SlideRunner - A Tool for Massive Cell Annotations in Whole Slide Images


License
GPL-2.0+
Install
pip install SlideRunner==2.2.0

Documentation

Logos

SlideRunner

DOI:10.1007/978-3-662-56537-7_81

SlideRunner is a tool for massive cell annotations in whole slide images.

It has been created in close cooperation between the Pattern Recognition Lab, Friedrich-Alexander-Universität Erlangen-Nürnberg and the Institute of Veterenary Pathology, Freie Universität Berlin. Development is continued now at Technische Hochschule Ingolstadt.

If you use the software for research purposes, please cite our paper:

M. Aubreville, C. Bertram, R. Klopfleisch and A. Maier (2018) SlideRunner - A Tool for Massive Cell Annotations in Whole Slide Images. In: Bildverarbeitung für die Medizin 2018. Springer Vieweg, Berlin, Heidelberg, 2018. pp. 309-314. link arXiv:1802.02347

Please find the authors webpage at: https://imi.thi.de

Version 2.0.0

With so many new features, it is time to declare Version 2. While we initially wanted to declare the version on this years BVM, it has all been delayed a bit. But for a good reason: Following requests by our pathologists, SlideRunner now has support for a much broader range of image formats.

List of new features:

  • New formats:
    • DICOM supprt (including DICOM WSI, through pydicom, new in 1.99beta4)
    • CellVizio MKT (Confocal Laser Endomicroscopy, new in 1.99beta4)
    • TIFF stacks (previously was only for 2D images, new in 1.99beta6)
    • NII files (thanks to nibabel package, new in 2.0.0)
  • Synchronization support with EXACT
  • Much enhanced polygon annotation support, including modifying and simplification
  • Support for z Stacks as well as time series
  • Support for non-clickable annotations (to aid some annotation tasks) (new in 2.0.0)
  • New plugins, e.g.:
  • and uncountable small fixes and improvements

Download and Installation

For Windows 10 and Mac OS X, we provide a binary file, that you can download here:

Operating System Version Download link
Windows 10 (google drive) V. 2.0.0 link
Mac OS X (10.15) (google drive) V. 2.0.0 link

Updates

Starting V. 1.31.0, SlideRunner has support for the DICOM WSI image format (thanks, pydicom team, for your support!). Use wsi2dcm to convert images into dicom format.

Starting V. 1.25.0, SlideRunner features a magic wand tool.

Watch the video

Installation - Source

SlideRunner is written in Python 3, so you will need a Python 3 distribution like e.g. Anaconda (https://www.anaconda.com/download/) to run it. Further, you need to install OpenSlide (http://openslide.org/download/).

Install using PIP

After OpenSlide is installed, we provide a convenient installation by using pip. On Linux or Mac, simply run:

sudo pip install -U SlideRunner

On windows, pip should install without sudo (untested):

pip install -U SlideRunner

Installation from repository

You need to clone this repository:

git clone https://github.com/maubreville/SlideRunner

In order to use the automated installation process, you need to have setuptools installed.

pip install -U setuptools

The installation procedure is then as easy as:

cd SlideRunner

python setup.py install

To run, the following libraries and their dependencies will be installed:

Library version link
PyQT5 >= 5.6.0 https://pyqt.sourceforge.net/
numpy >= 1.13 https://www.numpy.org
cv2 (OpenCV3) >= 3.1.0 https://opencv.org
sqlite3 >= 2.6.0 https://www.sqlite.org
openslide >= 1.1.1 https://www.openslide.org

Screenshots

SlideRunner Screenshot

Usage

SlideRunner features a number of annotation modes.

Annotation modes

View/Select

This mode is meant for browsing/screening the image, without the intention to add annotations.

Object center annotation

In this mode, annotations of object centers (e.g. cells) can be made with one click only. Select the class to annotate by clicking the button next to it in the class list or press the key corresponding to the number in the list of annotations (e.g. key 1 for the first class entry). Objects annotated in this mode are displayed with a circle having a diameter of 50 pixels in full resolution, and scaled appropriately when not viewed in full resolution.

Outline annotation (rectangular) This mode provides a rectangular annotation tool, that can be used to annotate rectangular patches on the image.

Outline annotation (circle) (new in ver 1.8.0) This mode provides a circle annotation tool, that can be used to annotate circular patches on the image.

Outline annotation (polygon) This mode creates a polygon, i.e. a connected line of points corresponding to a single annotation object. This can be handy for cellular structures or tissue types.

Annotation of important position Important positions are annotations shown with a cross in a circle. The size of this annotation does not change depending on the zooming level. An important position does not have a class attached to it.

News

New version 1.22.0 has a much improved plugin interface. Amongst the many improvements are annotation support from the plugin, as well as the ability to copy annotations from the plugin to the database, even annotation groups.

Databases

The MITOS2012 [http://ludo17.free.fr/mitos_2012/], MITOS-ATYPIA-2014 [https://mitos-atypia-14.grand-challenge.org/] and TUPAC16 [http://tupac.tue-image.nl] data sets are provided as SlideRunner database in the repository. Please download the original images (*.bmp, *.tif) in order to use them.

Tools

Mingrui Jiang has written a tool to extract patches around ROIS. Find it here: https://github.com/mingrui/SlideRunner-ROI-Patch-Maker

Demo workflows

To get started how to use SlideRunners databases, have a look into this notebook.

Plugins

There are currently three plug-ins available:

  • OTSU threshold: A simple OTSU threshold plugin. Positive pixels are shown in green color.
  • PPC: Positive Pixel Count, as in: Olson, Allen H. "Image analysis using the Aperio ScanScope." Technical manual. Aperio Technologies Inc (2006).
  • Macenko normalization: Image-based normalization using the method of Macenko et al. link

Database structure

The major entity of our database model is the annotation. Each annotation can have multiple coordinates, with their respective x and y coordinates, and the order they were drawn (for polygons). Further, each annotation has a multitude of labels that were given by one person each and are belonging to one class, respectively.

DB Structure

Troubleshotting

Programm won't start, error message: Missing or nonfunctional package openslide: Couldn't locate OpenSlide dylib. Is OpenSlide installed?

Unfortunately, the OpenSlide binary libraries can't be fetched using pip. You need to install OpenSlide libraries (and dependencies). See http://openslide.org/download/ for more.