Machine and Deep Learning for Disease Mapping

The use of Machine Learning and Deep Learning in the medical field has revolutionized the management and analysis of healthcare data, with particular focus on Electronic Health Records (EHRs). In this context, the TrustAlert project aims to create an integrated platform for analyzing patient medical trajectories to identify clusters of patients with similar characteristics. This thesis, conducted within the framework of TrustAlert, and in collaboration with the Department of Clinical and Biological Sciences, Università degli Studi di Torino, focuses on the analysis of data derived from structured health databases of a local health authority in Piedmont (ASL Cuneo 2), such as hospital discharge records, outpatient visits, and emergency room visits. To achieve this goal, Machine Learning and Deep Learning techniques have been used to analyze administrative healthcare data, including diagnosis and procedure codes, to identify relations and patterns between events. In particular, the core work of the thesis has focused on using a Convolutional Autoencoder (ConvAE), trained in a self-supervised way, for this task. This analysis will help identify clusters of patients with similar characteristics, providing an informational basis for resource allocation and priority setting in healthcare.