Aberration correction for real-time measurements in adaptive confocal microscopy

生物细胞或组织的三维检查需要具有高空间和时间分辨率的自动化显微技术。双光子显微镜可以提供所需的高分辨率，但作为一种基于点的技术，它需要进行三维扫描，导致时间分辨率有限。自适应光学元件原则上有可能打破体积扫描的这一限制。然而，自适应组件的使用伴随着像差的引入和调整，这对于双光子显微镜激发来说可能是一个问题。如前两个资助期所示，这些和样本引起的像差可以用自适应透镜来补偿。此外，通过使用用于轴向扫描的自适应透镜和用于横向扫描的自适应棱镜，可以实现无需机械运动的快速3D扫描。在该项目的第三阶段，将实施一种完全自适应的智能双光子显微镜，它将使用自适应显微镜透镜和自适应棱镜来实现快速扫描以及动态校正像差。为此，前几个项目阶段的自适应组件将首先用超短脉冲激光进行表征，如有必要，还将适应更高的能量密度。然后，用于像差校正和轴向扫描的自适应双驱动透镜将被集成到显微镜物镜中。这与自适应棱镜相结合，能够校正样品和系统引起的像差(散焦、散光、彗差和球差)，以及无机械位移地对样品进行3D扫描。通过使用压电陶瓷作为驱动器材料，在自适应元件的行为中引入了滞后。此外，在像差校正3D扫描系统中倍增致动器使得致动比单个自适应镜头复杂得多。对于这种高度复杂的多输入系统，控制器的设计变得越来越复杂，最优控制需要越来越多的迭代。作为经典控制工程的替代，将使用机器学习的方法，如神经网络和强化学习来控制整个系统，以实现自动化显微镜。由于新的自适应物镜和改进的控制，靶向双光子显微镜提供了小型化、自动化和开发紧凑而坚固的光学系统的可能性。首先，将采用测量方法来研究甲状腺激素对斑马鱼胚胎的影响。这一结果将对自适应光学元件和自动化智能显微镜的进一步发展做出重要贡献。

Automated microscopic techniques with high spatial and temporal resolutions are required for the three-dimensional examination of biological cells or tissues. Two-photon microscopy can provide the required high resolution, but as a point-based technique it requires scans in three dimensions, resulting in limited time resolution. Adaptive optical elements in principle have the potential to break this limitation for volumetric scans. However, the use of adaptive components is accompanied by the introduction and tuning of aberrations, which can be a problem for two-photon microscopy excitation. As presented in the first two funding periods, these and sample-induced aberrations can be compensated for with adaptive lenses. Furthermore, by using adaptive lenses for an axial scan and the adaptive prism for a lateral scan, the implementation of fast 3D scans without mechanical motion is possible. In the third phase of the project, a fully adaptive, smart two-photon microscope will be implemented, which will use an adaptive microscope lens and the adaptive prism to enable fast scans, as well as dynamic correction of aberrations. For this purpose, the adaptive components from the previous project phases will first be characterized with an ultrashort pulse laser and, if necessary, adapted to the higher energy density. Then, an adaptive bi-actuator lens for aberration correction and axial scanning will be integrated into a microscope objective. This, in combination with the adaptive prism, enables correction of sample- and system-induced aberrations (defocus, astigmatism, coma and spherical aberration) as well as 3D scanning of the sample without mechanical displacement. By using piezoceramics as actuator material, hysteresis is introduced into the behaviour of the adaptive components. In addition, multiplying the actuators in an aberration-correcting 3D scan system makes the actuation much more complex than for a single adaptive lens. For such highly complex multiple-input systems, the design of a controller becomes increasingly complex and optimal control requires more and more iterations. As an alternative to classical control engineering, methods from machine learning, such as neural networks and reinforcement learning will be used to control the overall system in order to realize an automated microscope. The targeted two-photon microscope offers the possibility of miniaturization, automation and the development of compact and robust optical systems due to the new adaptive objective and the improved control. For first demonstration measurements will be applied to study the effects of goitrogens in zebrafish embryos. The result will be an important contribution to the further development of adaptive optical elements and automated smart microscopy.