PhD position: Reinforcement Learning for Prognostic Modelling of NonInfectious Uveitis
PhD Proposed Start Date: October 2026
Required profil:
The ideal candidate holds a Master's degree (or equivalent) in one or more of the following disciplines: computer science, biomedical engineering, applied mathematics, or a related quantitative field. The following competencies are sought:
• Solid foundations in machine learning and deep learning (RL, attention mechanisms, ensemble methods).
• Proficiency in Python and relevant libraries (PyTorch or TensorFlow, scikit-learn).
• Experience or strong interest in prognostic analysis.
• Familiarity with semi-supervised or self-supervised learning is a plus.
• Capacity for interdisciplinary collaboration with clinical partners.
• Scientific writing skills and ability to communicate results to both technical and clinical audiences.
Abstract:
Non-infectious uveitis is a heterogeneous group of intra-ocular inflammatory diseases and remain a leading cause of preventable blindness among working-age adults. Its clinical course is characterized by unpredictable relapses that may lead to irreversible visual impairment while also generating frequent, resource-intensive followup visits.
This PhD project aims to develop an artificial intelligence framework to improve personalized management of patients with non-infectious uveitis. The first objective is to develop and externally validate a prognostic model that predicts the individual risk of relapse within one year using multimodal clinical data. The second objective is to translate these personalized risk estimates into tailored follow-up strategies that optimize the timing of clinical visits while minimizing unnecessary consultations. The central methodological hypothesis is that patient management should be viewed as a sequential decision-making problem rather than a series of independent clinical encounters: After developing a robust prognostic model, Reinforcement Learning (RL) methods will be investigated to learn adaptive follow-up policies that continuously integrate newly available clinical information, treatment modifications, and disease evolution to recommend individualized monitoring schedules.
The clinical state space of non-infectious uveitis is highly complex, encompassing diverse disease phenotypes, etiologies, imaging findings, therapeutic strategies, and longitudinal treatment responses. To improve prediction of relapse, visual outcomes, and ocular complications, structured clinical data will be integrated with automated analysis of multimodal ophthalmic imaging using state-of-the-art machine learning and computer vision approaches. Particular attention will be devoted to the development of interpretable AI models that satisfy the transparency and safety requirements necessary for future clinical implementation.
This interdisciplinary project lies at the interface of ophthalmology, artificial intelligence, medical imaging, and clinical epidemiology, and offers opportunities to develop novel machine learning methodologies while addressing major unmet needs in precision medicine.
Application: CV, Publications, Master transcripts, Motivation, Recommendation, English level.
Contact: robin.jacquot(AT)chu-lyon.fr, tao.wang(AT)insa-lyon.fr