Research Engineer (M/F) - Development and deployment of unsupervised anomaly detection models for neuroimaging.

Job Description

Missions

The PRODROM-ND project aims to develop tools for analyzing multimodal data (imaging, clinical data, omics) from the prodromal stages (i.e., preceding the onset of disease symptoms) of various neurodegenerative diseases. One of the project's research areas focuses on developing deep learning models to detect imaging biomarkers sensitive to weak signals associated with prodromal stages, particularly the presence of subtle anomalies in imaging data. The goal is to provide a state-of-the-art review of unsupervised anomaly detection (UAD) models for neuroimaging, define a standardized and reproducible evaluation framework for these methods across various neurodegenerative diseases, adapt existing models, if necessary, to the specific characteristics of the clinical problem, and to obtain an initial assessment of these models' performance, which will then serve as a basis for developing more robust models that are sensitive to the subtle anomalies characteristic of the early stages of these diseases.
The MYRIAD team at the CREATIS laboratory has developed strong expertise in UAD models for neuroimaging and has developed several original models
The successful candidate's role will be structured around four main areas.
-Work with the UAD models developed at the CREATIS laboratory and consolidate them into a unified code library.
-Select the most relevant UAD models for neuroimaging from the literature and integrate them into the UAD code library.
-Adapt the UAD models to the specific characteristics of neuroimaging data (three-dimensional nature, learning of dedicated representations, consideration of cross-sectional data, etc.).
- In collaboration with the other partners in the PRODROM-NN project, conduct a comparative study of the performance of various state-of-the-art UAD models and/or those developed by the various partners under defined and reproducible experimental conditions. This work will require centralizing several public databases and defining performance metrics tailored to the specific characteristics of prodromal neurodegenerative abnormalities.

Activity

- Conduct a review of the scientific literature to determine the state of the art regarding the most effective UAD models in neuroimaging.
- Develop the UAD model library by unifying the available code from the various models in the laboratory and in the literature.
- Propose strategies for adapting the developed models to the specific characteristics of neuroimaging data. Evaluate technical and computational feasibility.
- Establish a benchmark using public, multi-site cohorts of neurodegenerative diseases, including early-stage data.
- Contribute to the academic life of the group and the PRODROM-ND project by participating in and giving presentations at seminars, reading groups, and meetings.
- Contribute to the dissemination of research results through scientific publications and the provision of open-access tools.

Your Profil

Skills

• MD or PhD in Computer Science specializing in Neuroimaging data processing.
• Proficiency in Python and deep learning libraries such as PyTorch or TensorFlow
• Proficiency in Git and collaborative development practices.
• Previous experience in developing UAD models for neuroimaging is required.
• Ability to analyze the literature, evaluate available tools, and adapt them into solutions that can be used by a research team.
• Rigour, independence, ability to work with researchers, engineers and PhD students in a multidisciplinary environment.

An initial screening will be carried out based on the CV, experience, skills and cover letter. Shortlisted candidates will be invited to an interview. Applications must include the following documents:
• A full CV,
• A covering letter setting out the candidate's research interests in relation to the project,
• The contact details of one or two referees,
• If possible, one or two letters of recommendation,
• Transcripts from your Master's degree (M1 and M2) or equivalent, as well as any relevant certificates.

Your Work Environment

CREATIS is a multidisciplinary medical imaging research unit (approximately 200 people) in the Auvergne-Rhône-Alpes region affiliated with the University of Lyon 1, INSA Lyon, the CNRS, and Inserm. CREATIS's main areas of research are linked to two fundamental issues: 1) identifying major health challenges that imaging can address, and 2) Identifying theoretical challenges in biomedical imaging related to signal and image processing, modeling, and numerical simulation. The MYRIAD research team conducts upstream research leading to the design of advanced image processing and modeling methods. It has recognized expertise in machine learning and the prototyping of UAD models for neuroimaging. The person hired for this position will join the national PEPR PRODROM-ND project.

Compensation and benefits

Compensation

Starting at 3175 euros gross per month, depending on experience

Annual leave and RTT

44 jours

Remote Working practice and compensation

Pratique et indemnisation du TT

Transport

Prise en charge à 75% du coût et forfait mobilité durable jusqu’à 300€

About the offer

About the CNRS

The CNRS is a major player in fundamental research on a global scale. The CNRS is the only French organization active in all scientific fields. Its unique position as a multi-specialist allows it to bring together different disciplines to address the most important challenges of the contemporary world, in connection with the actors of change.

CNRS

The research professions