The ability to interpret and predict other people's actions is highly evolved in humans and is believed to play a central role in their cognitive behavior. However, there is no direct evidence that this ability confers a tangible benefit to sensory processing. Our quantitative behavioral experiments show that visual discrimination of a human agent is influenced by the presence of a second agent. This effect depended on whether the two agents interacted (by fighting or dancing) in a meaningful synchronized fashion that allowed the actions of one agent to serve as predictors for the expected actions of the other agent, even though synchronization was irrelevant to the visual discrimination task. Our results demonstrate that action understanding has a pervasive impact on the human ability to extract visual information from the actions of other humans, providing quantitative evidence of its significance for sensory performance.

The low-level deficits associated with amblyopia have been studied extensively, but very little is known about potential impairments to higher-level visual processing such as object recognition or structure-from-motion. Studies on biological motion, a complex form of structure-from-motion depicting human actions, have demonstrated that normal observers can analyze these patterns more effectively when they are shown in their original upright configuration as opposed to inverted upside-down (feet-up head-down). We measured this inversion effect quantitatively for both the dominant and amblyopic eyes of amblyopic observers. We found a modest ( approximately 30%) loss in sensitivity in the amblyopic eye for both upright and inverted actors, which we attribute to low-level deficits. However, we found no difference in the inversion effect between the two eyes, both showing an average 1/2 log-unit drop in sensitivity between upright and inverted displays. Our data provide a quantitative estimate of the inversion effect for biological motion, and demonstrate that higher-level processing in the motion hierarchy is not affected by amblyopia.

Amblyopia is a developmental disorder of spatial vision that results from abnormal early visual experience usually due to the presence of strabismus, anisometropia, or both strabismus and anisometropia. Amblyopia results in a range of visual deficits that cannot be corrected by optics because the deficits reflect neural abnormalities. Biological motion refers to the motion patterns of living organisms, and is normally displayed as points of lights positioned at the major joints of the body. In this experiment, our goal was twofold. We wished to examine whether the human visual system in people with amblyopia retained the higher-level processing capabilities to extract visual information from the synchronized actions of others, therefore retaining the ability to detect biological motion. Specifically, we wanted to determine if the synchronized interaction of two agents performing a dancing routine allowed the amblyopic observer to use the actions of one agent to predict the expected actions of a second agent. We also wished to establish whether synchronicity sensitivity (detection of synchronized versus desynchronized interactions) is impaired in amblyopic observers relative to normal observers. The two aims are differentiated in that the first aim looks at whether synchronized actions result in improved expected action predictions while the second aim quantitatively compares synchronicity sensitivity, or the ratio of desynchronized to synchronized detection sensitivities, to determine if there is a difference between normal and amblyopic observers. Our results show that the ability to detect biological motion requires more samples in both eyes of amblyopes than in normal control observers. The increased sample threshold is not the result of low-level losses but may reflect losses in feature integration due to undersampling in the amblyopic visual system. However, like normal observers, amblyopes are more sensitive to synchronized versus desynchronized interactions, indicating that higher-level processing of biological motion remains intact. We also found no impairment in synchronicity sensitivity in the amblyopic visual system relative to the normal visual system. Since there is no impairment in synchronicity sensitivity in either the nonamblyopic or amblyopic eye of amblyopes, our results suggest that the higher order processing of biological motion is intact.

No abstract

Accurate segmentation of retinal fluids in 3D Optical Coherence Tomography images is key for diagnosis and personalized treatment of eye diseases. While deep learning has been successful at this task, trained supervised models often fail for images that do not resemble labeled examples, e.g. for images acquired using different devices. We hereby propose a novel semi-supervised learning framework for segmentation of volumetric images from new unlabeled domains. We jointly use supervised and contrastive learning, also introducing a contrastive pairing scheme that leverages similarity between nearby slices in 3D. In addition, we propose channel-wise aggregation as an alternative to conventional spatial-pooling aggregation for contrastive feature map projection. We evaluate our methods for domain adaptation from a (labeled) source domain to an (unlabeled) target domain, each containing images acquired with different acquisition devices. In the target domain, our method achieves a Dice coefficient 13.8% higher than SimCLR (a stateof-the-art contrastive framework), and leads to results comparable to an upper bound with supervised training in that domain. In the source domain, our model also improves the results by 5.4% Dice, by successfully leveraging information from many unlabeled images.