The DeepTune framework for modeling and characterizing neurons in visual cortex area V4
The DeepTune framework for modeling and characterizing neurons in visual cortex area V4
复制标题
用于建模和表征视觉皮层 V4 区域神经元的 DeepTune 框架
DOI：
10.1101/465534
复制
发表时间：
2018
期刊：
影响因子：
--
通讯作者：
Bin Yu
中科院分区：
文献类型：
--
作者：
R. Abbasi;Yuansi Chen;Adam Bloniarz;M. Oliver;B. Willmore;J. Gallant;Bin Yu
深度神经网络模型最近被证明可以有效预测灵长类视觉皮层V4区域的单神经元反应。尽管这些模型的预测准确性很高，但通常很难解释。这限制了它们在表征V4神经元功能方面的适用性。在这里，我们提出了DeepTune框架作为一种方法，以引起对V4区单个神经元的基于深度神经网络的模型的解释。V4是腹侧视觉通路中的中层视觉皮层区域。其职能作用尚未得到很好的理解。使用由数千张静态自然图像刺激的71个V4神经元的记录数据集，我们构建了每个神经元18个基于神经网络的模型的集合，这些模型可以准确预测其对刺激图像的响应。为了解释和可视化这些模型，我们使用稳定性标准，通过将18个模型合并在一起来形成最佳刺激（DeepTune图像）。这些DeepTune图像不仅证实了先前关于V4区存在不同形状和纹理调谐的发现，而且还提供了单个V4神经元感受野的丰富，具体和自然的表征。对71个神经元的DeepTune图像的群体分析揭示了不同类型的曲率调谐如何分布在V4中。此外，它还表明了近一半的V4神经元的强抑制调谐。虽然我们只关注V4区域，但DeepTune框架可以更广泛地应用于增强对其他视觉皮层区域的理解。
Deep neural network models have recently been shown to be effective in predicting single neuron responses in primate visual cortex areas V4. Despite their high predictive accuracy, these models are generally difficult to interpret. This limits their applicability in characterizing V4 neuron function. Here, we propose the DeepTune framework as a way to elicit interpretations of deep neural network-based models of single neurons in area V4. V4 is a midtier visual cortical area in the ventral visual pathway. Its functional role is not yet well understood. Using a dataset of recordings of 71 V4 neurons stimulated with thousands of static natural images, we build an ensemble of 18 neural network-based models per neuron that accurately predict its response given a stimulus image. To interpret and visualize these models, we use a stability criterion to form optimal stimuli (DeepTune images) by pooling the 18 models together. These DeepTune images not only confirm previous findings on the presence of diverse shape and texture tuning in area V4, but also provide rich, concrete and naturalistic characterization of receptive fields of individual V4 neurons. The population analysis of DeepTune images for 71 neurons reveals how different types of curvature tuning are distributed in V4. In addition, it also suggests strong suppressive tuning for nearly half of the V4 neurons. Though we focus exclusively on the area V4, the DeepTune framework could be applied more generally to enhance the understanding of other visual cortex areas.
登录
查看更多内容
影响因子：
56.9
作者：
GALLANT, JL;BRAUN, J;VANESSEN, DC
通讯作者：
VANESSEN, DC
影响因子：
2.5
作者：
B. Willmore;J. Mazer;J. Gallant
通讯作者：
B. Willmore;J. Mazer;J. Gallant
影响因子：
2.5
作者：
Gallant, JL;Connor, CE;VanEssen, DC
通讯作者：
VanEssen, DC
影响因子：
2.5
作者：
David, Stephen V.;Hayden, Benjamin Y.;Gallant, Jack L.
通讯作者：
Gallant, Jack L.