The authors tested out a variety of machine learning techniques, including large language models and more traditional algorithms. They focused on data that can be gathered prior to treatment, ensuring cardiologists know as much as possible before the procedure.