Proposal for 2 PhD positions: Design of a smart microscope for rapid imaging of 3D biological processes.

Institut Fresnel, Marseille, France

We are proposing 2 funded thesis topics, with the aim of designing a new microscope driven by real-time data processing algorithms for imaging biological tissues.

Fluorescence microscopy is an essential tool in biology. It allows the generation of volumetric images of living tissue, by scanning the volume point by point with a laser beam. However, this technique is slow and requires a high dose of light, which can be damaging to the cells in the sample being imaged (phototoxicity).
In this project, the aim is to develop a "smart" microscope, whose acquisition strategy is guided by learning and real-time data processing. This system will automatically determine where and when to make new acquisitions, in order to retrieve the relevant information with the minimum amount of light sent onto the sample, thus speeding up the imaging process and reducing phototoxicity. The contribution of this approach will be demonstrated in particular for rapid imaging of morphogenetic processes during Drosophila development.

We propose 2 thesis subjects:
Thesis topic no. 1: Designing new smart acquisition strategies for microscopy. The main challenges of this thesis concern in particular the imaging of dynamic processes, for which it will be necessary to design new learning and statistical data processing techniques, in order to be able:
- to predict the evolution of the sample or parameters of interest as a function of time, from a very small number of measurements,
- but also to decide where and when new measurements should be taken to verify and refine the predictions.
We will also focus on the search for fast and unsupervised algorithmic solutions, i.e. that do not require the setting of parameters by a user.

Thesis topic no. 2: Experimental implementation of an intelligent microscope. Unlike commercial microscopes that capture the entire box encompassing the tissue of interest, the student will build a scanning microscope that adapts its path according to the spatio-temporal dynamics of the sample, in connection with the theoretical work mentioned. The project will address optical aspects, the physics of the laser scanning mechanism and how it constrains unsupervised algorithmic solutions, as well as biological application to the study of animal morphogenesis.
This multidisciplinary thesis therefore combines optics, mechatronics, data science and biophysics.

We are looking for two candidates at M2 or Engineering level:
Profile required for subject 1:
- Good knowledge of data or image processing.
- Skills in statistical data processing would be a plus.
- Depending on personal affinities, the student may focus on theoretical aspects, such as approaches combining estimation theory and information theory for experimental design, or on more applied aspects.
Profile required for subject 2:
- Good knowledge of applied physics and/or biophysics and/or data science.
- Skills in optical instrument development, particularly in the field of optical microscopy, would be a plus, as would programming skills (Matlab / Python) and knowledge of image processing.
- Depending on preferences, the candidate may focus on instrumentation, algorithm development or biological applications.

Funding: Each of these 2 PhD theses will be funded by the ANR TemporalSmartScan project over a 3-year period.

Candidates interested in either of the proposed topics are invited to submit a CV, a covering letter specifying the topic for which they are applying, and the contact details of two referees to: Loïc Le Goff (loic.legoff@fresnel.fr) and Frédéric Galland (frederic.galland@fresnel.fr)

Link to the PhD description: english version , french version