membrain-seg 0.0.1

membrane segmentation in 3D for cryo-ET

MemBrain-Seg

Membrain-Seg is a Python project developed by teamtomo for membrane segmentation in 3D for cryo-electron tomography (cryo-ET). This tool aims to provide researchers with an efficient and reliable method for segmenting membranes in 3D microscopic images. Membrain-Seg is currently under early development, so we may make breaking changes between releases.

Membrain-Seg is currently under early development, so we may make breaking changes between releases.

Overview

MemBrain-seg is a practical tool for membrane segmentation in cryo-electron tomograms. It's built on the U-Net architecture and makes use of a pre-trained model for efficient performance. The U-Net architecture and training parameters are largely inspired by nnUNet¹.

Our current best model is available for download here. Please let us know how it works for you. If the given model does not work properly, you may want to try one of our previous versions:

Other (older) model versions:

• v9 -- best model until 10th Aug 2023
• v9b -- model for non-denoised data until 10th Aug 2023

If you wish, you can also train a new model using your own data, or combine it with our (soon to come!) publicly-available dataset.

To enhance segmentation, MemBrain-seg includes preprocessing functions. These help to adjust your tomograms so they're similar to the data our network was trained on, making the process smoother and more efficient.

Explore MemBrain-seg, use it for your needs, and let us know how it works for you!

Preliminary documentation is available, but far from perfect. Please let us know if you encounter any issues, and we are more than happy to help (and get feedback what does not work yet).

[1] Isensee, F., Jaeger, P.F., Kohl, S.A.A., Petersen, J., Maier-Hein, K.H., 2021. nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nature Methods 18, 203-211. https://doi.org/10.1038/s41592-020-01008-z

Installation

For detailed installation instructions, please look here.

Features

Segmentation

Segmenting the membranes in your tomograms is the main feature of this repository. Please find more detailed instructions here.

Preprocessing

Currently, we provide the following two preprocessing options:

• Pixel size matching: Rescale your tomogram to match the training pixel sizes
• Fourier amplitude matching: Scale Fourier components to match the "style" of different tomograms
• Deconvolution: denoises the tomogram by applying the deconvolution filter from Warp

For more information, see the Preprocessing subsection.

Model training

It is also possible to use this package to train your own model. Instructions can be found here.

Patch annotations

In case you would like to train a model that works better for your tomograms, it may be beneficial to add some more patches from your tomograms to the training dataset. Recommendations on how to to this can be found here.