Deep learning, MR spectroscopy, parameter estimation, uncertainties estimation, signal processing
Magnetic resonance spectroscopy (MRS) is an important technique in biomedical research asit has the unique capability to give access non invasively to the biochemical content (metabolites) of scanned organs.In the literature, the quantification (the extraction of these potential biomarkers from the MRS signal), involvesthe resolution of an inverse problem based on a parametric model of the metabolite signal. However, quantificationresults in large uncertainties for most of the metabolites which is one of the main reason that prevents the use ofMRS in clinical routine.In this project, an original approach using deep learning will be used.Our first results demonstrate that our approach can constitute both a qualitative leap and important reductionof computation time. The recruited post doc would extend and improve these results: improves our network,provides an uncertainties estimator, investigates the features discovered by the network and also refines our NMRsignal model. Applications on real clinical data/problem and consequences on acquisition strategies will also beinvestigated in collaboration with our partners.
The project will be realized in partnership between the team ”Images and Models” (machine learning for med-ical imaging) and the team ”NMR and Optics: From Measure to Biomarkers” (MRS acquisition and quantification)of the CREATIS lab in Lyon (France).
Strong knowledge in the following fields is required:
The successful candidate is expected to be autonomous and show strong motivation and interest in multidisciplinaryresearch. He/She will need to understand the question and issues related to MRS quantification of in vivo data.The available code is written in Matlab and Python and use the deep learning framework Caffe. Knowledge ofthese tools is a plus but is not mandatory.
Interested applicants are required to send a cover letter, CV, reference letters, ... to: michael.sdika[at]creatis.insa-lyon.fr and helene.ratiney[at]creatis.insa-lyon.fr.
(c) GdR 720 ISIS - CNRS - 2011-2018.