In this tutorial, we show how to import FIB-SEM tomography data sets. While it is often difficult to distinguish between fore- and background in FIB-SEM tomography, it has some significant advantages. It provides a very good resolution resolving the detailed microstructure even for the binder in a cathode. Also, the crystal orientation and anisotropic properties could be considered by using an electron backscatter diffraction detector during the FIB-SEM measurement, which is not possible with a CT-scan.

We use modern AI-technology for the segmentation, teaching a neural network not only to distinguish between the material phases but also between fore- and background in the FIB-SEM images. For this we use two image stacks of the same cathode, recorded by different detectors through different channels. In this way, the neural network obtains the needed information for the segmentation.

In detail, this tutorial teaches you to do the following:

• Import two FIB-SEM measurements of a cathode with ImportGeo-Vol,
• Apply a post-processing on both images,
• Train an AI-model using both images as input to segment them,
• Post-process the segmented structure to cleanse it from remaining artifacts.