Geometric-Optical Illusions from Early Visual Cortex: Computational Principles and Top-Down Feedback

经典的几何视错觉（GI），即当视觉轮廓适当地相互作用时对几何特性的误解，数百年来一直在挑战理解它们的研究努力。 早期视皮层（EVC）的神经生理学机制长期以来一直与视知觉分组有关，但它们在GI中的作用仍然很大程度上未被探索。 尚未解决的问题包括：如何在EVC的视觉输入扭曲，导致错觉？从更高的皮层区域到EVC的自上而下的反馈对于地理标志的感知是必要的吗？这个跨学科的项目旨在通过数学建模和心理物理实验来探索这些问题。首先，我们建立了一个生物学上合理的，EVC的非线性动力学模型，捕捉其神经元的方向调谐特性和不同的感受野大小，以及这些神经元之间的神经连接在一个循环的神经回路，有或没有从更高的视觉区域自上而下的反馈的存在。 该模型将帮助我们探索视觉轮廓段的检测器之间的相互作用如何提供视觉输入的预注意失真，以及自上而下的反馈如何调节这种相互作用，从而产生GI。 其次，我们使用视觉心理物理实验来探测和测试的地理标志，使用视觉输入的启发，并希望新的地理标志预测，模型和计算的考虑。在建模和实验的组成部分，中心问题包括视觉输入元素之间的几何和规模关系的关键作用，和中央-周边二分法，即中央和周边视野之间的对比，在自上而下的反馈的可用性和视敏度。该项目将在图宾根大学计算机科学系李兆平教授的实验室进行，利用其实验室理论神经科学和实验心理学专业知识的独特组合以及一流的基础设施支持。

The classical geometric-optical illusions (GIs), i.e. misperceptions of geometric properties when visual contours interact appropriately have defied research efforts to understand them for hundreds of years. Neurophysiological mechanisms in early visual cortex (EVC) have long been implicated for visual perceptual grouping, but their role in the GIs remains largely unexplored. Unsolved questions include: how is visual input distorted in EVC to cause the illusory perceptions? Are top-down feedbacks from higher cortical areas to EVC necessary for the perceptions of GIs? This interdisciplinary project aims to probe these questions with mathematical modeling and psychophysical experiments. First, we build a biologically plausible, nonlinear dynamic model of EVC, capturing the orientation tuning properties and different receptive field sizes of its neurons and the neural connections between these neurons in a recurrent neural circuit, with or without the presence of top-down feedback from higher visual areas. The model will help us to explore how interactions between detectors of visual contour segments could provide a pre-attentive distortion of visual input, and how top-down feedback modulates such interactions, leading to the GIs. Second, we use visual psychophysical experiments to probe and test the GIs, using visual inputs inspired by, and hopefully new GIs predicted by, the model and computational considerations. In both the modeling and experimental components, the central questions include the critical role of the geometric and scale relationships between visual input elements, and the central-peripheral dichotomy, namely the contrast between the central and peripheral visual fields in the availability of top-down feedbacks and in visual acuity. This project will be conducted in Prof. Li Zhaoping’s lab in the Department of Computer Science in the University of Tübingen, taking advantage of the unique combination of theoretical neuroscience and experimental psychology expertise in her lab and the superb infrastructure support.