Self-supervised representation learning for anomaly detection in MRI neuroimaging
Recrutement en cours/passé: 
Recrutement en cours

See a detailled description in the attached pdf (english and french versions)


Host laboratory : Laboratoire CREATIS, 69 Villeurbanne- MYRIAD Team

Supervisors : Carole Lartizien -

                     Nicolas Pinon -

Keywords : Medical Image analysis and Modeling, Deep Learning, self-supervised supervised learning, Neuroimaging

Duration : 5-6 months.

Starting date : feb-april 2023

Gratuity  : ~560 euros/month


The vast majority of deep architectures for medical image analysis are based on supervised methods requiring the collection of large datasets of annotated examples. Building such annotated datasets is hardly achievable, especially for some specific tasks, including the detection of small and subtle lesions, which are sometimes impossible to visually detect and thus manually outline. This is the case for various brain pathologies including Parkinson's and epilepsy disease.

We  have developed an expertise in the field of anomaly detection methods for the analysis of multi-modality brain images, and recently applied it to the detection of early forms of Parkinson's disease or the detection of epileptogenic lesions in multiparametric MRI.

The purpose of this master project is to improve the  performance achieved with the current model architecture by exploring methodological research axes in the domain of self-supervised deep latent representation learning (see attached pdf file for details).  The objective is to continue the ongoing work on discrete models such as Vector Quantized Variational Auto-Encoders (VQVAE), to explore other learning models of the latent representation space based on Gaussian mixtures, and finally to explore anomaly detection methods such as OC-SVM or  LOF (local outlier factor) methods. 

The successful candidate will have access to the WMH public database ( containing multiple images of patients with subtle brain anomalies in different MRI sequences (T1, FLAIR) as well as to computing resources (CREATIS and/or CNRS supercomputer).

Candidate should have background either in machine learning and/or deep learning or image processing and some experience in both fields as well as good programming skills.

We are looking for an enthusiastic and autonomous student with strong motivation and interest in multidisciplinary research (image processing and machine learning in a medical context).

 The selected candidate will be in constant collaboration with the CREATIS team that is carrying out this project and will also benefit from the expertise of researchers from the Hubert Curien laboratory in St Etienne, who are partners on this project.