Motion Processing in the Retina

视网膜是大脑通向视觉世界的窗口。这个神经元回路通过将光转换成电信号来开始视觉过程，就像摄像机中的CCD一样。但这些信号并不是以原始形式传回大脑的：它们立即在视网膜内进行处理，以提取最重要的信息。换句话说，视网膜是一个计算的神经回路。视网膜进行的最重要的计算之一是检测物体的运动，使动物能够在世界中导航并检测捕食者或猎物。许多神经元还提供关于物体移动方向的信息。现代神经科学的关键挑战之一是理解这样的计算是如何通过神经元的生物硬件来实现的，这些神经元通过突触连接连接到电路中。理解信号如何在突触处传递是特别重要的，因为当视觉信号通过视网膜时，该过程涉及视觉信号的关键转换。为了理解视网膜是如何计算运动方向的，我们将观察视觉信号在不同类型的神经元之间传输时，同时在不同方向上传递刺激。为了实现这一目标，我们将使基因修饰的斑马鱼表达荧光蛋白，这些荧光蛋白在视网膜处理视觉刺激时发出神经元和突触活动的信号。这些信号由发射的绿色或红色光组成，其可以使用专用（多光子）显微镜观察，该显微镜可以用于观察活体动物视网膜内的事件。与摄像机相比，视网膜的一个缺点是感光神经元，即光感受器，比CCD需要更长的时间来感知光线。如果物体以任何速度移动，这都是一个问题，因为当视觉信号到达内层视网膜时，产生该信号的物体可能已经移动了相当大的距离。例如，安迪·穆雷（Andy Murray）发出的网球将在我们的光感受器做出反应的80毫秒内移动近5米。因此，视网膜进行进一步的计算，即基于物体沿直线移动的假设，将物体在视网膜上的位置外推到未来约80 ms。这个过程被称为“运动预测”，我们还将研究它是如何通过内部视网膜的神经回路来实现的。运动预测的过程并不完美：如果物体突然改变方向，预测就会出错。视网膜通过重新计算新的运动方向来迅速纠正这个错误。需要重新计算移动物体的轨迹是由一个“警报信号”突出的，该信号由大量神经元同步的电活动的瞬时爆发组成。了解这种警报信号是如何产生的是了解视网膜如何处理运动的一个关键方面。我们希望这项研究提供的见解将在广泛的应用中证明是有用的，从开发改进的视网膜修复术为盲人提供视觉到更高效的安全摄像头能够跟踪移动物体。我们还相信，我们正在开发的观察突触传递的方法将普遍用于分析大脑其他部分的神经元回路如何对传入信息进行计算。

The retina is the brains window to the visual world. This circuit of neurons begins the process of vision by converting light into an electrical signal, much like the CCD in a video camera. But these signals are not transmitted back to the brain in raw form: they are immediately processed within the retina to allow extraction of the most important information. In other words, the retina is a neural circuit that computes. One of the most important computations that the retina carries out is to detect the motion of objects, allowing the animal to navigate the world and detect predators or prey. Many neurons also provide information about the direction in which the object moves. One of the key challenges in modern neuroscience is to understand how such a computation is implemented by the biological hardware of neurons connected into circuits through synaptic connections. it is particularly important to understand how signals are transmitted at synapses because this process is involved in key transformation of the visual signal as it travels through the retina. To understand how the retina computes direction of motion we will observe the visual signal as it is transmitted between the different types of neuron while delivering stimuli moving in different directions. To achieve this, we will make genetically-modified zebrafish expressing fluorescent proteins that signal the activity of neurons and synapses as the retina processes visual stimuli. These signals consist of the emission of green or red light, which can be observed using a specialized (multiphoton) microscope that can be used to observe events within the retina of the live animal. One of the disadvantages of the retina compared to a video camera is that the light-sensitive neurons, the photoreceptors, take much longer to sense light than a CCD. This is a problem if the object is moving at any speed, because by the time the visual signal reaches the inner retina, the object generating that signal may have moved a considerable distance. For instance, a tennis ball served by Andy Murray will have moved nearly 5 m during the ~80 ms it takes our photoreceptors to respond. The retina therefore carries out a further computation, which is to extrapolate the position of the object on the retina ~80 ms into the future, based on the assumption that it is moving in a straight line. This process is called "motion prediction" and we will also investigate how it is implemented by the neural circuitry of the inner retina. The process of motion prediction is not perfect: if the object suddenly changes direction, the predictions will be wrong. The retina corrects this mistake quickly by re-calculating the new direction of motion. The need to recalculate the trajectory of the moving object is highlighted by an "alarm signal" consisting of a transient burst of electrical activity synchronized across a large number of neurons. Understanding how this alarm signal is generated is a key aspect of understanding how motion is processed by the retina.We hope that the insights provided by this research will prove useful in a wide range of applications, from developing improved retinal prosthetics to provide vision for the blind to more efficient security cameras able to track moving objects. We also believe that the methods we are developing for observing synaptic transmission will be of general use in analyzing how circuits of neurons in other parts of the brain operate to carry out computations on incoming information.