or clinical applications of RAN tests provide an objective method for concussion screening, the results are restricted to the total test times and error rates that preclude the study of the behavioral underpinnings of prolonged test times seen in concussions. The performance of RAN tests with simultaneous quantified eye movement recordings (i.e., eye tracking) provides a framework for the study of ocular motor behavioral changes [26,27,28]. The objective of this study was to perform a detailed temporal, spatial, and kinematic analysis of the saccades and the periods of fixation between the saccades during MULES performance in participants with a history of concussion.

2. Materials and Methods

2.1. Participants and Data

This study was IRB-approved and written informed consent was obtained from all participants. A convenience sample of 23 participants with a history of concussion was recruited from the institution’s outpatient concussion center. Participants were not invited based on particular clinical features or duration of concussion, though all reported persistent symptoms since the time of concussion. Exclusion criteria included co-existing ophthalmic disease unrelated to concussion (with the exception of corrected refractive error) that may have prevented clear visualization of MULES pictures. A control cohort of 50 participants with no history of traumatic brain injury, neurologic disease, visual impairment (other than corrected refractive error), or ocular motor abnormalities was recruited from institutional faculty and staff.

All participants completed a computerized version of the MULES test while simultaneously undergoing binocular eye movement recordings. The MULES test consisted of three screens (screen dimensions: width 53 cm × height 30 cm; distance between screen and participant: 75 cm) with a series of 18 pictures per screen arranged in a 3 × 6 grid for a total of 54 pictures to be named (Figure 2, top). The MULES test was digitized and computer-generated via custom MATLAB software to display the same pictures that comprise the ‘paper and pencil’ version of the MULES test, with the same ordering and relative spacing on each line. The images were separated into three sets instead of the two sets displayed on the laminated flip cards of the ‘paper and pencil’ version of the MULES test [21] to maximize image sizes on the computer monitor. Each screen of images was preceded by a central fixation circle, which they were required to fixate on for 1 s before the screen with the picture images was displayed to standardize the starting position of the eye across individuals and ensure that their attention was directed to the screen. Participants were instructed to name each picture as rapidly and accurately as possible starting at the upper left image and naming the images from left to right and top to bottom. The total digitized MULES time in seconds elapsed while naming the pictures (excluding time between screen changes) was recorded.

Video-oculographic recordings were binocularly obtained (Eyelink 1000+, SR Research, Ottawa, ON, Canada) at a sampling rate of 500 Hz and with 0.01° RMS precision. Participants’ heads were stabilized against a forehead rest, allowing for unimpeded speech throughout the session. Participants successfully completed a 13-point spatial calibration procedure prior to each session. Eye position was recorded continuously during presentation of all MULES pictures.

All analyses were performed offline using custom MATLAB code. Saccades were identified via an adaptive thresholding mechanism and velocities and accelerations were computed from position traces using a low-pass differentiator [29]. Inter-saccadic intervals (ISI) were extracted for further analysis.

2.2. Statistical Analysis

Statistical analysis was completed in MATLAB (R2020b). Differences between total MULES naming times and eye movement metrics were assessed via paired t-tests. Linear modeling was used to assess correlations among variables, and logistic (log-linear) models were used to assess the separability of patient and control data (the signal-to-noise ratio of detecting patients relative to controls; $d{}^{\prime }$) based on digitized MULES data [30,31]. Concussed versus control status with regard to MULES times and age was assessed by logistic regression. Temporal variables, such as MULES test time and ISI, are known to be right-skewed and these data were inverse-transformed prior to analysis. No results would be changed by substituting non-parametric tests. The alpha criterion for statistical significance was set at α = 0.05.

3. Results

3.1. Demographics

A total of 23 participants with a history of concussion (age median 34; age range 15–65; 52% female) met the eligibility criteria and were included. A total of 50 healthy adult volunteers (age median 22; age range 18–44; 46% female) were included in the control population. In the concussed cohort, the time from concussion to testing ranged from less than one month to 18 months (median duration 1.8 months). All participants with a concussion remained symptomatic from their concussion and continued to be actively managed medically in an outpatient concussion center. A logistic regression predicting concussed versus control participant status showed that the MULES times (p = 0.006), but not age (p = 0.11), were significantly predictive.

