Compressive Light Field microscopy for optogenetic neural activity tracking

摘要：了解活着的大脑产生感知、认知和行为的机制 需要能够同时记录和控制多个神经元的电活动。完全地 重建调节特定神经操作的神经活动模式对于完全 了解其潜在的机制。这就需要一种能够大规模测量神经活动的方法。 具有高时空分辨率的比例尺。基因编码的活动传感器的功能成像是其中之一 实现这一目标的最有希望的途径之一，因为它允许密集抽样和明确 分离附近的神经元1，2。然而，即使是目前可用的技术也不能捕捉到 同时，以高速和细胞分辨率的大容量大脑。此外，这些现有的 方法依赖于昂贵而复杂的硬件，并且不能很容易地适应在 无拘无束的动物。因为大脑计划的一个中心目标是实现对神经的大规模记录 为了更好地研究动物的行为活动，需要一种新的成像方法来克服这些技术挑战。 为了在紧凑和廉价的设备中实现高速、体积成像，我们建议开发 一种新的成像方式-压缩四维(4D)光场显微镜(LFM)。在这种方法中，我们 将结合光场显微镜和压缩传感的优势来提取活性 数以千计的单个神经元，通过散布组织，具有高空间和时间分辨率。在……里面 在传统显微镜下，光电探测器只对光子的强度进行采样。在线性调频中，传感器还 捕捉灯光的角度。这允许3D重建，因为位置和角度允许回溯 光线。这样的方案可以简单地通过在成像路径中放置小透镜阵列来实现(图1a)。 生成的4D光场在每个时间帧3-6提供完整的体积数据。因此，可以对3D活动进行采样 在相机受限的帧速率下，比多光子或光片等传统方法快得多 显微镜。出于这些原因，用于脑功能成像的LFM可能会给实验神经科学带来革命性的变化。 不幸的是，工作在单光子区域的成像方法会受到光散射的影响。至 在哺乳动物大脑的图像活动中，我们必须考虑组织散射的影响，通过3D体积。 我们建议应用一种新的方法来处理光场数据，以更好地重建神经元 散射介质37，38。由于散射的作用是传播光的角度，所以测量 角度信息本质上有助于表征和缓解散射效果。我们的方法利用 利用光发射在三维空间和时间上的稀疏性的压缩传感算法。 重要的是，我们方法的体积和分辨率限制不是由捕获的像素数来设置的，而是 而是由任何给定时间活跃神经元的数量决定的。因此，我们将能够本地化和测量 在大量大脑皮层组织中，数千或数百万个单独活跃的神经元的活动。

SUMMARY: Understanding the mechanisms by which the living brain derives perception, cognition and behavior requires the ability to record and control electrical activity in many neurons simultaneously. Completely reconstructing the pattern of neural activity that mediates a specific neural operation is critical for fully understanding its underlying mechanism. This requires an approach that can measure neural activity on a large scale with high spatial and temporal resolution. Functional imaging of genetically encoded activity sensors is one of the most promising avenues towards achieving this goal because it permits dense sampling and unambiguous separation of nearby neurons1, 2. However, even techniques currently available cannot capture neural activity in large volumes of the brain simultaneously, at high speed, and with cellular resolution. Furthermore, these existing approaches rely on expensive and sophisticated hardware, and cannot be readily adapted for imaging in unrestrained animals. Since a central goal of the BRAIN initiative is to achieve large-scale recording of neural activity in behaving animals, a new imaging approach is needed to overcome these technical challenges. To achieve high speed, volumetric imaging in a compact and inexpensive device, we propose to develop a new imaging modality - compressive four-dimensional (4D) light field microscopy (LFM). In this approach, we will combine the advantages of light field microscopy with compressed sensing to extract the activity of thousands of individual neurons with high spatial and temporal resolution, through scattering tissue. In conventional microscopy, the photo-detector only samples the intensity of photons. In LFM, the sensor also captures the angle of the light. This allows 3D reconstruction, since position and angle enable back-tracing of rays of light. Such a scheme can be achieved simply by placing a lenslet array in the imaging pathway (Fig. 1A). The resulting 4D light field gives complete volumetric data at each time frame3-6. 3D activity can thus be sampled at camera-limited frame rates, much faster than conventional methods such as multiphoton or light sheet microscopy. For these reasons, LFM for functional brain imaging could revolutionize experimental neuroscience. Unfortunately, imaging methods which operate in the one-photon regime suffer from light scattering. To image activity in the mammalian brain, we must consider the effects of tissue scattering through the 3D volume. We propose to apply a new approach to processing light field data for better reconstructions of neurons through scattering media37,38. Since the effect of scattering is to spread the angles of propagation of light, measuring angle information inherently helps to characterize and mitigate scatter effects. Our method leverages compressed sensing algorithms, which exploit the sparsity of the light emission in 3D space and time. Importantly, the volume and resolution limits of our method are not set by the number of pixels captured, but rather by the number of active neurons at any given time. Thus, we will be able to localize and measure the activity of thousands or millions of individually active neurons in large volumes of cerebral cortical tissue.