Bandits improve patients follow-up – BIP-UP

Machine learning is accompanied by many hopes in health. In project BIP-UP, we will explore the follow-up of patients who underwent bariatric surgery. Currently, this follow-up is the same for all patients though practicians are well aware that a personalized follow-up would better suit. Patient follow-up will be modeled as a contextual bandit problem, which precise definition will be detailed during the project. Training of a bandit algorithm in a context where data is scarce, context is made of features of different natures, risk-aware, in interaction with an autonomous dynamical system (the patient) will be the main challenges that we will study on a theoretical side. We will investigate these points with the necessary rigor, yet in a pragmatic way. We will use data from a cohort collected over the last 15 years at the CHU de Lille to train the algorithm, and we will test it on other validation cohorts. We will also use data about the patients that are available in the HDH to enrich the data and improve our predictions, hence the follow-up. Finally, the goal of BIP-UP is to develop a software companion for the medical team who will evaluate the tool in practice on their patients.
BIP-UP relies on a tight collaboration between Inria team-project Scool in Lille, expert in bandit theory and their applications, and INSERM U 1190 expert in bariatric surgery, whether in clinical terms, and for their research on the topic.
To summarize, BIP-UP has the ambition to lead exploratory research activities on a certain type of patients, design, develop and assess a software companion for patient follow-up. This approach may be adapted to the follow-up of other types of patients in the future.