National Vision Research Institute, Australian College of Optometry, Carlton, Vic. 3053, Australia.
Neuroscience. 2013 May 1;237:19-28. doi: 10.1016/j.neuroscience.2013.01.030. Epub 2013 Jan 26.
Neurons in the primary visual cortex are often classified as either simple or complex based on the linearity (or otherwise) of their response to spatial luminance contrast. In practice, classification is typically based on Fourier analysis of a cell's response to an optimal drifting sine-wave grating. Simple cells are generally considered to be linear and produce responses modulated at the fundamental frequency of the stimulus grating. In contrast, complex cells exhibit significant nonlinearities that reduce the response at the fundamental frequency. Cells can therefore be easily and objectively classified based on the relative modulation of their responses - the ratio of the phase-sensitive response at the fundamental frequency of the stimulus (F₁) to the phase-invariant sustained response (F₀). Cells are classified as simple if F₁/F₀>1 and complex if F₁/F₀<1. This classification is broadly consistent with criteria based on the spatial organisation of cells' receptive fields and is accordingly presumed to reflect disparate functional roles of simple and complex cells in coding visual information. However, Fourier analysis of spiking responses is sensitive to the number of spikes available - F₁/F₀ increases as the number of spikes is reduced, even for phase-invariant complex cells. Moreover, many complex cells encountered in the laboratory exhibit some phase sensitivity, evident as modulation of their responses at the fundamental frequency. There currently exists no objective quantitative means of assessing the significance or otherwise of these modulations. Here we derive a statistical basis for objectively assessing whether the modulation of neuronal responses is reliable, thereby adding a level of statistical certainty to measures of phase sensitivity. We apply our statistical analysis to neuronal responses to moving sine-wave gratings recorded from 367 cells in cat primary visual cortex. We find that approximately 60% of complex cells exhibit statistically significant (α<0.01) modulation of their responses to optimal moving gratings. These complex cells are phase sensitive and reliably encode spatial phase.
初级视皮层中的神经元通常根据其对空间亮度对比度的响应的线性(或其他)来分为简单或复杂。在实践中,分类通常基于细胞对最佳漂移正弦光栅的响应的傅里叶分析。简单细胞通常被认为是线性的,并且产生以刺激光栅的基频调制的响应。相比之下,复杂细胞表现出显著的非线性,从而降低了基频处的响应。因此,细胞可以基于其响应的相对调制(即刺激(F₁)的基频的相位敏感响应与相位不变持续响应(F₀)的比)轻松且客观地进行分类。如果 F₁/F₀>1,则细胞被分类为简单,如果 F₁/F₀<1,则细胞被分类为复杂。这种分类与基于细胞感受野的空间组织的标准广泛一致,因此被认为反映了简单和复杂细胞在编码视觉信息方面的不同功能作用。然而,尖峰响应的傅里叶分析对可用的尖峰数量敏感 - 即使对于相位不变的复杂细胞,F₁/F₀也会随着尖峰数量的减少而增加。此外,在实验室中遇到的许多复杂细胞表现出一定的相位敏感性,其特征在于其在基频处的响应调制。目前没有客观的定量方法来评估这些调制的重要性或其他方面。在这里,我们为客观评估神经元响应的调制是否可靠提供了一个统计基础,从而为相位敏感性的测量增加了一定的统计确定性。我们将我们的统计分析应用于从猫的初级视皮层中记录的 367 个细胞对移动正弦光栅的反应。我们发现,大约 60%的复杂细胞对最佳移动光栅的反应表现出统计学上显著(α<0.01)的调制。这些复杂细胞具有相位敏感性并且可靠地编码空间相位。