Thesis defense – Johannes Roos

Bioimage analysis workflows have transformed dramatically over the last years, accelerated by the emergence of deep learning. Indeed, once thought impossible challenges like 3D segmentation of complex microscopy-data or data-driven multidimensional microscopy, also called smart microscopy, seem now reachable, and new imaging modalities are breaking records in both resolution and acquisition speed.

This shift brings along a variety of new software platforms and tools, as well as the requirement of dedicated computing resources like GPUs. Plagued by the absence of a common framework for these tools however, tedious data management, complex orchestration and painful integration of new technologies have become a reality and currently restrict these advanced, distributed bioimage workflows to a limited set of few programming experts.

Additionally, most existing methods are still limited in their real-time capabilities and are usually restricted to off-line analysis workflows, where analysis happens after the acquisition, limiting emerging “smart” workflows, where the analytical result can influence the acquisition.

This PhD thesis introduces a new open-source software framework that acts a middleman between users and bioimage applications: Arkitekt. Arkitekt allows for the visual and user-friendly design of modern bioimage workflows, orchestrating existing popular bioimage software locally or remotely in a reliable, efficient and in real-time. It interfaces with popular interactive visualisation and analysis software, like ImageJ and Napari, but also easily integrates developer scripts and acquisition software.

This thesis is organized in 3 main parts. After a general introduction of bioimage analysis history and a detailed review of the modern analysis workflows it fully describes the main features of Arkitekt. It then illustrates and validates Arkitekt and its capabilities on representative advanced bioimage workflows, and discusses its limits and potentials.

Jury

• M. NAGERL Valentin, Professeur des universités, Université de Bordeaux, Directeur de these
• Mme. MANLEY Suliana, Professeure des universités, École polytechnique fédérale de Lausanne, Rapporteur
• M. KERVRANN Charles, Directeur de recherche, Centre INRIA de l’Université de Rennes, Rapporteur
• Mme. MONTCOUQUIOL Mireille, Directrice de recherche, Neurocentre Magendie, Examinateur
• Mme. TESTA Ilaria, Associate Professor, SciFiLab, Karonlinska Institute, Guest
• M. SIBARITA Jean-Baptiste, Research Engineer CNSR, HDR, IINS, Bordeaux, Guest

Publications

Imaging dendritic spines in the hippocampus of a living mouse by 3D-STED microscopy Stéphane Bancelin, Luc Mercier, Johannes Roos, Mohamed Belkadi, Thomas Pfeiffer, Sun Kwang Kim, U. Valentin Nägerl. PrePrint bioRxiv. 2023-02-01.  10.1101/2023.02.01.526326 – Fiche