3.2. MULES and Eye Movement Metrics

Total MULES naming times for individuals with a history of concussion were longer than in the healthy controls (32.4 s [95% CI 30.4, 35.8] vs. 26.9 s [95% CI 25.9, 28.0], $t=6.1$). The participants with a history of concussion produced a greater number of saccades per picture compared to the controls (3.6 [95% CI 3.3, 4.1] vs. 2.7 [95% CI 2.5, 3.0], $t=6.0$) (Figure 2 and Figure 3). Consistent with making a larger number of saccades, individuals with a history of concussion also made smaller saccades (7.1 deg [95% CI 6.8, 7.5] vs. 8.4 deg [95% CI 8.1, 8.6], $t=2.2$). There were no significant differences in the inter-saccadic intervals (ISI) between the two groups (195 msec [95% CI 192, 198] vs. 199 msec [95% CI 194, 204] in the control and concussion history groups, respectively). In addition, no group differences were identified for the number of ‘lookback’ saccades (i.e., saccades in the wrong direction for the task’s progression) or in the saccade peak velocity versus amplitude main sequence relationships (i.e., saccades were not slow in the concussion group).

There was a significant correlation (Figure 4A) between longer (worse) MULES times and increased numbers of saccades in both groups (r = 0.46, p = 0.026 and 0.63, p = <0.00001 in control and concussion history, respectively). There was also a significant correlation (Figure 4B) between ISI and the number of saccades in both groups (r = −0.62, p = 0.0016; r = −0.46, p = 0.00078 for control and concussion groups, respectively). No significant correlation was observed between the MULES times and ISI in either group (Figure 4C).

Control and concussion history participant performance could be discriminated $d^{\prime }\ge 1.0$ (Figure 5) using both MULES times, $d{}^{\prime }=2.3$ [95% CI 1.6, 3.3], and eye movement data, $d{}^{\prime }=2.8$ [95% CI 2.1, 3.8].

4. Discussion

4.1. Current Results

Eye movement behavior during a digitized MULES picture naming test was assessed in individuals with a history of concussion and compared to a cohort of healthy controls. There was a long duration between the last concussion and MULES performance in several participants in our cohort. Nonetheless, the cohort was a representative sample from an outpatient clinic at a comprehensive interdisciplinary academic concussion center, and these individuals remained symptomatic after their most recent concussion. The total MULES test times were prolonged in the participants with a history of concussion compared to the controls; this is consistent with prior sideline studies of concussed athletes [21,22]. In our study, MULES performance was combined with quantified eye movement recordings to assess eye movement behavior during the test and to examine potential factors contributing to prolonged test times in concussions. The results revealed that increased numbers of saccades (but not prolonged ISI) during MULES picture naming were associated with longer (worse) MULES times. This outcome is different from that observed with rapid number naming tests, in which strong correlations existed between both prolonged ISI and an increased numbers of saccades with worse overall test times [26].

4.2. Contributions and Limitations

The major differences between picture naming and letter or number naming tasks are (1) the extended and complex nature of pictures relative to single letters and digits, and (2) the ‘set size’ of possible responses from which the participants can expect to draw correct responses in each task. It is likely that the first difference breaks the correlation normally seen between the ISI and total number-naming test times [32] because most pictures require between two or three saccades before they are identified, even among the controls. This suggests that the pictures require additional time or exploration before they can be identified, and that the inverse correlation between the ISI and the number of saccades executed per picture indicates a tradeoff between these two variables that tends to maintain the amount of time spent per image. Thus, whereas a shorter ISI during the number naming typically means that the numbers are named more quickly, the negative correlation with saccade quantity seen here suggests that a shorter ISI during the picture naming would increase exploration within the images by increasing the number of saccades overall. This increased number of saccades would also increase the proportion of time that the participants are functionally blind, as the eye changes position during each saccade. The complex nature of pictures, which requires the processing of figure—ground relations, color, context, shadow, etc., also requires more time to complete. We can see the result of this additional processing in the longer per-item response times during the picture- relative to the number-naming tasks. The second major difference, the set size of the potential responses in the picture-naming task, will also tend to add to this increase in the per-item naming time. This is consistent with evidence from previous work comparing MULES performance to rapid number naming, showing that per-item responses were over twice as long with MULES (0.72 s vs. 0.33 s, respectively) [23].

