Microscopy Image Analysis: The Shift to Deep Learning?
D. Sage
Proceedings of the Ninth Edition of the CNRS Thematic School, Functional Microscopy for Biology (MiFoBio'21), Presqu'île de Giens, French Republic, November 5-12, 2021, pp. 39.
The quantification of microscopy images requires automatic tools to extract relevant information from complex data. To tackle this task, numerous image analysis algorithms have been designed, commonly based on prior knowledge and on physical modeling. However, the recent success of the deep learning (DL) in computer science have drastically changed the bioimage analysis workflows to a data-centric paradigm. While this DL technology remains relatively inaccessible to end-users, recent efforts has been proposed to facilitate the deployment of DL for some bioimage applications through new open-source software packages.
Here, we present a set of user-friendly tools that allows to test DL models and to gain proficiency in DL technology: the centralized repository of bioimage model (Bioimage Model Zoo), the ready-to-use notebooks for the training, and the plugin deepImageJ that can run a DL model in ImageJ.
We provide also good practice tips to avoid the risk of misuses. We address some practical issues such as the availability of massive amount of images, the understanding of generalizability concept, or the selection of the pre-trained models. The shift to deep learning also questions the community about the trust, the reliability and the validity of such trained deep learning models.
@INPROCEEDINGS(http://bigwww.epfl.ch/publications/sage2103.html,
AUTHOR="Sage, D.",
TITLE="Microscopy Image Analysis: {T}he Shift to Deep Learning?",
BOOKTITLE="Proceedings of the Ninth Edition of the {CNRS} Thematic
School, Functional Microscopy for Biology ({MiFoBio'21})",
YEAR="2021",
editor="",
volume="",
series="",
pages="39",
address="Presqu'{\^{i}}le de Giens, French Republic",
month="November 7-12,",
organization="",
publisher="",
note="")