About us
Espace utilisateur
INSTN offers more than 40 diplomas from operator level to post-graduate degree level. 30% of our students are international students.
Professionnal development
Professionnal development
Find a training course
INSTN delivers off-the-self or tailor-made training courses to support the operational excellence of your talents.
Human capital solutions
At INSTN, we are committed to providing our partners with the best human capital solutions to develop and deliver safe & sustainable projects.
Home   /   Thesis   /   Application of machine learning to ultra-fast electrical impedance tomography

Application of machine learning to ultra-fast electrical impedance tomography

Artificial intelligence & Data intelligence Engineering sciences Instrumentation Technological challenges


Close to MRI and CT scans, Electrical Impedance Tomography (EIT) is an innovative technology that images the inside of a body from measurements taken outside. The measurement consists in placing electrodes around an object to reveal its internal structure. Electric currents and potentials imposed and measured on these electrodes provide information on the distribution of impedance according to Ohm's law. This distribution ultimately provides information on the internal structure of the body. A new approach for EIT was developed at CEA to measure the water / air flows that appear in nuclear reactor accident scenarios. The performance obtained by the first prototypes motivated the ITIE start-up project to promote this technology to other industrial applications (food industry, hydrogen, chemical processes, medical, etc.).

While providing an extremely fast frame rate of several kHz [1], EIT however exhibits low spatial precision currently limiting this technology [2]. the ill-posedness of the reconstruction problem. To improve the images, it is thus proposed for this PhD to apply Machine Learning (ML) algorithms, along two main directions. The first concerns the use of ML algorithms at the end of the process to identify shapes in images reconstructed from the inverse problem, similar to facial recognition. This method is related to ML-based processing of the semantic segmentation type and to methods for correcting the loss of resolution in images. The second will consist in using ML algorithms to replace the inverse reconstruction and to perform image computation and identification in a single process. This method is an analysis of the raw data directly to both extract features and to provide output as a segmented image.

The work will be carried out at the Nuclear Technology Department of CEA, in a scientific environment characterized by great expertise in instrumentation, hydrodynamics, in close collaboration with other CEA teams who are experts in ML and within other collaborations, both national and international. The results will be promoted through scientific publications and through participation in national and international conferences.

[1] DARNAJOU, Mathieu, DUPRÉ, Antoine, DANG, Chunhui, et al. High Speed EIT with Multifrequency Excitation using FPGA and Response Analysis using FDM. IEEE Sensors Journal, 2020.

[2] DANG, Chunhui, DARNAJOU, Mathieu, RICCIARDI, Guillaume, et al. Performance analysis of an electrical impedance tomography sensor with two sets of electrodes of different sizes. Proc. WCIPT-9 (Bath, UK,), 2018.


Département de Technologie Nucléaire
Service de Technologie des Composants et des Procédés
Laboratoire d’Instrumentation Systèmes et Méthodes
Top pencilenvelopegraduation-hatlicensebookuserusersmap-markercalendar-fullbubblecrossmenuarrow-down