No abstract

Traveling waves have been studied to characterize the complex spatiotemporal dynamics of the brain. Several studies have suggested that the propagation direction of α traveling waves can be task dependent. For example, a recent electroencephalography (EEG) study from our group found that forward waves (i.e., occipital to frontal, FW waves) were observed during visual processing, whereas backward waves (i.e., frontal to occipital, BW waves) mostly occurred in the absence of sensory input. These EEG recordings, however, were obtained from different experimental sessions and different groups of subjects. To further examine how the waves’ direction changes between task conditions, 13 human participants were tested on a target detection task while EEG signals were recorded simultaneously. We alternated visual stimulation (5-s display of visual luminance sequences) and resting state (5 s of black screen) within each single trial, allowing us to monitor the moment-to-moment progression of traveling waves. As expected, the direction of α waves was closely linked with task conditions. First, FW waves from occipital to frontal regions, absent during rest, emerged as a result of visual processing, while BW waves in the opposite direction dominated in the absence of visual inputs, and were reduced (but not eliminated) by external visual inputs. Second, during visual stimulation (but not rest), both waves coexisted on average, but were negatively correlated. In summary, we conclude that the functional role of α traveling waves is closely related with their propagating direction, with stimulus-evoked FW waves supporting visual processing and spontaneous BW waves involved more in top-down control.

Travelling waves have been studied to characterize the complex spatiotemporal dynamics of the brain. Several studies have suggested that the propagation direction of travelling waves can be task-dependent. For example, a recent EEG study from our group found that forward waves (i.e. occipital to frontal, FW waves) were observed during visual processing, whereas backward waves (i.e. frontal to occipital, BW waves) mostly occurred in the absence of sensory input. These EEG recordings, however, were obtained from different experimental sessions and different groups of subjects. To further examine how the waves' direction changes between task conditions, 13 participants were tested on a target detection task while EEG signals were recorded simultaneously. We alternated visual stimulation (5 s display of visual luminance sequences) and resting state (5 s of black screen) within each single trial, allowing us to monitor the moment-to-moment progression of travelling waves. As expected, the waves' direction was closely linked with task conditions (visual processing vs. rest state). First, FW waves from occipital to frontal regions, absent during rest, emerged as a result of visual processing, while BW waves in the opposite direction dominated in the absence of visual inputs, and were reduced (but not eliminated) by external visual inputs. Second, during visual stimulation (but not rest), both waves coexisted on average, but were negatively correlated. That is, when the FW waves were stronger than expected based on alpha amplitude alone, the BW waves tended to be weaker, and vice-versa. In summary, we conclude that the functional role of travelling waves is closely related with their propagating direction, with stimulus-evoked FW waves supporting visual processing and spontaneous BW waves involved more in top-down control.

Modern feedforward convolutional neural networks (CNNs) can now solve some computer vision tasks at super-human levels. However, these networks only roughly mimic human visual perception. One difference from human vision is that they do not appear to perceive illusory contours (e.g. Kanizsa squares) in the same way humans do. Physiological evidence from visual cortex suggests that the perception of illusory contours could involve feedback connections. Would recurrent feedback neural networks perceive illusory contours like humans? In this work we equip a deep feedforward convolutional network with brain-inspired recurrent dynamics. The network was first pretrained with an unsupervised reconstruction objective on a natural image dataset, to expose it to natural object contour statistics. Then, a classification decision layer was added and the model was finetuned on a form discrimination task: squares vs. randomly oriented inducer shapes (no illusory contour). Finally, the model was tested with the unfamiliar "illusory contour" configuration: inducer shapes oriented to form an illusory square. Compared with feedforward baselines, the iterative "predictive coding" feedback resulted in more illusory contours being classified as physical squares. The perception of the illusory contour was measurable in the luminance profile of the image reconstructions produced by the model, demonstrating that the model really "sees" the illusion. Ablation studies revealed that natural image pretraining and feedback error correction are both critical to the perception of the illusion. Finally we validated our conclusions in a deeper network (VGG): adding the same predictive coding feedback dynamics again leads to the perception of illusory contours.