New retinal biomarkers for cardiovascular diseases: analysis of retinal imaging based on deep learning algorithms

his PhD project falls within the field of oculomics, an emerging and innovative area of research that leverages retinal imaging to identify ocular biomarkers reflecting systemic diseases. Thanks to recent advances in high-resolution imaging techniques—such as retinography, optical coherence tomography (OCT), and OCT angiography (OCT-A)—it is now possible to observe the retinal microcirculation in vivo, thereby providing direct access to key vascular biomarkers.

This approach offers a unique opportunity to study the correlations between retinal microvascularization and systemic vascular pathologies, particularly neurocardiovascular diseases. Currently, the assessment of vascular disease risk relies on score-based calculations derived from clinical data (cardiovascular risk factors, family history, lifestyle). These clinical data are used to build risk scales (such as the Framingham Score or HeartScore). Oculomics could complement these tools by incorporating individual vascular data, thereby refining risk scores and contributing to a more personalized approach to medicine.

Oculomics sits at the intersection of ophthalmology, artificial intelligence, and medical specialties such as cardiology and neurology. It relies on image analysis strategies and deep learning algorithms to fully harness the potential of retinal vascular data by automating the detection of subtle pathological vascular signatures. This paves the way for early detection of neurocardiovascular diseases, improved individualized cardiovascular risk assessment, and better prediction of vascular recurrence (e.g., myocardial infarction, stroke).

The aim of this PhD project is to develop and apply deep learning techniques to analyze retinal images (retinography and OCT-A) in order to better predict the onset of systemic vascular diseases and identify vulnerable patients.

This project will be based on existing and well-structured datasets: retinal imaging and rich neurocardiovascular clinical data from the Montrachet Cohort and the open-source RASTA database. A second phase of the project will involve the implementation of a prospective longitudinal cohort (BioRetDL project), which will include patients post-myocardial infarction and post-stroke (250 patients in each group, with a 2-year follow-up) at the University Hospital of Dijon (CHU de Dijon).

L'équipe IFTIM (Imagerie Fonctionnelle et moléculaire et Traitement des Images Médicales) du laboratoire ICMUB a pour objectif de couvrir deux domaines essentiels en imagerie clinique : le déploiement de nouveaux traceurs d’imagerie en préclinique puis en clinique associé au développement de nouvelles applications, et le traitement et l’analyse d’images. Cette multidisciplinarité est atteinte grâce à une équipe constituée d’hospitalo-universitaires impliqués dans la recherche appliquée en imagerie médicale, de biologistes et physiciens impliqués dans la recherche fondamentale et appliquée en imagerie médicale et d’informaticiens spécialisés dans le traitement et l’analyse d’images médicales. L’équipe bénéficie d’un environnement très favorable pour le développement de projets de recherche, avec notamment les deux établissements publics de santé de la métropole Dijonnaise (CHU et CGFL), disposant ainsi d’un plateau d’imagerie très complet et d’une plateforme d’imagerie préclinique unique en France.

skill and competencies in medical imaging/python/deep learning

Fluent in english