About the alignment and 3D reconstruction of sparse cryo-scanning transmission electron tomography datasets

In electron microscopy, sparse imaging consists in the collection of a limited subset of the image pixels, which can be used to reduce electron beam damage. Scanning transmission electron microscopy (STEM) is particularly adapted to sparse imaging owing to the scanning nature of the method, scan patterns can be designed where fewer sample locations are targeted. However, since some of the pixels are not scanned, there is an inherent loss of information. Several algorithms were developed to reconstruct missing pixels with high fidelity. Whereas sparse imaging and missing pixel reconstruction in 2D experiments are mature methods, the application of sparse imaging in 3D scanning transmission electron tomography (STET) is rare and still under development. The main difficulty encountered in tomography studies is the tilt-series alignment, which must be accurate to ensure high-quality 3D reconstruction. Because sparse images contain only a certain portion of the original information, the images constituting sparse tilt-series might not share enough mutual information to guarantee an accurate alignment, even after missing pixel reconstruction. This work presents for the first time a thorough analysis of the fiducial alignment and reconstruction of sparse (cryo-)STET tilt-series. Furthermore, the limits of sparse imaging are explored to estimate the minimum amount of information required to obtain good-quality 3D reconstructions. The use of a cryo-fixed biological sample is motivated by the fact that cryo-samples are typical highly beam-sensitive samples, and that the intricate nature and structure complexity of biological samples place them among the most difficult to reconstruct with high details.