My work is organized around four complementary areas, ranging from the structuring of hospital data to diagnostic support and the generation of research hypotheses.
Area 1. Multimodal integration and information retrieval. This first area aims to bring together and make searchable data of very different kinds: clinical reports, biological and genomic data, and biomedical literature. The Dr. Warehouse project, a health data warehouse centered on clinical text, forms its foundation. It makes it possible to search and cross-reference the information contained in patient records across the entire Institute. The goal is to bring hospital data and knowledge drawn from the literature into dialogue within a single research environment.
Area 2. Information extraction from hospital data. The second area develops methods to transform the free text of medical records into usable data: entity recognition (symptoms, diagnoses, treatments), normalization to medical terminologies, and phenotype detection. These natural language processing methods feed into all of my other work by making the raw clinical material more reliable and better structured.
Area 3. Reducing diagnostic delay through computational patient representation. This third area tackles head-on one of the major challenges of rare diseases: the delay before diagnosis. By building a computational representation of the patient, a synthetic profile based on all of their data, I seek to measure similarity between patients and bring clinically similar cases closer together. The AIDY project fits into this diagnostic-support approach, helping to identify earlier those patients likely to share the same, as yet uncharacterized, genetic disease.
Area 4. Exploration and hypothesis generation. The final area explores the tools that allow researchers and clinicians to navigate the data and bring out new avenues of research. The Dr Cloud project opens up these analytical capabilities to a wider community by offering an accessible exploration environment. The aim is to move from a passive consultation of data to a genuine exploratory approach that generates clinical and scientific hypotheses.
Together, these four areas form a continuum: structuring the data, extracting meaning from it, bringing patients closer together for better diagnosis, and equipping research to make discoveries.
