Algoritmos de aprendizado de máquina aplicados na previsão clínica de pacientes com insuficiência cardíaca e uma revisão para dados limitados

Abstract

The use of machine learning in the healthcare field represents a significant revolution in diagnostic and treatment methods for diseases. This technology harnesses advanced computational capability to identify complex patterns in medical data. This approach holds significant potential to substantially improve the efficiency of diagnoses related to serious illnesses such as Congestive Heart Failure. However, a continuous challenge in this area is the effective application of machine learning in data-scarce scenarios, i.e., situations where available samples are limited. This challenge stimulates the study of machine learning with small samples, an approach that seeks to adapt and optimize machine learning models to function efficiently even with a restricted number of data points. This dissertation, comprised of a collection of articles, explores applications of machine learning algorithms in the clinical prediction of patients with Heart Failure and reviews machine learning techniques in low-data scenarios. The first article presents a practical application using machine learning models to predict survival and hospitalization time for Congestive Heart Failure patients, focusing on the most significant variables associated with risk factors. The results highlight the potential of machine learning models in predicting clinical outcomes in patients with CHF. The second article conducts a theoretical analysis, investigating the applicability and effectiveness of machine learning techniques in contexts characterized by data scarcity. This study underscores the importance of machine learning models, techniques such as variable selection, and the use of medical data in the field of machine learning, with contributions to the healthcare sector, where data-driven decisions can have considerable impacts on patients' lives.