I am a PhD student working at the Paris Brain Institute in the Aramis team, supervised by Stanley Durrleman.
My work is focused on the modelisation of neurodegenerative diseases. A question which my research revolves around is the one of longitudinal subtyping in heterogeneous cohorts. In other words, I try to formalize a notion of typical subtypes in the progression of a disease and I devise methods to discover and understand them in a variety of clinical data. I use rigorous statistical frameworks to assess the robustness of discovered subtypes and make sure they are of clinical relevance. I believe that new methods blending the clinical insights of longitudinal studies with the dense information lying in high-dimensional data such as genetic or imaging data could greatly improve our knowledge about such disorders. Ultimately I hope that some of those tools will find their way to clinical use and be useful for tasks such as performing early diagnosis or designing better clinical trials.
My research is guided by my skills and interests whose scope ranges from applied mathematics and computer science to contemporary medical and clinical questions. Among other things, I have gathered experience in modern signal, image and shape processing methods, that involved the use of probability theory, differential geometry, optimal transport. I have also faced many bio-medical contexts and I worked using data from a variety of modalities such as sEEG, structural and functional MRI, genetic data, clinical scores and bio-banks.
Here are some of the projects I partook in before starting my PhD.
Tractogram and Atlasing With Optimal Transport, Using Multiscale and Gpu-Friendly Schemes
I worked under the supervision of Pietro Gori at Telecom ParisTech on tractogram segmention, and atlas transfer using Optimal Transport tools. This was a joint work with Pierre Rousillon and Jean Feydy. See the dedicated page for some more details and results.
High-Frequency Oscillations Detection
I studied the detection of High-Frequency Oscillations in electroencephalograms using modern methods in time-frequency analysis of signals. See the dedicated page soon, and the proof of concept detector. Here is a recent presentation (2022) that I did about this work.
You can also find some of the other work and projects that I did during my studies, either at the ENS or before.
With a rather more fun perspective, the page anecdote gathers some anecdotes that were generated by a gpt-2 model. They do not reflect any kind of truth, although they can sometimes be fun to read.
I also maintain a page gathering some LaTeX and TikZ tips, that is mostly intended to be a personal aide-mémoire, but it could be useful to others as well.