Bioimage analysts from all over Europe and beyond met this year in the beautiful city of Szeged, Hungary for the second annual Network of European Bioimage Analysts (NEUBIAS) conference, which consisted of a symposium preceded by two training schools.
The symposium was hosted at the Biological Research Center of the Hungarian Academy of Sciences from January 31st to February 2nd. There were 255 participants coming from 37 countries. The design of the program was to foster an open environment to promote a cohesive bioimage analysis community. Successful signature sessions from the previous year were again the highlights of this year’s symposium. In the “Call4Help” session, life scientists and bioimage analysts presented image analysis roadblocks encountered in their research. The floor was opened to all the symposium attendees to give live feedback, which was recorded in a live Google document. The diverse backgrounds of symposium attendees cultivated spirited discussions on a variety of approaches to solve each presented problem. The Open Source Software Lounge was an interactive event of 23 stations equipped with large monitors featuring live demonstrations of cutting edge image analysis software (table 1 – see online). This year’s NEUBIAS panel discussion focused on questions regarding what it means to be a bioimage analyst, what skills are necessary to become one, as well as what kinds of training is needed and how to train people as bioimage analysts.
The current developments in the image analysis community were highlighted through exciting scientific sessions with international keynote speakers and a bustling poster session showcasing the research of NEUBIAS community. Of particular interest from the community was how to effectively visualize, navigate, and showcase large image datasets. Presentations from software programs showed an increased effort for cross-platform integration. Several recipients of the Short Term Scientific Mission grants gave summary reports illustrating the importance of cross communication for information exchange between scientists with different expertise to help mutually advance bioimaging research.
Also, an enlightening talk about publishing trends was given by Natalie De Souza, chief editor at Nature Methods, highlighting hot publication topics including, but not limited to, faster brain imaging, spatial transcriptomics, mass spectrometry imaging, and the need for advanced image analysis for organoids.
To supplement the exciting research community developments, symposium delegates had the opportunity to attend several company workshops led by generous sponsors (i.e. Zeiss Microscopy, Leica Microsystems, Bitplane, and ThermoFischer Scientific).
In the days leading up to the NEUBIAS 2018 symposium, two training schools were held to train new and experienced bioimage analysts. One training school was aimed at teaching early career scientists the basics of bioimage analysis. Each teaching session was dedicated to a different component of image analyses (e.g. deconvolution, segmentation, particle tracking) and covered both the practical aspects of these tools as well as the underlying algorithms. Hands-on participation taught students how to implement these components into workflows using different software including ImageJ, Icy, Matlab, CellProfiler, and Drishti. This offered an overview of advantages and challenges of each software. With this introduction, participants were able to analyze their own image data. The course provided early career scientists - mainly involved in image production – with a needed understanding about bioimage analysis to further their research.
The aim of the other training school was to update experienced bioimage analysts on cutting-edge image analysis tools and to broaden their existing skill set. Five outstanding workflows, submitted via the school application, were chosen by the instructors to be summarized via short presentations in the school by the applicants. An additional three applicants with exemplary workflows were invited as instructors to lead the school through a detailed, in-depth deconstruction of their workflow examining each component and statistical analyses. Their peers reproduced these detailed workflows on their own computers. On the last day, students teamed up to modify and improve components of one of the three workflows and presented their results. In addition to in-depth workflow deconstruction, invited speakers, who are experts in the field, presented overviews of recent developments in specific workflow components (e.g. co-localization, image segmentation, big data and 3D visualization). There were also community discussions on various topics from strategies used to find solutions to new analysis problems, to content and methods of teaching bioimage analysis.
Both training schools had complementary ethics training aimed at providing guidelines for optimal image analysis and preventing unintentional data manipulations.
Taggathon: Web Tool Development and Benchmarking
Two complementary NEUBIAS working groups convened alongside the training schools in a Taggathon to work on improving the Bio Imaging Search Engine web tool, BISE (http://biii.eu/), and to work on benchmarking for comparing image analysis methods. Finding the right software for the bioimage analysis task at hand in the sea of tools available in the field can be challenging to even the most experienced analysts. BISE is a response to this problem to help simplify the discovery of bioimage analysis software tools. Improvements to BISE from this year’s meeting included extending and refining the EDAM-bioimaging ontology as a foundation for searchable keywords (https://f1000research.com/posters/7-180), adding more functions to the search and filter web interfaces, and integrating a FIJI search bar. An extension of the database structure was needed to augment the capability of BISE to allow the reflection of complex workflows comprised of different steps accomplished in different software tools. This extension was designed and implemented by the team and will soon be online. Lastly, the content of BISE was expanded by experts of light-sheet and electron microscopy tagging tools focused on those imaging topics.
A second aim of the Taggathon was to create an online platform to allow users to benchmark and compare workflows for common bioimage analysis problems. Cytomine was previously chosen as the basis for this platform. Improvements to the Cytomine architecture, commissioned by NEUBIAS and implemented by the Cytomine cooperative (https://cytomine.coop/), were applied to the benchmarking server and tested during the Taggathon. A standardized set of inputs and outputs for bioimage analysis problems was agreed upon and a set of metrics for core bioimage analysis problems was chosen. Several proof-of-concept workflow execution pipelines for ImageJ, Icy, and CellProfiler including the import of the results into the Cytomine database were implemented. Strategies allowing easier contribution of new workflows to the benchmarking tool were explored including the automated use of GitHub and Docker repositories and automatic generation of wrapper scripts. Scripts for benchmark metric calculation were also developed re-using existing code where possible.
The next steps for benchmarking development include making further improvements to the outward-facing Cytomine interface, finishing the implementation of automatic benchmark metric calculation, extending the tool to additional image analysis problems and software applications, and simplifying and better documenting the procedure for new workflow integration.
With the conclusion of both training schools, Taggathon, and symposium, the NEUBIAS community welcomed newcomers, trained the next generation of bioimage analysts, forged new collaborations, and shared great science. Many conference participants left inspired with the acquisition of new tools and research ideas, and the ability to toast in Hungarian: Egészségére!
1Center for Genomics and Systems Biology, New York University, USA
2Institut de Génomique Fonctionnelle de Lyon, École Normale Supérieure de Lyon, France
3Dundee Imaging Facility, University of Dundee, Dundee, UK
4Center for Microscopy and Image Analysis, University of Zurich, Switzerland
5Institute of Molecular Systems Biology, ETH Zürich, Switzerland
Joyce Y. Kao
Institute of Molecular Systems Biology