2 Post-doctoral Researchers in Foundation Models for Neuro Imaging

Background. Radiology has taken a central role in the diagnosis and management of brain diseases. In the present project, we work of next generation AI models to extend the molecular imaging capacity of Magnetic Resonance Imaging (MRI). Building on the various contrasts mechanisms available in multi-parametric MRI, we aim to establish deep learning models that predict biomarkers of diseases progression and response to therapies, with applications in brain tumours and neuro-degenerative diseases. Our data comes from large public cohorts as well as local proprietary databases of clinical and preclinical medical images, including Positron Emission Tomography, multi-modal MRI and Histology, which we use for models training and validation. 

Objectives. A key research focus of our group is the optimization of medical image-to-image translation models. In the present project, we will build foundation models for multimodal brain imaging and apply them to clinical radiology applications, including biomarkers prediction, brain image synthesis (from one modality to another) and response to radiotherapy assessment. These models will rely on the Joint-Embedding Predictive Architecture architectures with Vision Transformers. They will leverage Mixture of Experts, Chain of Thought and Retrieval-Augmented Generation to maximize their generalization and privacy. Related topics such as robustness and explainability will be also covered in tight collaboration with medical doctors and radiologists involved in the projects.

Key Skills, Experience and Qualifications

• PhD in machine learning, computer vision or a related field;
• Established expertise in deep learning methods applied to images analysis. Experiences with generative models, image-to-image translation models and medical images analysis are considered important assets;
• Established expertise in key machine & deep learning frameworks and toolsets;
• Experience in GPU computing, HPC, Containers & Image processing tools would be appreciated;
• A strong track record of publications in high profile conferences or journals in the field;
• Independent and self-motivated person with scientific creativity and originality, strong team spirit and collaborative capacity. Excellent time management, rigour, perseverance and strong writing skills;
• Fluency in English is mandatory.

Training and research environment. The Brain Imaging & Neuro Epidemiology group (BraINE) develops advanced methods for medical images analysis, combining computer vision, radiomics and deep learning approaches. The team has a double affiliation to the Department of Cancer Research and the Department of Precision Health of the LIH, and has access to state-of-the-art preclinical imaging equipment and deep learning computing infrastructures. The post-doctoral fellows will be supported through FNR CORE projects and will join an interdisciplinary project team, including biologists, medical imaging & computer science experts. The position foresees occasional traveling to and interactions with partner groups in Europe.

Researchers are supported by easy access to scientific and medical expertise, well-equipped facilities, an active seminar program as well as opportunities for conference attendance and collaborations with other research organisations.

Scientific contacts: Dr Olivier Keunen, olivier.keunen@lih.lu and Dr. Salah Ghamizi, salah.ghamizi@lih.lu 

More info on the activities of the group can be found on the https://researchportal.lih.lu/en/organisations/brain-imaging-neuro-epidemiology

Gender Equality

The LIH is an equal opportunities employer. We are fully committed to removing any discriminatory barrier related to gender, and not only, in recruitment and career progression of our staff. The LIH is attentive to gender representation among its leadership staff and aims to eliminate obstacles to the recruitment and promotion of female leaders and their career development.