An Autonomous Rapidly Adaptive Multiphoton Microscope for Neural Recording and Stimulation

用于神经记录和刺激的自主快速自适应多光子显微镜

• 批准号: 10739050
• 负责人: Mark A Foster
• 金额: $ 200.86万
• 依托单位: JOHNS HOPKINS UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 2023
• 资助国家: 美国
• 起止时间: 2023-09-20 至 2026-08-31
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10739050
• 关键词: 3-Dimensional; Action Potentials; Algorithms; Architecture; Auditory area; Behavior; Behavioral; Benchmarking; Brain; Cells; Code; Computer software; Coupled; Detection; Dimensions; Feedback; Fluorescence; Generations; Goals; Hybrids; Image; In Vitro; Label; Light; Maps; Methods; Microscope; Microscopy; Monitor; Motion; Nature; Neurons; Neuropil; Noise; Optics; Output; Pattern; Recovery; Research; Resolution; Resources; Running; Sampling; Scanning; Signal Transduction; Slice; Source; Speed; Spottings; Stimulus; Structure; System; Techniques; Technology; Time; Tissues; Validation; Visualization; calcium indicator; code development; detector; improved; in vivo; lightspeed; machine learning algorithm; multi-photon; multiphoton microscopy; neural; neural network; neurotransmission; novel; parallelization; prevent; programs; response; sound; spatiotemporal; temporal measurement; theories

PROJECT SUMMARY/ABSTRACT Multiphoton microscopy of cells labeled with genetically encoded calcium indicators (GECIs) enables detection and correlation of fine neuronal structure to functional activity with cellular resolution. However, the point-scanning nature of conventional multiphoton systems makes it difficult to achieve sufficient temporal resolution for activity mapping over volumes spanning multiple circuits. Advances have largely come from developing faster methods of raster scanning. To this end, recent techniques focus on developing passive optical scanners that sequentially scan a focused spot in one dimension and these techniques demonstrate recording from an impressive 2.2 million neurons/sec. These unparalleled recording rates are enabled by passive axial multiplexing and optimized spatial sampling to maximize SNR, leading the microscope to be limited in speed by the fluorescence lifetime. Given this limit, further improvements in neurons/sec activity recording will necessitate either bright fluorescent indicators with shorter lifetime or better use of the fluorescence lifetime limited sampling rate. Since labeled neurons occupy a small volume fraction (< 5%) of a typical FOV, significant gains in neuronal recording rates are possible through the combination of these optimized scanning techniques with parallelized coded excitation. Through this research program, we will develop such an adaptive multiphoton microscope. Our approach will leverage a hybrid volumetric scanning architecture that combines the benefits of passive axial multiplexing and optimal sampling with simultaneous multi-beam volumetrically patterned excitation. This rapid and highly agile microscope platform, which we term Coded Raster Scanning Hybrid (CRaSH), will be coupled with machine learning algorithms and high-speed feedback circuits to adaptively adjust the scan conditions in response to the observed experimentally relevant activity and motion. Our goal is to develop a microscope that scans smarter and autonomously optimizes the use of resources to maximize the number and SNR of recorded neurons in response to their motion and activity. First (Aim 1), we will develop and construct the CRaSH microscope system. Our approach leverages a novel axial multiplexing approach that we term Binary Expansion Axial Multiplexing Module (BEAMM). We plan to develop the microscope in three stages starting with a non-adaptive scanning BEAMM microscope, then moving to an adaptive 2D CRaSH, and finally moving to the full adaptive 3D CRaSH. Second (Aim 2) we will develop an adaptive, hardware/software solution that uses computationally efficient algorithms running on FPGAs, to recover neural signals and adapt the excitation codes of our CRaSH microscopes. At first, we will optimize the acquisition hardware and software architecture for in vitro application. Subsequently, we will extend our algorithms to tackle in vivo challenges such as motion and uncorrelated activities. Lastly (Aim 3), we will benchmark are various microscope realizations (scanning BEAMM, 2D CRaSH, and 3D CRaSH) in brain slices and in vivo and then investigate the application of each realization to studying the functional representation of sounds in the auditory cortex.

项目摘要/摘要 用遗传编码的钙指示剂（GECIS）标记的细胞的多光子显微镜可以检测和相关性神经元结构与细胞分辨率的功能活性。但是，常规多光子系统的点扫描性质使得在跨越多个电路的体积上实现足够的时间分辨率来实现足够的时间分辨率。进步很大程度上来自开发更快的栅格扫描方法。为此，最近的技术着重于开发被动光学扫描仪，这些扫描仪会在一个维度上依次扫描一个集中点，这些技术证明了来自令人印象深刻的220万神经元/秒的记录。这些无与伦比的记录率是通过被动轴向多路复用和优化的空间采样来启用的，以最大化SNR，从而导致显微镜在荧光寿命上受到限制的速度。鉴于此限制，神经元/SEC活性记录的进一步改善将需要寿命较短或更好地使用荧光寿命有限的采样率的明亮荧光指标。由于标记的神经元占据了典型FOV的少量分数（<5％），因此通过这些优化的扫描技术和并行化的编码激发的结合，可以通过结合这些优化的扫描技术来实现神经元记录率的显着增长。通过该研究计划，我们将开发出这样的自适应多光子显微镜。我们的方法将利用一种混合体积扫描结构，将被动轴向多路复用和最佳采样的好处与同时多光束体积图案的激发相结合。这个快速且高度敏捷的显微镜平台，我们将其定为编码的栅格扫描混合动力（Crash）将与机器学习算法和高速反馈电路相结合，以适应响应观察到的实验相关的活动和运动，以适应扫描条件。我们的目标是开发一个显微镜，该显微镜更明智地扫描并自主优化资源的使用，以最大程度地利用记录的神经元的数量和SNR，以响应其运动和活动。首先（AIM 1），我们将开发和构建崩溃显微镜系统。我们的方法利用了一种新型的轴向多路复用方法，该方法将我们称为二元扩展轴向多路复用模块（BEAMM）。我们计划以三个阶段开发显微镜，从非自适应扫描BEAMM显微镜开始，然后转到自适应2D崩溃，最后移至完整的自适应3D崩溃。第二（AIM 2）我们将开发一种自适应硬件/软件解决方案，该解决方案使用在FPGA上运行的计算有效算法，恢复神经信号并调整崩溃显微镜的激发代码。首先，我们将优化用于体外应用程序的采集硬件和软件体系结构。随后，我们将扩展算法以应对运动和不相关活动等体内挑战。最后（AIM 3），我们将基准在脑切片和体内进行各种显微镜实现（扫描Beamm，2D崩溃和3D崩溃），然后研究每个实现在研究听觉皮层中声音的功能表示方面的应用。