Embedding-based Instance Segmentation in Microscopy

Table of Contents

• Introduction
• Dependencies
• Getting Started
• Datasets
• Training & Inference on your data
• Animated figures
• Contributing
• Issues
• Citation
• Acknowledgements

Introduction

This repository hosts the version of the code used for the preprint Embedding-based Instance Segmentation of Microscopy Images.

We refer to the techniques elaborated in the publication, here as EmbedSeg. EmbedSeg is a method to perform instance-segmentation of objects in microscopy images, based on the ideas by Neven et al, 2019.

With EmbedSeg, we obtain state-of-the-art results on multiple real-world microscopy datasets. EmbedSeg has a small enough memory footprint (between 0.7 to about 3 GB) to allow network training on virtually all CUDA enabled hardware, including laptops.

Dependencies

We have tested this implementation using pytorch version 1.10.0 and cudatoolkit version 10.2 on a linux OS machine. One could execute these lines of code to run this branch:

conda create -n EmbedSegEnv python==3.7
conda activate EmbedSegEnv
conda install pytorch torchvision cudatoolkit=10.2 -c pytorch
git clone https://github.com/juglab/EmbedSeg.git
cd EmbedSeg
pip install -e .

Getting Started

Look in the examples directory, and try out one of the provided notebooks. Please make sure to select Kernel > Change kernel to EmbedSegEnv.

Datasets

3D datasets are available as release assets here.

Training and Inference on your data

*.tif-type images and the corresponding masks should be respectively present under images and masks, under directories train, val and test. (In order to prepare such instance masks, one could use the Fiji plugin Labkit as suggested here). The following would be a desired structure as to how data should be prepared.

$data_dir
└───$project-name
    |───train
        └───images
            └───X0.tif
            └───...
            └───Xn.tif
        └───masks
            └───Y0.tif
            └───...
            └───Yn.tif
    |───val
        └───images
            └───...
        └───masks
            └───...
    |───test
        └───images
            └───...
        └───masks
            └───...

Animated Figures

Contributing

Contributions are very welcome. Tests can be run with tox.

Issues

If you encounter any problems, please file an issue along with a detailed description.

Citation

If you find our work useful in your research, please consider citing:

@misc{lalit2021embeddingbased,
      title={Embedding-based Instance Segmentation of Microscopy Images}, 
      author={Manan Lalit and Pavel Tomancak and Florian Jug},
      year={2021},
      eprint={2101.10033},
      archivePrefix={arXiv},
      primaryClass={eess.IV}
}

Acknowledgements

The authors would like to thank the Scientific Computing Facility at MPI-CBG, thank Matthias Arzt, Joran Deschamps and Nuno Pimpao Martins for feedback and testing. Alf Honigmann and Anna Goncharova provided the Mouse-Organoid-Cells-CBG data and annotations. Jacqueline Tabler and Diana Afonso provided the Mouse-Skull-Nuclei-CBG dataset and annotations. This work was supported by the German Federal Ministry of Research and Education (BMBF) under the codes 031L0102 (de.NBI) and 01IS18026C (ScaDS2), and the German Research Foundation (DFG) under the code JU3110/1-1(FiSS) and TO563/8-1 (FiSS). P.T. was supported by the European Regional Development Fund in the IT4Innovations national supercomputing center, project number CZ.02.1.01/0.0/0.0/16013/0001791 within the Program Research, Development and Education.