Fuzzy multi-relational clustering of chronic pain trajectories

Abstract

Trajectory analysis has recently received increasing attention in the healthcare domain, due to a significant increase in the volume of individual patient follow-up data. The identification of care trajectory patterns is thus becoming a major challenge in the perspective of personalized medicine. However, this task becomes more difficult when trajectory data is complex, imprecise and subjective. It is all the more difficult when medical information is represented by a self-reported discrete time series. In this work, we extend multichannel sequence analysis to the extraction of sequence trajectories described by discrete time series, covering different aspects of chronic pain. In addition, we exploit the advantages of weighted fuzzy relational clustering based on multiple distance matrices. The results show that this approach improves the interpretability of the trajectory typologies identified for medical professionals, and makes it possible to simultaneously consider several dimensions intervening in a care trajectory, thus facilitating scalability.

Armel Soubeiga, Violaine Antoine and Sylvain Moreno. " Clustering multi-relationnel flou des trajectoires de la douleur chronique " LFA 2024 : rencontres francophones sur la Logique Floue et ses Applications