MemBrain v2

MemBrain v2 is a deep learning-enabled program aimed at the efficient analysis of membranes in cryo-electron tomography (cryo-ET). It consists of the three main modules „MemBrain-seg”, „MemBrain-pick”, and „MemBrain-stats”.
MemBrain-seg is a generalizable membrane segmentation network that can be used out-of-the-box for a wide variety of tomograms. We achieve this by combining a well-curated and diverse training dataset with strong data augmentations, as well as a custom loss function that aims to capture the correct membrane topology.
MemBrain-pick is our interactive framework for localizing proteins on membrane surfaces. To this end, we convert membrane segmentations to mesh representations and perform learning directly on these meshes, leading to improved data efficiency compared to common voxel-based approaches. MemBrain-pick works interactively with membrane surface annotation tools to enable going back-and-forth smoothly between annotation and training processes.
MemBrain-stats leverages the outputs of MemBrain-pick and MemBrain-seg to generate several membrane statistics, like protein concentrations, geodesic nearest neighbor analysis, and Ripley’s statistics.