Applying machine learning to dissociate between stroke patients and healthy controls using eye movement features obtained from a virtual reality task

Conventional neuropsychological tests do not represent the complex and dynamic situations encountered in daily life. Immersive virtual reality simulations can be used to simulate dynamic and interactive situations in a controlled setting. Adding eye tracking to such simulations may provide highly detailed outcome measures, and has great potential for neuropsychological assessment. Here, participants (83 stroke patients and 103 healthy controls) we instructed to find either 3 or 7 items from a shopping list in a virtual super market environment while eye movements were being recorded. Using Logistic Regression and Support Vector Machine models, we aimed to predict the task of the participant and whether they belonged to the stroke or the control group. With a limited number of eye movement features, our models achieved an average Area Under the Curve (AUC) of .76 in predicting whether each participant was assigned a short or long shopping list (3 or 7 items). Identifying participant as either stroke patients and controls led to an AUC of .64. In both classification tasks, the frequency with which aisles were revisited was the most dissociating feature. As such, eye movement data obtained from a virtual reality simulation contain a rich set of signatures for detecting cognitive deficits, opening the door to potential clinical applications.

Image-based and eye-based influences on binocular rivalry have similar spatial profiles

Binocular rivalry occurs when the images presented to the two eyes do not match. Instead of fusing into a stable percept, perception during rivalry alternates between images over time. However, during rivalry, perception can also resemble a patchwork of parts of both eyes' images. Such integration of image parts across eyes is relatively rare compared to integration of image parts presented to the same eye, suggesting that integration across space during rivalry is primarily rooted at the early monocular level of processing. However, recent evidence suggests that rivalry, and potentially also integration across space during rivalry, has its basis at multiple stages of processing, including stages at which monocular signals are minimal. As such, integration and competition at these later stages would be driven more by image-based factors, such as continuity and color than by eye of origin. Because "higher" visual areas also have increasingly larger receptive fields, image-based integration may occur over a larger spatial extent compared to monocular, eye-based integration. We therefore used rival images containing two separate image parts and varied the interimage-part distance (IIPD) to assess the relative contributions of eye of origin and image features to integration across space at increasing IIPDs. Our hypothesis was that the balance between these contributions would shift toward image features as IIPD increased. Instead, results show that the relative contributions of both factors to grouping remain constant as a function of IIPD. This indicates that image-based grouping is subject to similar spatial constraints as monocular, eye-based grouping, suggesting both kinds of grouping rely on similarly sized receptive fields.

Orientation bandwidths are invariant across spatiotemporal frequency after isotropic components are removed

It is well established that mammalian visual cortex possesses a large proportion of orientation-selective neurons. Attempts to measure the bandwidth of these mechanisms psychophysically have yielded highly variable results ( approximately 6 degrees -180 degrees ). Two stimulus factors have been proposed to account for this variability: spatial and temporal frequency; with several studies indicating broader bandwidths at low spatial and high temporal frequencies. We estimated orientation bandwidths using a classic overlay masking paradigm across a range of spatiotemporal frequencies (0.5, 2, and 8 c.p.d.; 1.6 and 12.5 Hz) with target and mask presented either monoptically or dichoptically. A standard three-parameter Gaussian model (amplitude and width, mean fixed at 0 degrees ) confirms that bandwidths generally increase at low spatial and high temporal frequencies. When incorporating an additional orientation-untuned (isotropic) amplitude component, however, we find that not only are the amplitudes of isotropic and orientation-tuned components highly dependent upon stimulus spatiotemporal frequency, but orientation bandwidths are highly invariant ( approximately 30 degrees half width half amplitude). These results suggest that previously reported spatiotemporally contingent bandwidth effects may have confounded bandwidth with isotropic (so-called cross-orientation) masking. Interestingly, the magnitudes of all monoptically derived parameter estimates were found to transfer dichoptically suggesting a cortical locus for both isotropic and orientation-tuned masking.