Pavel Potocek

SEM 3D Reconstruction of Stained Bulk Samples using Landing Energy Variation and Deconvolution

Three dimensional reconstruction of bulk biological samples in SEM is currently achieved by Serial Sectioning or Serial Block Face (SBF) techniques. To image deeper layers of a sample these approaches differ mainly in the physical cutting method employed: diamond-knife microtome slicing or Focused Ion Beam (FIB) milling. During stack acquisition slicing steps are alternated with imaging steps to build the 3D volume. While these methods are proving very useful for a variety of Cell Biology, and in particular Neuroanatomy studies, they are still limited to a z-resolution of around 5nm for FIB techniques, 15-30nm for Serial Block Face (microtome) Slicing and around 2540nm for serial section techniques. Furthermore most of these techniques are destructive to the imaged samples. What TEM/STEM tomography can achieve for thin sections, being in principle non-destructive and providing high levels of resolution, is clearly missing in 3D SEM on bulk. Nonetheless, recent work [1] on computational modeling of beam-sample interactions in epoxyembedded osmium-stained samples shows remarkable signal linearity, offering well-proven deconvolution techniques as a tool for 3D reconstruction. Such models, significantly improved and extended with our own experimental work, indicate that the Point Spread Function (PSF) governing backscatter (BS) image formation remains well confined in the lateral directions for over 100 nm below the sample surface for commonly used landing energies (around 2keV).

How should a fixed budget of dwell time be spent in scanning electron microscopy to optimize image quality

In scanning electron microscopy, the achievable image quality is often limited by a maximum feasible acquisition time per dataset. Particularly with regard to three-dimensional or large field-of-view imaging, a compromise must be found between a high amount of shot noise, which leads to a low signal-to-noise ratio, and excessive acquisition times. Assuming a fixed acquisition time per frame, we compared three different strategies for algorithm-assisted image acquisition in scanning electron microscopy. We evaluated (1) raster scanning with a reduced dwell time per pixel followed by a state-of-the-art Denoising algorithm, (2) raster scanning with a decreased resolution in conjunction with a state-of-the-art Super Resolution algorithm, and (3) a sparse scanning approach where a fixed percentage of pixels is visited by the beam in combination with state-of-the-art inpainting algorithms. Additionally, we considered increased beam currents for each of the strategies. The experiments showed that sparse scanning using an appropriate reconstruction technique was superior to the other strategies.

High-Throughput Large Volume SEM Workflow using Sparse Scanning and In-painting Algorithms Inspired by Compressive Sensing

We are presenting a new extension to our Cell and Tissue/Neurobiology large volume imaging workflow, with the goal of increasing acquisition speed by more than five times. Instead of scanning dense square-grid frames, in the conventional way, our approach is here to explore the use of sparse scanning and inpainting techniques inspired by Compressive Sensing (CS) [1]. Sparse samples are obtained by pseudo-random scan patterns, and reconstruction algorithms are used to recover the dense volume data. The goal is to recover 3D datasets with minimum loss of information. Techniques inspired by CS gained wide attention over the last decade and are now being used in various applications where sensor bandwidth is a limiting factor. They have been recently explored for SEM and STEM applications [2][3]. In the context of nano-scale cell biology volume acquisition, we expect these techniques to ultimately increase the imaging throughput by nearly an order of magnitude. We will discuss additional advantages of this approach, such as the low-dose imaging of sensitive specimens, and the good compatibility with backscatter electron imaging. A key enabler of any sparse scan application to EM is the accurate control of scan locations. It has been shown in [2] and in our own experiments that precise positioning of the beam at the planned sampling locations is essential for a good CS reconstruction. We have developed advanced minimum-path scanning strategies to address this issue. The scanning technique is illustrated in Fig. 1, where the left two images show a conventional raster scan at 300ns dwell visiting a random set of points with the compressive sensing reconstruction obtained from such scan strategy. The right two images of Fig. 1 show an example minimum-path scan pattern and a much improved reconstruction result from images acquired with this second method. In future work we will compare pseudo-random sparse sampling, in combination with a reconstruction algorithm based on CS-inspired in-painting, to conventional grid sampling of the same effective dose, in combination with a de-noising algorithm, also based on CS. CS machine learning algorithms build patch-dictionaries, which are used as the building blocks for data representation [3]. During live acquisition runs, such dictionaries can be used to in-paint with high fidelity, the sparsely sampled datasets (Figure 2). We are implementing the new sparse scanning modules on SEM platforms, which also employ the Multi Energy Deconvolution SEM (MED-SEM) technology and Serial Block Face (SBF) imaging [4]. By incorporating CS, we will have an instrument allowing for both high-resolution isotropic imaging, and the fast acquisition of very large datasets (Figure 3).