🟢 Void

easymode segment void

Void refers to regions of tomograms that are outside the sample (e.g. above or below the lamella), contain poor-quality data (e.g. thick or poorly reconstructed areas), or are dominated by artefacts (e.g. dense ice contamination). One application of segmenting void is to measure or approximate the distance from a particle to the edges of a lamella; sampling the void segmentation output at the location of picked particles offers a practical way of estimating this distance.

A second use of the void model is to approximate the quality of entire tomograms, as demonstrated in this preprint. Simply running easymode segment void on all tomograms in a dataset, tabulating the average void output value for each tomogram, and sorting low to high is a practical way to filter good from worse tomograms.

The model is relatively fast, as it works at 20 Å/px; but input is still expected to be at 10 Å/px and is automatically downsampled.

Example output

Example of easymode segment void output overlaid on a tomogram from EMPIAR-11899 (FIB-milled D. discoideum), a dataset which was not used to train this model.