Annonce

Les commentaires sont clos.

Deep learning-based approaches for dose images denoising in Monte Carlo simulations

13 December 2022


Catégorie : Post-doctorant


Scientific context

Monte Carlo simulations (MCS) are using random sampling methods for solving physical and mathematical problems. They play a key role in medical applications, especially to treat cancer by radiotherapy. MCS simulate the interactions of the particles through the matter allowing to estimate the irradiation dose deposition within human tissue. Dose simulation is used to study new protocol in radiotherapy or in radiation protection.

 

The main drawback of MCS is the need of long time of calculation to obtain a result with a sufficient statistic. One possibility is to use denoising or inpainting approach to improve the image quality of the result provided by the simulation. However standard image processing technics are quite limited especially when the sampling is extremely low (noise very high) or when information is missing, which is mainly the case in MCS (see fig. 1-left). Recent advances in IA, especially in deep learning, show a promising solution to deeply improve the MCS in order to obtain a fast and accurate image of the dose deposition within tissue.

 

Job description and missions

The job consists in investigating new methods in deep learning for MCS. The aim is developing approaches that denoise images obtained from a fast simulation (fig. 1-left), in order to obtain an equivalent image of a long-time simulation (fig. 1-right). The main issue is the genericity. The proposed method must work for any kind of MCS. A possible way is investigating patch-based or dictionary learning approaches.

 

Profile

Candidate with a PhD in computer sciences, image processing, computer vision or applied mathematics. Good programming skills is an important requisite, especially in python. Autonomy, open-mindedness and motivation, as well as good English speaking/writing skills, are also expected. Some experience in deep learning is appreciated. Experience in MCS is not required, since the work will be focused on the dose map which are basically images.

 

Position context

The postdoc will join the INSERM UMR1101 Laboratory of Medical Information Processing (LaTIM, Brest, France). Our research group is composed of 15 peoples including PhD students and other postdocs. The future recruited postdoc will work in collaboration with different academic and hospital partners within the context of MoCaMed project funded by the French National Research Agency. The position will be for an initial duration of one year and could be renewable. Salary is about 2100 € net/month, depending on the candidate’s experience. The position starts as soon as possible and the deadline to apply is March 2023.

 

Contact and additional information

For application, a folder that contains a CV, a motivation letter, a complete list of publications, letters of recommendation, and a copy of your thesis manuscript (if possible) have to be sent to the following e-mails:

 

Julien Bert (julien.bert@univ-brest.fr)