VollSeg Napari Plugin

This project provides the napari plugin for VollSeg, a deep learning based 2D and 3D segmentation tool for irregular shaped cells. VollSeg has originally been developed (see papers) for the segmentation of densely packed membrane labelled cells in challenging images with low signal-to-noise ratios. The plugin allows to apply pretrained and custom trained models from within napari.

Installation & Usage

Install the plugin with pip install vollseg-napari or from within napari via Plugins > Install/Uninstall Package(s)…. If you want GPU-accelerated prediction, please read the more detailed installation instructions for VollSeg.

You can activate the plugin in napari via Plugins > VollSeg: VollSeg. Example images for testing are provided via File > Open Sample > VollSeg.

If you use this plugin for your research, please cite us.

GPU_Installation

This package is compatible with Python 3.6 - 3.9.

1. Please first install TensorFlow (TensorFlow 2) by following the official instructions. For GPU support, it is very important to install the specific versions of CUDA and cuDNN that are compatible with the respective version of TensorFlow. (If you need help and can use conda, take a look at this.)

2. VollSeg can then be installed with pip:

   • If you installed TensorFlow 2 (version 2.x.x):

     pip install vollseg
     

Examples

VollSeg comes with different options to combine CARE based denoising with UNET, StarDist and segmentation in a region of interest (ROI). We present some examples which are represent optimal combination of these different modes for segmenting different cell types. We summarize this in the table below:

Example Image	Description	Training Data	Trained Model	GT image	Optimal combination	Notebook Code	Model Prediction	Metrics
	Light sheet fused from four angles 3D single channel	Training Data ~320 GB	UNET model		UNET model, slice_merge = False	Colab Notebook
	Confocal microscopy 3D single channel 8 bit	Training Data	Denoising Model and StarDist Model		StarDist model + Denoising Model, dounet = False	Colab Notebook
	LaserScanningConfocalMicroscopy 2D single channel	Dataset	UNET Model		UNET model	Colab Notebook		No Metrics
	TIRF + MultiKymograph Fiji tool 2D single channel	Training Dataset	UNET Model		UNET model	Colab Notebook		No Metrics
	XRay of Lung 2D single channel	Training Dataset	UNET Model		UNET model	Colab Notebook
	LaserScanningConfocalMicroscopy 2D single channell	Test Dataset	Private		UNET model	Colab Notebook		No metrics
	LaserScanningConfocalMicroscopy 3D single channell	Test Dataset	Private		UNET model + StarDist model + ROI model	Colab Notebook

Troubleshooting & Support

• The image.sc forum is the best place to start getting help and support. Make sure to use the tag vollseg, since we are monitoring all questions with this tag.
• If you have technical questions or found a bug, feel free to open an issue.