Apprentissage profond pour la détection de foyers épileptiques en imagerie IRM-TEP
Recrutement en cours/passé: 
Recrutement passé

Key-words: multimodality (PET and MRI) medical imaging, neuroimaging, machine learning (deep learning)


For patients suffering from drug resistant epilepsy, surgical removal of the epileptogenic zone offers the possibility of a cure. The analysis of neuroimaging data (MRI, PET..) is increasingly used in the presurgical work-up of patients to localize the epileptogenic zone. We are currently developing a software based on statistical machine learning that will automatically process these medical imaging data so as to provide maps highlighting abnormal regions in the image. We recently prototyped a first system based on features automatically learned from magnetic resonance imaging (MRI) based on the deep learning methodology (see figure in the pdf file for illustration). The objective of this master project is to improve the diagnostic performance of this system by accounting for the information provided by a nuclear imaging modality referred to as positron emission tomography (PET) (see figure 2). The successful candidate will explore different ways to incorporate this new data modality into the pipeline. The first step will consist in building the MRI-PET image database, and more specifically processing the PET data. This task will be supervised by the CERMEP team. The second step will consist in designing and evaluating different fusion strategies of the PET and MR data within the deep feature learning architecture. This task will be supervised by the CREATIS team. Dr Julien Jung who is a neurologist at HCL and CRNL (Centre de Recherche en Neurosciences de Lyon) will also actively contribute to this project by providing his strong clinical expertise in the evaluation of the epilepsy lesion maps that will be generated by the software as well making the MR and PET data available.


Strong knowledge in at least one of the following fields is required:

  • machine learning (deep learning/ outlier detection);
  • Image processing
  • Applied mathematics

The available code is written in Matlab (SPM toolbox) for the image database processing and Python for the machine learning part.

The successful candidate is expected to be autonomous and show motivation and interest in multidisciplinary research (image processing and machine learning in a medical context).


Interested applicants are required to send a cover letter, CV and any other relevant documents (reference letter, recent transcripts of marks,...) to and

Gratuity : ~550 euros/mois.

See the attached pdf file for illustrations and further details.