This paper summarizes the achievements in the 3D reconstruction of microscopic specimens through the tomographic algorithm applied to a set of projection\images obtained in the SEM. This approach is complementary to the serial-sectioning and the slice-and-view methods presently implemented in the SEM platform, and benefits from a compressed sensing approach to refine the reconstruction from a limited number of projections.

A Si-based electron detector has been specifically developed for the purpose of operating the microscope in the scanning-transmission imaging mode for the tomographic application, and the detection strategy has been tailored in order to maintain the projection requirement over the large tilt range, a requirement needed for the reconstruction workflow [1]. Either inorganic or biological samples have been investigated to demonstrate the adaptability of the compressed sensing refinement to the specimen characteristics: the former system is formed by cobalt particles within a carbon tube and the latter features collagen fibrils in dermal tissue.

Figure 1 shows a STEM image from the tilt series of Co nanoparticles inside a carbon tube. The contrast in the STEM image is determined by local specimen thickness and composition, the Co particles being visible with the highest contrast. The reconstruction has been obtained starting from 53 projections at 2°steps, and refined through compressive sensing with regularization parameters emphasizing sparsity in the gradient domain.

Figure 2 highlights the complex structure of the dermal tissue as revealed by the STEM imaging mode in the SEM. Cellular membranes and circular structures are mixed with bundles of collagen fibrils. The bundles were truncated by the fine sectioning and their disposition is clearly visible. A small bundle of collagen was selected as the region of interest for the tomographic reconstruction. Starting from 91 projections at 40.000× magnification and ranging between -50° to +40°. Compressed sensing was adapted to deal with the inherent complexity of biological images, and the final tomogram turned out to preserve the finest details of the fibrils. The known periodical striation (about 60 nm periodicity) of collagen was indeed recovered with adequate spatial resolution.

The proposed system exploits the capability of the STEM imaging mode, which can be applied for both biological and physical science for the 3D analysis of volumes below 100 mm³. The limit in resolution is posed by the probe size of the microscope, specimen composition and thickness, and the number of projections that can be acquired without significant beam damage of the sample. Compressed sensing is effective in improving the quality of the reconstruction. Owing to the flexibility of the SEM platform, cryo-preservation of the specimen as well as site-selective sample preparation could be pursued within the proposed approach for tomography in the SEM.

[1]  M Ferroni et al., Journal of Physics: Conference Series 644 (2015): 012012

Figures:

(Left) STEM image from the tilt series of a C tube with Co particles. Front- and side- views of the reconstructed tomogram (Center and Right, respectively).

STEM images and reconstruction of a biological sample: (Up - Left) STEM BF panoramic view of cellular structures and collagen fibers in human derma. (Up - Right) STEM DF image of the fibers, showing a periodical striation along the fibers. (Down - Left) STEM DF image from the tilt series of the collagen bundle selected for tomographic reconstruction, (Down - Right) Tomogram of the reconstructed collagen bundle.

« Back to The 16th European Microscopy Congress 2016

EMC Abstracts - https://emc-proceedings.com/abstract/compressed-sensing-tomography-of-inorganic-and-biological-samples-in-the-scanning-electron-microscope-operated-in-the-transmission-mode/