Here, we present the first quantitative assessment of smooth pursuit performance in individuals with central visual field loss. We used a novel SLO approach that allowed us to look at target placement directly on the retina during smooth pursuit, thereby allowing us both to describe smooth pursuit characteristics in our patient population and to determine the retinal area used to follow the target in the presence of central field loss.
The eleven individuals with macular degeneration we tested could smoothly pursue a slowly moving (5°/s–6°/s) target. Interestingly, performance did not depend on overall scotoma size, even though in several patients, scotoma size exceeded 20° of visual angle. Performance on the smooth pursuit task, in terms of gain and latency, was significantly worse in patients than in age-matched controls—a result consistent with previous findings in healthy participants with a small (≤6°), dense, artificial scotoma (Pidcoe & Wetzel,
2006) and those pursuing perifoveal targets (Winterson & Steinman,
1978). The present study is an important extension of the work by Pidcoe and Wetzel, which focused on the formation and use of PRLs in young (ages 28–37), healthy controls. In that study, the participants were presented with periodic (sinusoidal) and nonperiodic (sum of sines) targets moving between 4°/s and 16°/s. The authors did not compute smooth pursuit metrics, such as latency and gain, but instead focused on the offset of eye position to determine PRL formation. Qualitatively, the authors reported poorer pursuit in the nonperiodic condition. The across-patient variability of pursuit in our study is in agreement with previous qualitative observations of Schuchard, Naseer, and de Castro (
1999). Patients in this study reported being able to visualize the target for at least part of the trial, and being able to pursue it, which is consistent with our data.
Some of our patients were able to follow the target even when it fell into their scotoma region for at least part of the trial (
Figure 7, end part of the trial). Given the size of the target (1.7°) and variation in the scotoma boundary, we cannot exclude the possibility that a part of the target continued to stimulate healthy retina, facilitating pursuit in these patients. Overall, there was a significant effect of scotoma location relative to the target direction for both pursuit initiation (latency) and pursuit maintenance (steady state gain).
Several studies have shown that the PRL measured during fixation is not always used as an oculomotor reference. For example, White and Bedell (
1990) showed that only observers with long-standing central field loss successfully rereferenced saccades to their fixational PRL. Those with more recent macular degeneration used a mixture of the original fovea and the PRL as an oculomotor reference. Furthermore, although monkeys with induced retinal lesions formed a fixational PRL within a day of the lesions, it took much longer for them to use the PRL for saccades (Heinen & Skavenski,
1992). Thus, it was not clear whether observers with CFL would use their PRLs for smooth pursuit. Our data indicate that patients placed the target on their fixational PRLs about 42% of the time analyzed. We saw no correlation between disease duration and use of a fixational PRL; patients with some of the longest disease durations had some of the fewest trials with the target primarily within the PRL region. Perhaps the discrepancy between our pursuit data and previous saccade data is that pursuit can be accomplished successfully in the periphery (Winterson & Steinman,
1978). Furthermore, the size of our target, and its annular shape, may have allowed the patients to follow a peripheral portion of the target. Additionally, the resolution of our analysis method may be limited by some jitter (<0.5°) that would place the target outside the fixation area. However, for both patients and controls, we saw trials that were well outside of the range of the extent of the target and analysis error. Interestingly, we saw evidence of PRL switching in two of our observers. However, in both cases one of the two PRLs used was the fixational PRL (e.g.,
Figure 7A).
PRL awareness and placement can be important considerations for successful object viewing in patients with AMD (Verghese & Janseen,
2015). Therefore, understanding the effects of PRL (and thus scotoma) placement, relative to target motion, on pursuit performance is important and can lead to potential successful training paradigms, as shown previously for other tasks (Janssen & Verghese,
2015; Seiple, Szlyk, McMahon, Pulido, & Fishman,
2005). To that end, we analyzed target placements and scotoma-PRL configuration in two ways. For the dynamic analysis, we analyzed the position of the target on the moving eye, analyzing performance relative to the health of the retina traversed by the target. For the static analysis, we looked at pursuit performance as a function of target direction relative to the scotoma, taking into account heading and starting directions. Both analyses need to be considered, as it is important to note that scotomas can vary in shape across individuals, and even between eyes within an individual. In some patients, the PRL is located along the edge of a scotoma, whereas in others it can be entirely or partially surrounded by a scotoma. Therefore, simply looking at retinal health in “heading” and “starting” directions individually is insufficient to understand the complex interplay of target motion, target location, and PRL location relative to the target and the scotoma. Conversely, to look exclusively at retinal health at locations corresponding to the moving target does not take into account patients' experience with their scotoma.
For the dynamic analysis we considered the actual retina traversed by the target. We did so by looking at both the trial as a whole (broken up into three epochs), and the portion of the trial for which gain was computed. We were surprised to find that the health of starting and traversed retina did not have an effect on the latency or gain. Furthermore, whether the target started in healthy retina did not predict whether the patient could maintain the target on healthy retina for the majority of the trial. This outcome is surprising and may require further investigation. In our experiment, we probed only the absolute scotoma, as our flash stimuli were well above threshold in brightness and were therefore easy to detect. The measure of retinal health in a given location and its relationship to pursuit quality may be more nuanced, and the region of relative scotoma may have to be taken into account for future investigations.
