VoidSeg: Leveraging Self-Supervised Denoising for Image Segmentation

Mangal Prakash, Tim-Oliver Buchholz, Manan Lalit, Pavel Tomancak, Florian Jug¹, Alexander Krull¹

¹Joint supervision

Deep learning (DL) has arguably emerged as the method of choice for the detection and segmentation of biological structures in microscopy images. However, DL typically needs copious amounts of annotated training data that is for biomedical projects typically not available and excessively expensive to generate. Additionally, tasks become harder in the presence of noise, requiring even more high-quality training data. Hence, we propose to use denoising networks to improve the performance of other DL-based image segmentation methods. More specifically, we present ideas on how state-of-the-art self-supervised CARE networks can improve cell/nuclei segmentation in microscopy data. Using two state-of-the-art baseline methods, U-Net and StarDist, we show that our ideas consistently improve the quality of resulting segmentations, especially when only limited training data for noisy micrographs are available.

Paper: https://arxiv.org/abs/1911.12239

Installation

This implementation requires Tensorflow. We have tested VoidSeg on Linux Mint 18.1 Serena (Ubuntu 16.04), CUDA Version 9.0.176.4, NVIDIA-SMI Driver Version 396.26 using Python 3.6 and tensorflow-gpu 1.12.0.

From Scratch

We recommend using miniconda. If you do not yet have a strong opinion, just use it too!

After installing Miniconda, the following lines are likely the easiest way to get Tensorflow and CuDNN installed on your machine (Note: Macs are not supported, and if you sit on a Windows machine all this might also require some modifications.):

$ conda create -n 'voidseg' python=3.6
$ source activate voidseg
$ conda install tensorflow-gpu=1.12 keras=2.2.4
$ pip install jupyter

Note: It is very important that the version of keras be 2.2.4 or 2.2.5, hence the explicit installation above. Once this is done (or you had tensorflow et al. installed already), you can install VoidSeg with the following commands.

Install from Sources

This option is ideal if you want to edit the code. Clone the repository:

$ git clone https://github.com/juglab/VoidSeg.git

Change into its directory and install it:

$ cd VoidSeg
$ pip install -e .

You are now ready to run our notebooks.

Datasets

We use datasets from Data Science Bowl 2018 and Broad Bioimage Benchmark Collection. For more details on the datasets, please see the paper http://arxiv.org/abs/1911.12239.

Network Architectures

We show all our results for two different network architectures - 3 class U-Net (https://arxiv.org/pdf/1802.07465) and StarDist (https://arxiv.org/pdf/1806.03535).

Getting Started

Have a look at our jupyter notebooks:

• [U-Net baseline (U-Net_Baseline.ipynb)
• U-Net Sequential (U-Net_Sequential.ipynb)
• U-Net finetune (U-Net_Finetune.ipynb)
• U-Net finetune sequential (U-Net_Finetune_Sequential.ipynb)
• StarDist baseline (StarDist_Baseline.ipynb)
• StarDist Sequential (StarDist_Sequential.ipynb)

How to cite:

@article{prakash2019leveraging,
  title={Leveraging Self-supervised Denoising for Image Segmentation},
  author={Prakash, Mangal and Buchholz, Tim-Oliver and Lalit, Manan and Tomancak, Pavel and Jug, Florian and Krull, Alexander},
  journal={arXiv preprint arXiv:1911.12239},
  year={2019}
}

Cluster Execution

Please note that this repository contains the jupyter notebooks version of VoidSeg http://arxiv.org/abs/1911.12239. If you want to run many experiments at the same time, cluster scripts are preferable which can be found at https://github.com/juglab/VoidSeg_cluster. The cluster version is specifically adapted for MPI-CBG cluster. For your own cluster, please consider making appropriate changes.