In addition, this set-size difference should create a larger ‘learning effect’ in picture naming relative to number- or letter-naming tests. A learning effect is measured as the reduction in the test times when the identical test is repeated. One of the major reasons repeat testing should be completed more quickly is because test items, and to some extent their ordering, will have been primed by the first test [33,34]. Before the first administration of a picture naming test, the set of images that our participants might expect to see is exceedingly large. After taking the test, this number is drastically reduced, and one would predict a corresponding drop in the total picture-naming time. In contrast, there are only ten numbers (zero through nine) that one could expect to identify in a number-naming task, and this is known before the first test; the ten numbers are in a sense ‘pre-primed’ when the instructions are given in the number-naming tests. Therefore, one would expect to see a much smaller learning effect in number- relative to picture-naming [35].

The differences outlined above between picture- and number/letter naming suggest differences in the brain pathways most heavily recruited while performing these tests. Both number/letter and picture naming rely on higher cortical control of saccadic eye movements, in concert with cognitive and attentional activity. RAN tasks probe the integrity of the cortical structures involved, such as the frontal eye fields (FEF) and the dorsolateral prefrontal cortex (DLPFC) [36]. Furthermore, number/letter and picture tasks require different cortical pathways for completion. For instance, color vision required for picture assessment is processed in the early visual cortex, as well as throughout the ventral pathway including V2, V4, and the hemifield maps in the ventral occipital regions VO1 and VO2 [37,38], whereas shapes are processed extensively in the lateral occipital complex (LOC) [32,39], including some overlap in the VO cluster [40,41]. Processing in the lateral occipital cortex (LOC) also contributes to repetition priming of objects, and repetition of the MULES could be used to provide evidence regarding potential priming deficits and their underlying causes following traumatic brain injury [42,43,44].

The presence of abnormal eye movements has been interpreted to indicate subnormal brain function in post-concussive states and as a proxy for cognitive or attentional deficits [12,14,25]. Objective data from eye movement recordings of the eye movement types that are highly dependent on cognition and attention (e.g., memory-guided saccades to the remembered location of a visual target) in concussion show increased saccadic latencies and saccadic errors. Errors in predictive target tracking and vergence eye movements are also commonly observed following a head injury [45,46,47,48,49,50,51,52,53,54]. Eye movement recordings have been combined with rapid number naming to better understand the mechanisms underlying prolonged test times following a concussion [36]. Prior studies have demonstrated that both the intersaccadic interval (ISI) (i.e., the time between saccades) and the number of saccades predict the overall KD test times [26,36]. Given that rapid picture naming is more challenging than number naming and that it recruits additional visual networks to processes color, shadow, texture, figure–ground separation, occlusion, etc., relative to number naming, the MULES may be an even more sensitive measure of a concussion [20]. Herein, we have demonstrated the differences in eye movement metrics between rapid number and picture naming, in an attempt to further explore the behavioral changes contributing to the prolonged test times seen in both tasks.

Limitations: The extended duration and variability in time since the concussive injury in our subjects is a potential limitation; thus, the population sample studied here may differ neurobiologically from populations with acute concussions, and the findings from this study may not be entirely generalizable to those in the earliest phases of recovery from concussion. Another limitation is that the age of the group with a history of concussions was slightly greater than the age of the control group; however, age was not a significant factor predicting concussion status. Future studies will be needed to assess RAN task performance across various age groups from childhood to early and later adulthood.

4.3. Conclusions and Future Directions

Prior studies have shown prolonged MULES test times following concussion [21,22]; eye movement behavior underlying these worsened times has not been formally examined. We examined the eye movements of participants with a history of concussion during MULES performance. Prolonged rapid picture naming times in MULES were significantly associated with an increased number of saccades, but not with prolonged ISI values. As such, a combination of MULES times and the number of saccades, but not ISI, was predictive of a concussion. This contrasts with prior studies examining rapid number naming in concussion, which showed significant changes in both the number of saccades and ISI [26]. These eye movement differences suggest that the brain may adopt different strategies when performing rapid picture- versus number-naming tasks following head injury. Understanding the deficits uncovered in rapid picture-naming tasks may ultimately assist in identifying the precise neuroanatomic substrates that are disrupted following concussion. In addition, a better understanding of the neurocognitive network underlying visual–verbal tasks may assist in optimizing outcome measures to track recovery and to design more tailored rehabilitation regimens in patients with concussion.