For the static analysis, we looked at the heading direction of the target relative to the scotoma. For this analysis we considered two parts of the trajectory, the starting and heading ROI, relative to the PRL. Our initial hypothesis was that patients would perform most poorly when the target motion initiated in the scotoma, i.e., when the starting ROI was in unhealthy retina (
Figure 8B). We anticipated low gain and an increase in latency due to the initial disappearance of the target in the scotoma. Our results show that a high proportion of scotoma in the starting ROI increased latency. However, pursuit gain was not affected by the starting ROI (
Table 2). Interestingly, both gain and latency were adversely affected by the presence of scotoma in the heading direction (
Figure 8A), suggesting several possible interpretations. One possible explanation is that in a situation where pursuit gain is less than one, and therefore the PRL falls behind the target; the target will approach and eventually be occluded by the scotoma (
Figure 8A). If the participant does not reacquire the target, the gains should diminish towards the end of these trials (as is the case in
Figure 7A,
E). Previous research suggests that when a smooth pursuit target is occluded, smooth pursuit eye velocity starts to decelerate approximately 190 ms after target disappearance. The eyes continue to move at a much lower “residual velocity” that develops after as little as 300 ms of target presentation and persists for an extended period of time (<4 s) (Becker & Fuchs,
1985; Bennett & Barnes,
2006). However, we did not consistently observe a decrease in gain towards the end of the trial. This discrepancy could be due to participants refixating the target several times throughout the trial; however, we find this possibility unlikely, as the majority of the participants did not exhibit searching behavior, such as saccades in different directions, at later stages of the trial. Interestingly, the best strategy in this target-PRL configuration might be to maintain a gain of 1 or slightly above—values that are present in some participants for certain directions, but seldom when the target is heading into the scotoma.
The health of the retina may also have been an important factor. When the target is moving toward the scotoma (
Figure 8A), it is moving into progressively less healthy retina. The longer latencies suggest that the target could be very close to the scotoma by the time pursuit begins. Conversely, when the target is moving into healthy retina (
Figure 8B), as the eye lags, the target moves away from the scotoma, which may be especially important as the PRL is frequently close to the scotoma border (Fletcher & Schuchard,
1997). Our results indicate that this interpretation on its own is unlikely, as retinal health at the location of the target did not directly predict changes in gain, indicating that an investigation that includes relative and dense scotoma boundaries (Ergun et al.,
2003) in the analysis is warranted. Finally, our finding may be related to the physiological constraints on patients' eye movements, with movement into the scotoma requiring a more eccentric eye position than is comfortably possible. However, several of our participants had close-to-central PRLs and nonetheless exhibited gain variations based on scotoma location.
For latency, one would easily expect an increase when the target steps into the scotoma at trial onset, as the patient would require a longer time to find the target. However, the influence of heading direction on latency is more surprising. A possible explanation is a change in the patients' strategy: While the target is visible, the patients may not be motivated to immediately move their eyes to start tracking it, as that may introduce additional position error. Patients may even delay making the initial saccade toward the target since vision is suppressed during saccades, reducing the amount of time patients have to visualize the target. As the target approaches the scotoma, patients would then initiate pursuit, albeit with less intact retina and lower gains. An alternative explanation is a measurement artifact: Because gains were lower in these trials, latencies were more difficult to detect precisely. We believe this explanation unlikely, since we saw this phenomenon across patients who had differing ranges of gain modulation, i.e., lowest gains of one patient were comparable to some of the higher gains for another, but they all exhibited this trend in latency. Interestingly, we also found a negative interaction between latency and the extent of scotoma in both ROIs. This interaction indicates that as the amount of scotoma in both regions increases, pursuit latency decreases. One potential explanation is that participants are more likely to move their eyes quickly when the target is likely to disappear into their scotoma, soon after it becomes visible. This behavior suggests that prior experience with their scotoma affects pursuit strategy.
Smooth pursuit can be an important tool for individuals with vision loss, as it can afford additional time for target visualization, analysis, and identification, as in the case of trying to identify the number of a passing bus. The current study is the first to examine smooth pursuit in individuals with macular degeneration, and it provides key first steps to understanding this behavior. However, because of the limitations of the SLO, the study was performed monocularly using above-threshold viewing conditions. To extend our findings to smooth pursuit in the real world for persons with central field loss, we need a greater range of target velocites under binocular viewing. Therefore, future investigation of smooth pursuit under more natural viewing conditions and at a variety of target speeds can provide important information about successful strategies that individuals with central visual field loss can employ. These eye tracker-based studies can be related to the monocular measurements presented here to frame the behavior in the context of each patient's individual pursuit locus.