Deciding Where to Look Next: Frontal Eye Field's Role during Natural Viewing

 描述(由申请者提供)：这项研究的长期目标是了解大脑如何决定我们在现实世界中的位置。很多因素都会影响我们的眼动(眼跳)。例如，我们更有可能看到突出的物体(即那些显眼的物体)，比如蓝天中的一个鲜红的气球。我们也更有可能查看与目标相关的对象(即那些与我们的目标有共同特征的对象)，例如搜索香蕉时的黄色对象。几十年来，计算机视觉的研究人员一直在开发基于这些因素的模型，以预测我们将眼睛移动到的位置。神经生物学的研究人员也一直在研究眼跳选择，他们认为前部眼场(FEF)起着很大的作用，因为FEF同时编码视觉特征和眼动。但由于FEF同时编码视觉特征和眼跳，因此很难在自然观看过程中解析FEF活动。出于这个原因，过去的实验主要是使用简单的、受限制的人工刺激任务来研究FEF。在这个项目中，我将使用自然场景的图像，这更接近真实世界的复杂性。我将用细胞外电极记录清醒的恒河猴在观看自然场景时的FEF。为了确定FEF在决定在自然场景中下一步在哪里扫视时所扮演的角色，我将研究FEF如何编码预测眼跳的视觉特征。在我的两个目标中，我将测试FEF如何编码显著(目标1)和目标依赖(目标2)。我将建立一个模型，使用视觉特征(突显和目标依赖)以及眼睛运动来解释神经活动，这是神经活动的一个令人困惑的来源。该模型将利用计算机视觉和机器学习算法来观察这些自然场景中的大量相关性和视觉特征的效果。为这些目的开发的神经数据分析方法将使研究许多大脑区域的研究人员更容易使用自然场景。此外，了解大脑如何在自然场景中选择在哪里扫视对神经学和精神病学的健康和疾病具有重要影响。包括精神分裂症、自闭症和帕金森氏症在内的几种疾病会影响眼跳的选择。更好地了解视觉特征、眼球运动和FEF活动之间的联系有望增加对这些疾病的理解，并允许开发新的诊断工具。

 DESCRIPTION (provided by applicant): The long-term goal of this research is to understand how the brain decides where we look in the real world. Many factors influence our eye movements (saccades). For instance, we are more likely to look at salient objects (i.e. those that are conspicuous), such as a bright red balloon in a blue sky. We are also more likely to look at goal-dependent objects (i.e. those that share features with our goals), such as a yellow object when searching for a banana. For several decades, researchers in computer vision have been developing models based on these factors to predict the locations to which we move our eyes. Researchers in neurobiology have also been studying saccade selection, and have suggested the frontal eye field (FEF) plays a large role, as the FEF encodes both visual features and eye movements. But because the FEF encodes both visual features and saccades, it is very difficult to parse FEF activity during natural viewing. For this reason, past experiments have primarily investigated the FEF using simple, constrained tasks with artificial stimuli. In this project, I wil use images of natural scenes, which better approximate the complexity of the real world. I will record with extracellular electrodes from the FEF of awake, behaving rhesus monkeys, while they view natural scenes. In order to determine the FEF's role in the decision of where to saccade next in natural scenes, I will investigate how the FEF encodes visual features that predict saccades. In my two aims, I will test how the FEF encodes salience (Aim 1) and goal-dependence (Aim 2). I will build a model that explains neural activity using visual features (salience and goal-dependence) along with eye movements, which are a confounding source of neural activity. This model will take advantage of computer vision and machine learning algorithms in order to look at the effects of large numbers of correlates and visual features in these natural scenes. The neural data analysis methods developed for these aims will allow researchers that study many brain areas to more easily use natural scenes. Additionally, understanding how the brain chooses where to saccade in natural scenes have important consequences for neurologic and psychiatric health and disease. Several diseases including schizophrenia, autism, and Parkinson's impair the choice of saccades. A better understanding of the link between visual features, eye movements, and FEF activity promises to increase understanding of these diseases and allow the development of novel diagnostic tools.