Oculomotor analysis to assess brain health: preliminary findings from a longitudinal study of multiple sclerosis using novel tablet-based eye-tracking software

Abstract

A growing body of evidence supports the link between eye movement anomalies and brain health. Indeed, the oculomotor system is composed of a diverse network of cortical and subcortical structures and circuits that are susceptible to a variety of degenerative processes. Here we show preliminary findings from the baseline measurements of an ongoing longitudinal cohort study in MS participants, designed to determine if disease and cognitive status can be estimated and tracked with high accuracy based on eye movement parameters alone. Using a novel gaze-tracking technology that can reliably and accurately track eye movements with good precision without the need for infrared cameras, using only an iPad Pro embedded camera, we show in this cross-sectional study that several eye movement parameters significantly correlated with clinical outcome measures of interest. Eye movement parameters were extracted from fixation, pro-saccade, anti-saccade, and smooth pursuit visual tasks, whereas the clinical outcome measures were the scores of several disease assessment tools and standard cognitive tests such as the Expanded Disability Status Scale (EDSS), Brief International Cognitive Assessment for MS (BICAMS), the Multiple Sclerosis Functional Composite (MSFC) and the Symbol Digit Modalities Test (SDMT). Furthermore, partial least squares regression analyses show that a small set of oculomotor parameters can explain up to 84% of the variance of the clinical outcome measures. Taken together, these findings not only replicate previously known associations between eye movement parameters and clinical scores, this time using a novel mobile-based technology, but also the notion that interrogating the oculomotor system with a novel eye-tracking technology can inform us of disease severity, as well as the cognitive status of MS participants.

Keywords: BICAMS; EDSS; MSFC; SDMT; digital biomarker; machine learning; mobile eye-tracker; multiple scleorsis.

Figure 1

Eye-tracking tasks. (A) Fixation: participants fixated a stationary target for 7 s, at one of 5 locations. (B) Pro-saccades: participants initially fixated a central fixation point, which disappeared after 1.0–3.5 s, after which a different target appeared at one of 8 eccentric locations for 1.5 s. (C) Anti-saccades: participants initially fixated a central fixation point, which disappeared after 1.0–3.5 s, after which a round target appeared at 10° to the left or right from the center. Participants were instructed to move their gaze in the opposite direction to the round target, where after 1,200 ms they were shown a square with an arrow inside that pointed in one of 4 random directions (left, right, up, or down; shown during 400 ms). The users then had to direct their gaze toward the arrow orientation corresponding to the arrow they saw in the preceding step. (D) Smooth pursuit: after initially fixating on a central cross, participants followed a moving target with a constant velocity of 8°/s (in this example, a downward-moving target).

Figure 2

Radar plot illustrating the z-score values for low- and high-EDSS subgroups for each oculomotor parameter. Significant subgroup differences are highlighted with bold labels. S, short amplitude pro-saccade; L, large amplitude pro-saccade. *p < 0.05, **p < 0.01.

Figure 3

Spearman correlations between select eye-tracking parameters and functional scores. (A) Fixation: BCEA95, (B) Pro-Saccades: large amplitude saccade latency, (C) Anti-Saccades: time to target, and (D) Smooth pursuit: average pursuit gain. All Spearman’s rho correlation values were calculated using the raw data. For visualization purposes only, the MSFC x-axes were rescaled [0.1–0.9] and log2-transformed. *p < 0.05 (corrected for multiple comparisons).

Figure 4

(A–D) Scatterplots of the relationship between the study participants’ clinical scores and the corresponding predicted value obtained by partial least squares regression analysis using the oculomotor parameters as predictors. (E) Heatmap visualization of the relative contribution (normalized absolute value of standardized regression coefficients) of each oculomotor parameter to each partial least squares regression predictor. Dark squares indicate lesser contributions to the model whereas lighter/yellow squares indicate greater contributions. Absent squares indicate that the parameter was not used in the final model. S, short amplitude pro-saccade; L, large amplitude pro-saccade.