Health

The increase in medico-social costs and the growing pandemic risk require health structures to consider a digital transition, which, in addition to the use of the current information systems supposes an automatic aid to medical practice. This can be built, among other things, on the aggregation, analysis and interpretation of (more or less structured) data; increasing scientific results and health data being already incompatible with the exhaustive nature on which the medical decision is based. This decision support, in order to be accepted and profitable, both by the patient and the practitioner, supposes that it can be explained. This notion of explainability of an automatic decision is one of the open questions falling within the fields of XAI. Concerning Health, it relies on fundamental concepts and innovative technical solutions, but also on the ability of practitioners to interpret and integrate (as part of their diagnosis) a response resulting from an analysis process and automatic interpretation.

Thus, heterogeneous and sensitive data coupled with medical expertise constitute the pillars of the Health axis in MAIA. The topic relating to precision medicine, which is essentially personalized, is based on data allowing us to take into account the heterogeneity of the patient's genetics, physiology and environment in order to identify and understand the etiology of the disease(s) he/she suffers from. The research activity concerning Health is organized around three scientific objectives:

• The first objective concerns the consolidation of medical diagnoses thanks to ML techniques, by limiting indeterminism, identifying the possible a priori of the practitioner, proposing new markers (multi-omics and holistic approaches), modeling and explaining the ``healthy patient", so as to strengthen acceptability.
• The second objective is about the prediction of the patient's future: identifying groups at risk, weak signals, in order to predict the evolutionary trajectory of the patient under treatment, and to limit the side effects induced by therapies.
• The third objective relates to constraint-based optimization in health. This concerns the organization of care via the identification of trivial cases, but also issues such as the anticipation of capacity, the waiting time in emergency services, and even surgical planning.