Periodic Reporting for period 1 - Nanocubic (Nanoscale Isotropic 3D Resolution using Omni-view Structured Light Sheet Microscopy)
Reporting period: 2022-09-01 to 2025-02-28
A major challenge in the life sciences is to gain access across scales, from the molecular building blocks, via cells, to organization and dynamics at the composite tissue level. Fluorescence microscopy as the primary biological imaging tool has partly addressed these needs with three major revolutions. The first is the use of 2D sheets of light for illumination so as to only illuminate the part of the sample volume that is image on the camera, which minimizes light induced damage to the cells. The second is a suite of super-resolution techniques where single molecule imaging and patterned illumination are used to image down to and below the 100 nm scale. The third is the use of adaptive optical techniques for imaging heterogeneous samples such as cells or tissues, or for manipulating the illumination and/or imaging light field to a desired shape.
The next frontier in microscopy lies in overcoming trade-offs between imaged volume, resolution, and sample longevity by combining and extending these three innovations. The goal of the project is to create a microscopy platform in which a sequence of patterned light sheets is applied to the sample, where the 2D light sheet pattern, as well as the position and orientation of the 2D light sheet in 3D space is changed in the sequence. This creates a very large diversity of 3D illumination patterns, and this is what enables isotropic super-resolution across large volumes. A second crucial ingredient is the use of computational reconstruction techniques to render the final volumetric image. Taking the heterogeneity of the imaging conditions across the cellular or tissue volume into account is the major goal of algorithm development.
The next frontier in microscopy lies in overcoming trade-offs between imaged volume, resolution, and sample longevity by combining and extending these three innovations. The goal of the project is to create a microscopy platform in which a sequence of patterned light sheets is applied to the sample, where the 2D light sheet pattern, as well as the position and orientation of the 2D light sheet in 3D space is changed in the sequence. This creates a very large diversity of 3D illumination patterns, and this is what enables isotropic super-resolution across large volumes. A second crucial ingredient is the use of computational reconstruction techniques to render the final volumetric image. Taking the heterogeneity of the imaging conditions across the cellular or tissue volume into account is the major goal of algorithm development.
The project relies on the construction in the lab of an optical setup that enables the precise projection of the envisioned patterned light sheets into the sample, and that offers the intricate control over the position and orientation of the patterned light sheet in all three spatial dimensions. This requires extensive calibration and alignment, systematic use of raytracing models for tolerance analyses, and adaptive optics and control loops for system control and robustness. Once created the platform will be used to demonstrate isotropic super-resolution over large imaging volumes. Specifically, a first goal is to achieve isotropic super-resolution approaching 100 nanometer over a large volume of about 100 micron in all three spatial directions, enough to contain several cells and be able to visualize sub-cellular structures. A next goal is to further zoom in at the single digit nanometer resolution scale using single molecule imaging techniques enhanced with 3D patterned illumination information, for providing access to the structure of molecular machinery inside the cell. On the computational side the biggest challenge is to invent computational reconstruction algorithms that are purely based on physical principles and information extracted from the imaged data itself. In this way any arbitrariness or black box approaches are avoided, providing the most objective assessment of the structural information provided by the data. Specific attention is given to the limiting role of noise and propagation of noise effects through the algorithmic chain of image processing operations, and to adapting the algorithm to the level of noise and aberrations in image formation that can now vary with space and time.
The first result of the project pertains to the assessment of image resolution using a novel image splitting procedure, where two sub-images are created from a single image acquisition, and subsequently checked for consistency across length scales. This correlation measure appears to be a suitable measure for the information content of the original image. A similar procedure is envisioned for the final volumetric reconstruction derived from the whole set of acquired images for different light sheet patterns, positions and orientations. A second result is an analysis of the impact of noise on the input image on the outcome of a so-called image deconvolution algorithm, the so-called Richardson-Lucy algorithm. It appears that the noisy input necessarily implies ill convergence of the algorithm. This result points to new ways and new algorithms to mitigate or even overcome this computational breakdown.