IMT Atlantique is opening a postodoc position about convolutional neural networks on superpixel graphs for medical image segmentation.
Goals Extend convolutional neural networks on superpixel graphs to perform medical image segmentation.
Research areas artificial intelligence, biomedical image processing, deep learning, semantic segmentation, graph
theory, mathematical morphology
Advisors Pierre-Henri Conze and Nicolas Farrugia, IMT Atlantique.
Collaborations Vincent Gripon, IMT Atlantique and Montreal Institute for Learning Algorithms (MILA), Montréal
and Gwenolé Quellec, Inserm.
Location IMT Atlantique, Brest, France
Start date March 2019, for a duration of 10 months
Deadline for application Accepting applications now, will remain open until filled
Our project addresses medical image segmentation using techniques based on artificial intelligence. To efficiently assist clinicians, computer-aided diagnosis and follow-up tools require reliable methods to segment anatomical and pathological structures. In order to increase the reliability and speed of these methods, we aim at extending convolutional neural networks to support regions that adapt to the medical content: superpixels.
Superpixels are visual primitives generated by aggregating neighboring pixels sharing similar caracteristics to form homogeneous and regular regions. Deep learning applied to superpixel graphs represents a promising perspective, requiring the adaptation of existing graph signal processing techniques to semantic segmentation. The generalization of convolutional neural networks to graphs has been formely studied for image classification. A research effort is needed to adapt these approaches to medical image segmentation based on superpixel graphs.
The purposes of this study is to propose methodological contributions at the interface between several disciplines (artificial intelligence, image processing, graph theory, mathematical morphology) and to contribute to concrete medical applications including massive diagnosis and follow-up of breast cancer using mammographies.
Funding The present postdoctoral position is funded by the Télécom Société Numérique Carnot Institute (Carnot TSN) which carries out scientific resourcing.
Host organization IMT Atlantique is a public technological university focusing on the training of engineers at the MSc. level and junior researchers at the PhD level. The Brest campus of IMT Atlantique is ideally located at the sea front (http://inovadys.com/29/telecom_bretagne/) and benefits from a high quality of life.
Send an email to firstname.lastname@example.org and email@example.com with the following:
(c) GdR 720 ISIS - CNRS - 2011-2018.