Application d’algorithmes de prédiction interprétables sur les données du Transdiagnostic Connectome Project : Construction d’un réseau bayésien à partir de tests psychométriques

Details

Supervisors

Briganti, Giovanni

Faculty

Faculté de pharmacie et des sciences biomédicales

Degree label

Master [120] en sciences biomédicales, à finalité spécialisée: sciences biomédicales cliniques

Abstract

enThe Transdiagnostic Connectome Project (TCP) is a study conducted in a cohort of patients with the aim of making available to the scientific community a set of data to establish reliable associations between psychiatric disorders suffered by patients. In this thesis, we propose to exploit these data to build, by machine learning, a set of Bayesian networks in order to model the interactions between these (psychiatric) disorders. More specifically, these networks model the likelihood of developing psychiatric disorders based on patient responses to psychometric tests. Preliminary results suggest that such modelling is relevant, by comparison with the scientific literature which has already highlighted, by conventional network modelling, the interconnections between certain psychiatric disorders (for example, the set of models obtained effectively links anxiety, stress and depression). These results encourage the adoption of Bayesian networks, allowing to increase the effectiveness of clinical diagnosis, while we currently know a peak of onset of psychiatric disorders.