Development of algorithms for image registration and spectrum estimation in multispectral imaging applied to endoscopy

3 Juillet 2024

---

Catégorie : Post-doctorant

---

Hosting structure

ImViA Laboratory, Université de Bourgogne, Dijon, France

SATT Sayens, Dijon, France

 

Supervisors

Yannick Benezeth, Franck Marzani, Flavie Courtaut

 

Keywords

Multispectral imaging, image registration, spectrum estimation, classification, deep learning, clinical endoscopy.

 

Context

The ImViA laboratory research team behind this project has been working on digital chromoendoscopy for some ten years, in collaboration with a clinical endoscopy research team. The stomach and colon are the two major organs targeted in studies carried out to characterize inflammatory lesions at an early stage. In this context, the multi-dimensionality (spatial and spectral) provided by multispectral imaging is a source of new approaches.

As a result of the ANR EMMIE project, the team has been awarded two years' maturation funding. A prototype for multispectral image acquisition in endoscopy is currently being developed and will be operational in autumn 2024. The person recruited will work for 18 months on the development of software tools for analyzing images from the above-mentioned prototype. His/her role will be at the interface between research and innovation, as he/she will be tasked with developing and implementing innovative multispectral image analysis algorithms.

 

Subject

Three major and complementary software blocks will be developed.

1. As the imager provides multispectral images in the form of videos, an image registration step is essential, particularly as in-vivo acquisitions in the stomach are subject to numerous movements.
2. The estimation of reflectance spectra consists in transforming the signal present in multispectral images into a reflectance spectrum specific to the scene under study, with maximum freedom from the impact of the light source and camera response.
3. Finally, classification and characterization of the spectral reflectance values of the gastric mucosa are the ultimate goal of the project.

 

Profile required

Master's degree, engineer or doctorate.

Skills in image analysis, machine learning and deep learning. Knowledge of multispectral imaging would be a plus.

Python + Opencv, tensorFlow / Pytorch, scikit learn. ...

Ability to work independently and as part of a team.

Taste for transdisciplinarity and research for innovation.

 

Salary

According to profile

 

Duration

Oct., 2024 – April, 2026

 

Documents required to apply

CV, cover letter, recommendations, most recent transcripts

To be sent by August 31, 2024 to:

yannick.benezeth@u-bourgogne.fr

franck.marzani@u-bourgogne.fr