Cytomine ULiège Research & Development

@ Montefiore Institute (Dept. EE & CS), University of Liège, Belgium

At the Montefiore Institute (University of Liège), we are software developers and computer science researchers developing machine/deep learning algorithms and big data software modules to make life easier for multidisciplinary teams who have to deal with very large imaging data. Our software tools enable remote collaboration through sharing of images, algorithms, and quantitative results over the web. Our developments are used worldwide in biomedical research and beyond (e.g. digital collections, geology, industrial quality control, …).

We continuously contribute to the Cytomine open-source project initiated by our research unit in 2010. The official “Community Edition (CE)" is validated and maintained by the Cytomine open company. Our Cytomine ULiège version includes base features plus experimental features driven by our research projects and collaborations. Once validated, these developments are integrated into Cytomine CE.

Ongoing research & results

The Cytomine open-source software (with a permissive licence) is a rich internet application using modern web and container technologies, geospatial and noSQL databases, and machine/deep learning to foster active and distributed collaboration and ease large-scale image exploitation. The software can be used remotely, using a web browser, e.g. by life scientists to help them better evaluate drug treatments or understand biological processes using various imaging modalities (including whole-slide tissue images), by pathologists to share and ease their diagnosis (digital pathology), by teachers and students for image-based training purposes (e.g. histology courses), and by computer and data scientists willing to integrate and apply their image recognition algorithms on large imaging datasets using REST/Python/Java APIs.

In our group, our main contributions are related to (see our latest R&D overview slides):

  • Development of novel algorithms, efficient data structures, back-end modules, and web user interfaces for multimodal imaging data
  • Development of efficient workflows and novel algorithms (deep learning and tree-based machine learning) for content-based image retrieval, object detection, recognition, and segmentation in very large (multimodal) images
  • Reproducible benchmarking of algorithms on realistic datasets
  • Applications in biomedical domain with a specific focus on digital pathology and multimodal microscopy datasets
  • Applications in any other domain with large sets of large images (geology, digital collections, astronomy, industrial control, …).
  • Development of novel algorithms for user behavior analytics
  • Software architecture, distributed software deployment, reproducibility, open science, …

In addition to our open-access scientific publications, our results are first distributed through the Cytomine ULiège open-source software repository and later (after code review and validation) through the official Community edition (CE) repository.

Content-based image retrieval

Content-based image retrieval

Algorithms for large-scale, interactive, explainable content-based image retrieval.

Reproducibility and interoperability

Reproducibility and interoperability

Web framework and container-based execution architecture for reproducible deployment and benchmarking of image analysis workflows (Rubens et al., Cell Patterns, 2020).

Algorithms and web user interfaces for multimodal datasets

Algorithms and web user interfaces for multimodal datasets

User interfaces and data models for multiple modalities beyond histology incl. MALDI-IMS/multispectral imaging (Rubens et al., Proteomics Clin Appl, 2019).

Tools for user behavior analytics

Tools for user behavior analytics

Data analysis modules to study user behavior in educational settings (Fries et al., Anatomical Sciences Education 2023; Vanhee et al., J. Pathology Informatics 2019).

Anatomical landmark detection algorithms

Anatomical landmark detection algorithms

Machine learning algorithms for morphometric change measurements e.g. in developmental and toxicological studies. See e.g. Kumar et al. (Biomolecules 2023, ECCV 2022 Bioimage Computing) and Vandaele et al. (Nature Scientific Reports, 2018).

Image classification and object recognition algorithms

Image classification and object recognition algorithms

Deep/Machine learning algorithms for diagnostic or for phenotyping, e.g. in developmental and toxicological studies. See e.g. Marée et al. (PRL 2016; ISBI 2016); Jeanray et al. (PLOS One 2015); Mormont et al. (CVPRW 2018, IEEE JBHI 2020).

Cell Counting algorithms

Cell Counting algorithms

Benchmarking of algorithms for cell counting in specific regions of interests within tissues.

Workflows for sorting various types of cells

Workflows for sorting various types of cells

Image analysis workflows e.g. to detect abnormal cells for early cytological diagnosis. See e.g. Delga et al., Acta Cytologica 2014; Mormont et al., 2016.

Image segmentation techniques

Image segmentation techniques

Self-training algorithms for exploiting sparse annotations (Mormont et al., ECCV AIMIA 2022) and algorithms for the quantification and delineation of tissue areas in whole tissue slides (Marée et al. ISBI 2014).

Bigpicture research project

Since 2021, we are involved in the BigPicture EU IMI project (2021-2027) to establish the biggest database of pathology images to accelerate the development of artificial intelligence in medicine. We extend Cytomine used as the main software platform to visualize and annotate BigPicture WSI repository.

Bigpicture research project
COMULIS EU COST Action: Kick-off

We are involved in the COMULIS EU-funded COST network for (Correlated Multimodal Imaging in Life Sciences).

COMULIS EU COST Action: Kick-off

Software

(Installation, source code, documentation)

Cytomine can be installed on high-end servers for large-scale studies but also on laptops for small-scale works (then loosing collaboration functionalities).

