Postdoctoral Research position on learning methods for hadrontherapy in imXgam

Contract Period : 24 months

Expected date of employment : September 2021

Proportion of work : Full time

Remuneration : Between 2728 et 3145 euros according to experience

Desired level of education : PhD

Experience required : Indifferent

Missions

The research work is part of the TIARA project, which aims to provide an instrumental, methodological proof of concept on real data of the real time monitoring of the emission distribution of Gamma Prompts (GPs) in the context of hadrontherapy. A detection system placed around the patient - composed of an hodoscope and gamma detectors with a temporal resolution of 100 ps - provides time of flight measurements of the particles involved in the treatment. The first objective of the proposed mission is to exploit these data to be able to detect as soon as possible a significant divergence between the real treatment and the treatment plan (simulated upstream), and to stop the treatment if necessary. The final goal is to be able to estimate the distribution of GPs' vertices with a sufficient accuracy to adjust the treatment in real time with respect to the treatment plan. Access to this type of information could also pave the way for measuring the actual dose distribution delivered.

Activities

He/she will develop a research activity at the interface between data sciences and medical physics in order to explore applications of artificial intelligence for time-of-flight data in hadrontherapy. He/she will pursue the following main objectives: i) to improve our approach to 3D GPs vertex reconstruction based on an optimization problem that incorporates the physics of hadrontherapy (e.g. by analyzing simulated and real available data in order to incorporate suitable regularizations like low-rank+sparse). The results obtained at this stage will allow in particular to optimize the design of the instrument for a maximum spatial resolution; ii) to evaluate the contribution of machine learning and deep learning methods to jointly estimate the 3D distribution of GP vertices and the anatomical changes (i.e. of electron density) between the time of the treatment plan and the time of the actual treatment. This essential quantity is often estimated beforehand with significant uncertainties that must be compensated. Different strategies will be studied (transfer learning, use of Generative Adversarial Networks) and the most relevant ones will be evaluated on real data with the feedback of physicians and radiophysicists.

Skills

The candidate should have operational skills in data science, optimization methods and/or machine or deep learning.

Interest in applications in the medical field will be appreciated. Knowledge of the context and/or science of medical physics, specifically radio/hadron therapy will be appreciated.

Ability to work in an interdisciplinary environment.

Work Context

The researcher will be part of the TIARA project (Time-of-flight Imaging ARrAy for real-time monitoring in hadrontherapy) financed by the Inserm Cancer Plan, PCSI 2020 (Physics, Chemistry or Engineering Sciences applied to Cancer), a tripartite collaboration between the Laboratoire de Physique Subatomique & Cosmologie (LPSC) in Grenoble, the Centre de Physique des Particules de Marseille (CPPM) and the Centre Antoine Lacassagne (CAL) in Nice. In this project, the CPPM is responsible for the development of the data processing methodology.

He/she will benefit from the multidisciplinary, rich and stimulating work environment of the imXgam team, the CPPM and the Luminy Campus, including access to the know-how and computing facilities of the IN2P3 Computing Center.

Constraints and risks

The research work will be done in close collaboration with the LPSC team specialized in hadrontherapy physics and will require 1-2 days working visits at LPSC, Grenoble and CAL, Nice.

Additional Information

For more information, contact Yannick Boursier (boursier@cppm.in2p3.fr).

Application here : https://bit.ly/36n0skG