Distributed computations for collaborative medical projects
Personalized medical care in the fields of statistics relies on comparing new patient profiles to existing medical records, in order to predict patient response to treatments or risk of disease based on their individual characteristics. The chances of finding past profiles similar to new patients, and therefore of more accurately treating them, naturally increase with the number of individuals in the database. For this reason, gathering patient information from different hospitals promises better health care. However, there are technical and social barriers to sharing medical dat Indeed, as the size of the database increases, it becomes more and more difficult to handle and store it. Simultaneously, institutions are usually reluctant to share their data due to privacy concerns.
During my four months at Stanford, I worked with Dr. Balasubramanian Narasimhan in the department of statistics on a technique called distributed computation, which allowed us to overcome both obstacles. It consists of leaving the data on sites and doing calculations separately, so that hospitals only share some intermediate results instead of the raw data. We implemented statistical methods in a software that will be available to hospital practitioners and will help them carry out collaborative medical projects.
My experience at Stanford was fruitful both scientifically and personally. I had the opportunity to discuss my topic with top statisticians and receive feedback from them. Scientifically, my stay at Stanford was very productive, and resulted in an article now submitted to one of the best journals in statistics. I was very lucky to be advised by several professors there, whose fields of expertise was different from my advisors in France. For this reason, I was able to develop new aspects of my research topic.
In terms of work experience, the most interesting part was to witness the differences between doing research in my lab in France and at Stanford, in terms of topics, methods, and organization. While French labs, as far as I know, are often driven by theoretical advances and beautiful math, the department at Stanford seemed more driven by potential applications. In the end, both produce theoretical and applied works, but the incentive seemed different.
One of the best things at Stanford is that students are very international, and I met Ph.D. students and postdoctoral students from all over the world. The university is also located in a beautiful area, between San Francisco—an extremely beautiful and lively city—mountains, the desert and the ocean.
The FSCIS fellowship will very likely be a crucial point in my young academic career for several reasons. First of all, scientifically, since my time at Stanford enabled me to produce research results that I could not produce in France, and I believe this will make my Ph.D. dissertation much more convincing. It also gave me the opportunity to meet professors who are experts in their field, and therefore in the future, I will know who I can contact for help or advice on similar topics. This experience also allowed me to take a step back from my specific research topic and broaden my culture on modern statistical subjects and applications.