SC4: Assistance machine pour acquisition, analyse, interprétation/modélisation de données de microscopies

SC4: Assistance machine pour acquisition, analyse, interprétation/modélisation de données de microscopies 

A. Demortière (UPJV, Amiens), P. Paul-Gilloteaux (IRS, Nantes)

Keywords: méthodes basée deep learning, transfer learning, model zoo, segmentation et analyse multimodal  , visualisation de données complexes ou multimodale, reconstruction de données (tomo ou ptychographie), aide à la prise de décision lors de l’acquisition/acquisition intelligente basée traitement en ligne d’image.

Invited speakers: D. Sage (EPFL Lausanne), N. Pustelnyk (ENS Lyon)


Digital technology has taken a prominent place today in microscopy, whether in the field of materials sciences, medicine or biology.The use of deep learning approaches has become an essential solution in the processing and analysis of big data acquired in microscopy. They allow their classification or their modeling on the basis of statistical laws, and can also assist us in decision-making and the visualization of these data. “Ad hoc” algorithms and “deep learning” approaches have demonstrated in recent years their ability to be able to help decision-making on the acquisition parameters to be chosen and even in some cases replace the operator. For example, analyzing images during the acquisition can help decide what to do next and guide the microscope, or allow you to search for rare events automatically. They also make it possible to extract information from the image that did not seem to be present, using a priori knowledge, formalized or not, as in the case of denoising, artifact correction or resolution improvement of the image.They are also able to extract phenotyping or characterization data automatically and without speed comparison with a human operator, with reproducible quantification bias, by segmentation, tracking, or analysis of signals and derived measurements. The development of these algorithms and in particular machine learning models for all these applications requires the sharing of annotated data in order to develop them, which also calls for new sharing and annotation solutions.


En poursuivant votre navigation sur ce site, vous acceptez l’utilisation de cookies pour réaliser des statistiques de visites.