En poursuivant votre navigation sur ce site, vous acceptez le dépôt de cookies dans votre navigateur. (En savoir plus)

PhD student M / F in medical imaging and AI

This offer is available in the following languages:
Français - Anglais

Date Limite Candidature : jeudi 3 décembre 2020

Assurez-vous que votre profil candidat soit correctement renseigné avant de postuler. Les informations de votre profil complètent celles associées à chaque candidature. Afin d’augmenter votre visibilité sur notre Portail Emploi et ainsi permettre aux recruteurs de consulter votre profil candidat, vous avez la possibilité de déposer votre CV dans notre CVThèque en un clic !

Faites connaître cette offre !

General information

Reference : FR3423-CHRFER-002
Workplace : POITIERS
Date of publication : Thursday, November 12, 2020
Scientific Responsible name : Thierry Urruty
Type of Contract : PhD Student contract / Thesis offer
Contract Period : 36 months
Start date of the thesis : 1 February 2021
Proportion of work : Full time
Remuneration : 2 135,00 € gross monthly

Description of the thesis topic

Ph.D. in computer science at the CNRS: Interactive deep learning for aneurysm detection and segmentation
We propose to improve the automatic detection of AI in a semi-interactive way. Thanks to the latest advances in self-learning systems and on the explainability of neural networks, the scientific challenge of this approach is to bring the medical expert into the learning loop of the selected AI approach in order to guide the choices made to better index the data.
Due to the large quantity of unannotated MRI images available in a hospital, we propose to simply improve the segmentation and classification results by exploiting them together with few images annotated by an expert in the learning phase. This process can be adapted to almost any MRI (TOF) imaging technique and task using self-supervision. The self-supervision technique introduced by Doersch et al. in [1] aims to reduce the need for a large number of annotated samples. This technique is based on contextual learning [2] and transfer of learning [3].
We assume that the implementation of self-supervision on a CNN can guide the supervised learning task within a CNN already designed for MRI image analysis [4], thus reducing the rate of error and improving the overall quality of segmentation and / or classification. In this self-supervision task, we will be interested in the interactive indexing of data by allowing the expert to intervene in the iterative indexing loop. The self-learning / interaction coupling will make it possible to obtain effective learning dedicated to the final application and involving a minimum of the expert.
References
[1] C. Doersch, A. Gupta, and A. A Efros. Unsupervised visual representation learning by context prediction. In Proceedings of the IEEE International Conference on Computer Vision, pages 1422–1430, 2015.
[2] SJ Pan and Q Yang. A survey on transfer learning. ieee transaction on knowledge discovery and data engineering, 22 (10), 2010.
[3] J. Sun and D. W Jacobs. Seeing what is not there: Learning context to determine where objects are missing. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 5716–5724, 2017
[4] A. Fenneteau, P. Bourdon, D. Helbert, C. Fernandez-Maloigne, C. Habas and R. Guillevin. Learning a CNN on multiple sclerosis lesion segmentation with self-supervision 3D Measurement and Data Processing, IS&T Electronic Imaging 2020 Symposium, Jan 2020, San Francisco, United States. Best paper award

Work Context

The thesis is part of a regional project on artificial intelligence analysis of aneurysm detection and segmentation. The partners of this project are the CHUs of Limoges and Poitiers and the research laboratories XLIM and LMA, within the CNRS research federation MIRES, in Mathematics and Data Sciences.
The work will be carried out at the CNRS, in Poitiers, between the Futuroscope site of the XLIM Laboratory at the University of Poitiers and the Poitiers University Hospital (around ten kilometers), within the framework of the I3M joint laboratory, between the CNRS, the SIEMENS company, the CHU and the University of Poitiers. This joint laboratory is co-supported by the XLIM and LMA laboratories, at the University of Poitiers, within the CNRS MIRES research federation. It uses data from a 7 Tesla MRI, unique in France for research and clinics, at the Poitiers CHU.

Constraints and risks

Short assignments in France and possibly internationally

Additional Information

Required profile :
Master 2 in computer science or bioinformatics.
Required Skills :
- Computer programming, Machine Learning and Deep Learning (Transfer learning, GAN, Auto-encoder, Self-learning system)
- Knowledge of biology, understanding of NMR phenomenon, physiology and pathophysiology of aneurysm, histology related to the appearance of aneurysm (connective tissue)
- Medical imaging method: arteriography, TOF angiography, MRI
- English: good level
- Have done a research internship (a scientific publication is a plus)

We talk about it on Twitter!