High Speed, Multi-sensor Light Field Deconvolution Microscopy for Whole Brain Recording of Neuronal Activity

 描述(由申请人提供)：以毫秒级的时间精度记录3D卷中的所有神经活动是大脑倡议的一个关键目标。最近，在一个协作性的 在与Vaziri实验室(IMP，维也纳)合作的项目中，我们采用了光场显微镜的策略，用于荧光神经钙反应的3D体积成像(Prevedel 2014)。这项技术通过同时捕捉入射光线的角度和强度，实现了从图像中计算重建3D体积。成像可以以荧光神经活动报告允许的最快速度进行(陈，2013);我们以20赫兹的频率对整个斑马鱼幼体的大脑进行了成像。然而，光场显微镜的空间分辨率很低，导致低信噪比(SNR)以及自动分割神经解剖结构的困难，而神经解剖结构是将神经活动与底层电路联系起来的关键。这种空间分辨率限制是光场显微镜的一个基本问题，因为要获得3D成像能力，就必须牺牲空间分辨率：相机上只有这么多像素。因此，我们在此建议通过将现有系统的总像素数增加一个数量级，同时利用帧速率提高六倍来提高信噪比，以单神经元分辨率对斑马鱼幼体进行全脑记录。我们现有的光场成像系统(Prevedel 2014)和其他系统(Levoy 2006，Cohen 2014)使用微透镜阵列来实现空间分辨率与轴向分辨率的权衡。另一种方法是在没有任何微透镜的情况下，使用一组相机捕捉光场。我们新颖的设计以前所未有的规模将这两种方法结合在一起。我们的可扩展数据采集系统(Willow，参见初步数据)与内部设计的摄像头相结合，将我们系统的成本限制在传统双光子显微镜的十分之一。通过这种方式，我们的目标是开发一种用户友好的系统，能够以与神经活动的自然时标相当的速度对3D体积中的所有神经元进行成像，同时着眼于为市场和广泛使用的系统打磨。

 DESCRIPTION (provided by applicant): Recording all the neural activity in a 3D volume with millisecond timescale precision is a key goal of the BRAIN initiative. Recently, in a collaborative project with the Vaziri lab (IMP, Vienna), we adapted the strategy of lightfield microscopy for 3D volumetric imaging of fluorescent neural calcium responses (Prevedel 2014). This technology enables computational reconstruction of a 3D volume from an image by simultaneously capturing the angle of incident light rays in addition to their intensity. Imaging can occur as quickly as the fluorescent neural activity reporter allows (Chen 2013); we imaged the entire larval zebrafish brain at 20 Hz. However, the spatial resolution for lightfield microscopy is poo, resulting in low signal to noise ratio (SNR) as well as difficulty in automatically segmenting neural anatomy, which is key to linking neural activity to underlying circuitry. This spatial resolution limit is a fundamental issue with lightfield microscopy, since to gain 3­D imaging capability, one must sacrifice spatial resolution: there are only so many pixels on the camera. Accordingly, we here propose to perform the first whole brain recording of a larval zebrafish with single neuron resolution by increasing the total pixel count of our existing system by an order of magnitude whilst improving the SNR by leveraging a six fold increase in frame rate. Our existing lightfield imaging system (Prevedel 2014), and others (Levoy 2006, Cohen 2014), use an array of microlenses to effect the trade­off of spatial for axial resolution. An alternativ approach captures the lightfield using an array of cameras, without any microlenses. Our novel design combines both approaches at an unprecedented scale. Our scalable data acquisition system (Willow, see preliminary data) combined with cameras designed in­house limit the cost of our system to 1/10 of a traditional two­photon microscope. In this way we aim to develop a user­friendly system capable of imaging all of the neurons in a 3D volume, at speeds comparable to the natural timescales of neural activity, whilst keeping an eye towards polishing our system for marketability and widespread use.