We release our research results mainly through the Cytomine ULiège R&D repository, which is based on the “Community Edition (CE)” available on the official “Community Edition” repository.

To install and use the ULiège R&D version you can follow experimental version installation instructions and its full documentation wiki (it includes guides for administrators, developers, and a Cytomine guide for end-users (to be updated)). To install the “Community Edition”, please follow official “Community Edition” installation instructions.

Access to our R&D server can be obtained in the Cytomine user guide (default demo accounts). Access to an official, stable, version can be requested to Cytomine cooperative.

We kindly ask you to cite Cytomine website url (www.cytomine.org) and our main publication (Marée et al., Bioinformatics 2016) when you use our software in your own work.

Get Started View Documentation

See here for access to the official, stable version maintained by the Cytomine cooperative.

ULiège Cytomine team members

Current team and past contributors

Researchers

Avatar

Raphaël Marée

Montefiore Institute

Head of Cytomine ULiège R&D

Avatar

Navdeep Kumar

Montefiore Institute

PhD Student in machine learning (deep learning, domain adaptation)

Avatar

Axelle Schyns

Montefiore Institute

PhD Student in large-scale CBIR

Avatar

Maxime Amodei

Montefiore Institute

PhD Student in multimodal machine learning

Avatar

Ba Thien Le

Montefiore Institute

Research software engineer

Alumni

Ulysse Rubens

Research software engineer (2017 - 2022)

Laurent Vanhee

Education software modules contributor (2018)

Romain Mormont

Machine learning researcher (2016 - 2022)

Gino Michiels

Software contributor (2018)

Renaud Hoyoux

Software developer (2014-2017, now at Cytomine cooperative)

Rémy Vandaele

Machine learning researcher (2014-2019)

Mehdy Ouras

UI contributor (2018)

Benjamin Stevens

Software developer (2010-2014)

Gilles Louppe

Machine learning contributor (2013)

Loic Rollus

Software developer (2010-2015)

Guillaume Vissers

Software contributor (2019)

Jean-Michel Begon

Workflow contributor (2014)

Julien Confetti

Software contributor (2015)

Pierre Ansen

Education module software contributor (2013)

Prof. Pierre Geurts and Prof. Louis Wehenkel are long-term machine learning collaborators.

Through various collaborations (including the NEUBIAS COST Action), other direct or indirect contributors were/are coming from IRB Barcelona, Pasteur Institute (Paris), VIB Ghent, MRI Biocampus Montpellier, …

If you want to contribute and work with our team (scientific research stays, internships, collaborations, …), contact us.

Funding

We are/were involved in these projects

BigPicture (2021-2027)

BigPicture (2021-2027)

Funded by H2020 Innovative Medicine Initiatives.

This project will establish the biggest database of pathology images to accelerate the development of artificial intelligence in medicine.

ARIAC/TRAIL (2021-2025)

ARIAC/TRAIL (2021-2025)

Funded by Wallonia/DGO6

ARIAC by DigitalWallonia4.ai / SPW-Recherche. Applications et Recherche pour une Intelligence Artificielle de Confiance

BioMedaqu (2018-2022)

BioMedaqu (2018-2022)

Funded by H2020 MSCA ITN network.

This project aims at developing new image analysis modules for Zebrafish imaging.

COMULIS (2018-2022)

COMULIS (2018-2022)

Funded by EU COST Action.

This project aims at developing new multimodal imaging visualization and annotation modules.

IDEES (2017-2020)

IDEES (2017-2020)

Funded by FEDER (ERDF/Wallonia).

This project aimed at developing new big data software modules and deployment mechanisms.

DeepSport (2017-2020)

DeepSport (2017-2020)

Funded by Wallonia/DGO6

This project aims at developing new visualization and annotation modules for video data.

NEUBIAS (2016-2020)

NEUBIAS (2016-2020)

Funded by EU COST Action.

This project aimed at developing image analysis reproducible benchmarking modules, see BIAFLOWS.

ADRIC (2017-2021)

ADRIC (2017-2021)

Funded by Pole Mecatech, Wallonia/DGO6.

This project aims at developing new AI modules for industrial quality control.

HistoWeb (2014-2017)

HistoWeb (2014-2017)

Funded by Wallonia/DGO6.

This project aimed at developing Cytomine education software modules

SMASH (2012-2014)

SMASH (2012-2014)

Funded by Wallonia/DGO6.

Cytomine Business development project.

Cytomine (2010-2016)

Cytomine (2010-2016)

Funded by Wallonia/DGO6.

Initial Cytomine R&D project.

Contact and Community

We recommend to use the Image.sc forum as the discussion channel for user related questions.

For bugs or feature requests related to our experimental features, we recommend to post your issue on Cytomine ULiège R&D Github repository. For the official version, please post issues on the official repository.

To ask for a specific demo account, to support the open-source project, or for service requests (e.g. installation, hosting, maintenance, trainings, specific software developments, slide scanning), please contact Cytomine cooperative.

For any questions related to our research activities or for potential research collaborations or internships @ ULiège, feel free to contact our ULiège R&D team: