• Probabilistic programming
• Bayesian Neural Network
• Machine learning
• Deep Learning
• Personalized medicine.
• Data

Personalized medicine presents itself as the future of clinical practice, with the potential of revolutionizing the approach to pathology, from a dual disease/health paradigm to one in which the continuum from health to disease state is properly taken into account, tailoring each pathway to the specific subclass of patients or even to the specific patient. This thesis work fits into the relationship between artificial intelligence and personalized medicine, attempting to provide a sorting of available methods and of their current and future potential applications, exploring the critical issues of these solutions and of ML and AI in more general terms. An attempt will then be made to propose a solution, at least in part, to the problems with which AI is plagued, especially regarding the application to personalized medicine. This attempt will be pursued through the implementation of probabilistic models encoded through a probabilistic programming language for an example case of classification on structured data (MHELP database, Monignoso Heart Lung Project) representative of clinical practice, and their comparison with classical methods and deterministic neural networks predominantly used at the present time. The objective is to provide a
comprehensive view of the current situation and to propose a possible solution and future direction in the use of AI methods in personalized medicine.