Using the scientific Python stack to analyze Low Energy Electron Microscopy data

Low Energy Electron Microscopy (LEEM) is a specialized surface-sensitive microscopy technique utilizing electron with energies more than 1000 times lower than regular EM. This provides unique  measurement opportunities, but also challenges in the analysis of the data.

Here, we showcase how we utilize Numpy, Scipy, Dask and Scikit Learn and other parts of the scientific python stack to implement image analysis techniques, previously described for other microscopy techniques, but adapted to the specific challenges of LEEM [1,2]. Amongst others, we implement fast, parallelized, image (stack) registration and image stitching using Dask.

We show that the image registration algorithm is, in the best-case, accurate to the sub-pixel level

results and fast enough to enable registration of 500 images within 7 minutes on a regular desktop CPU, enabling per-pixel analysis of spectroscopic datasets, where energy is added to the images as a third dimension. Similarly, the stitching algorithm allows for the creation of 100Mpixel+ overview images from tiles with estimated positions.

In summary, we show that the use of the scientific python stack allows for easy adoption to specific peculiarities of different imaging techniques and even individual datasets. We anticipate the code from this work can be adapted to be applied to other forms of electron microscopy such as PEEM, scanning tunneling microscopy, and others, providing a open source, Python alternative to existing closed source / undisclosed implementations in often proprietary languages.

[1] T.A. de Jong et al., Quantitative analysis of spectroscopic Low Energy Electron Microscopy data: High-dynamic range imaging, drift correction and cluster analysis, Ultramicroscopy, Volume 213, 2020, https://doi.org/10.1016/j.ultramic.2019.112913.

[2] https://github.com/TAdeJong/LEEM-analysis