I am a French PhD student in machine learning. I wanted to apply data mining to the medical field in order to bring a meaning to what I do, so I chose to do my PhD on the automatic diagnosis and prognosis of neurodegenerative diseases at AramisLab. I also enjoy signal and image processing, which I studied during my engineering studies at the INSA of Rouen.
I love to read and take pictures, and I am always eager to travel and discover new places and cultures. This is why I spent a year as an exchange student in the US in 2011, and I would enjoy having another international experience after my PhD.
What I do
I am currently doing my PhD at AramisLab, a joined team between Inria, CNRS, Inserm and Sorbonne University, located at the Brain and Spine Institute in Paris. I am co-supervised by Didier Dormont and Stanley Durrleman, and I am working on automatically diagnosing Alzheimer's disease and predicting its evolution.
We worked on lowering the recruitment cost for Alzheimer's disease clinical trials, by proposing a pre-screening tool for selecting subjects having amyloid plaques. We benchmarked a variety of algorithms: Random Forest, SVM, logistic regression, adaptive boosting, and adaptive logistic regression. We studied the impact of the input features (cognitive or imaging variables, cross-sectional or longitudinal), feature selection and data set size. We showed this could lead to a significant reduction in recruitment cost.
Predicting diagnosis evolution
We focused on predicting the evolution from Mild Cognitive Impairment to Alzheimer's Disease status. We first conducted a systematic quantitative review, to identify current trends and the methodological elements that most impact the performance of the predictor. We also proposed guidelines to ensure the performance evaluation reflects what can be achieved in clinical practice. In a second part, we proposed a method for predicting the diagnosis evolution while maximizing interpretability. Our method is composed of two parts: (1) Prediction of the evolution of the features at the given date; (2) Prediction of the corresponding diagnosis using the predicted features.
Using real-world data
We now want to identify subjects at risk of having Alzheimer's disease in a data set of patients followed in French hospitals or by general practitioners. We first identified biases that can be encountered in clinical routine data bases compared to research ones. We are now working on identifying at-risk subjects based on their medical history, treatments and co-morbidities.
My most relevant publication to date is:
Ansart, M., Epelbaum, S., Gagliardi, G., Colliot, O., Dormont, D., Dubois, B., Hampel, H.,Durrleman, S., 2019. Reduction of recruitment costs in preclinical AD trials: validation of automatic pre-screening algorithm for brain amyloidosis. Statistical methods in medical research, p.0962280218823036. doi:10.1177/0962280218823036 Available on HAL
A complete list of my publication record can be found on HAL.
I presented my work at various conferences, via poster (AAIC 2018, OHBM 2019) or oral presentations (MICCAI 2017 - MLCDS workshop, plenary session at CTAD 2017). I also presented at the WiMLDS Paris meet-up. General public presentations of my work can be found in the special issue on AI and medicine of Liberation (a French journal of national audience), and in the scientific radio show "Le Club de la Tête au Carré" on France-Inter (most-listened-to radio station in France) on January 9th 2019.
I have been teaching at Sorbonne University for three semesters. I have given practical classes in programming (in C and Python) and in Artificial Intelligence.