Yazdanbakhsh Arash, Grossberg Stephen
Department of Cognitive and Neural Systems, Boston University, Boston, MA 02215, USA.
Neural Netw. 2004 Jun-Jul;17(5-6):707-18. doi: 10.1016/j.neunet.2004.06.005.
Perceptual grouping is well known to be a fundamental process during visual perception, notably grouping across scenic regions that do not receive contrastive visual inputs. Illusory contours are a classical example of such groupings. Recent psychophysical and neurophysiological evidence have shown that the grouping process can facilitate rapid synchronization of the cells that are bound together by a grouping, even when the grouping must be completed across regions that receive no contrastive inputs. Synchronous grouping can hereby bind together different object parts that may have become desynchronized due to a variety of factors, and can enhance the efficiency of cortical transmission. Neural models of perceptual grouping have clarified how such fast synchronization may occur by using bipole grouping cells, whose predicted properties have been supported by psychophysical, anatomical, and neurophysiological experiments. These models have not, however, incorporated some of the realistic constraints in which groupings in the brain are conditioned, notably the measured spatial extent of long-range interactions in layer 2/3 of a grouping network, and realistic synaptic and axonal signaling delays within and across cells in different cortical layers. This work addresses the question: Can long-range interactions that obey the bipole constraint achieve fast synchronization under realistic anatomical and neurophysiological constraints that initially desynchronize grouping signals? Can the cells that synchronize retain their analog sensitivity to changing input amplitudes? Can the grouping process complete and synchronize illusory contours across gaps in bottom-up inputs? Our simulations show that the answer to these questions is Yes.
众所周知,知觉分组是视觉感知过程中的一个基本过程,特别是跨越未接收对比性视觉输入的场景区域进行分组。错觉轮廓就是这种分组的一个经典例子。最近的心理物理学和神经生理学证据表明,即使分组必须跨越未接收对比性输入的区域完成,分组过程也可以促进由分组绑定在一起的细胞的快速同步。同步分组由此可以将可能由于各种因素而变得不同步的不同物体部分绑定在一起,并可以提高皮质传递的效率。知觉分组的神经模型已经阐明了通过使用双极分组细胞这种快速同步是如何发生的,其预测特性已得到心理物理学、解剖学和神经生理学实验的支持。然而,这些模型并未纳入大脑中分组所依赖的一些现实约束条件,特别是分组网络第2/3层中长程相互作用的测量空间范围,以及不同皮质层内和细胞间现实的突触和轴突信号延迟。这项工作解决了以下问题:在最初使分组信号不同步的现实解剖学和神经生理学约束条件下,遵循双极约束的长程相互作用能否实现快速同步?同步的细胞能否保持其对变化的输入幅度的模拟敏感性?分组过程能否完成并使跨越自下而上输入中的间隙的错觉轮廓同步?我们的模拟表明这些问题的答案是肯定的。