Aprendizado de máquina aplicado na previsão do tempo de sobrevida em pacientes com câncer de pulmão

Abstract

Lung cancer is one of the most common and lethal types of cancer worldwide. Early diagnosis and appropriate treatment play a fundamental role in reducing mortality associated with this disease. Artificial intelligence has become a promising tool in several areas, including medicine. A Systematic Literature Review (SLR) on survival time prediction in cancer patients revealed that, among the 64 studies analyzed, 55 used some machine learning technique. In this context, this study proposes applying supervised machine learning techniques to develop and evaluate models capable of classifying lung cancer patients according to their survival time (short or long). The results indicate that the algorithms with the best performance in this task were Random Forest, with an accuracy of 87.12%, and Decision Tree, with an accuracy of 86.94%. To understand the results of Machine Learning (ML) models, we used SHAP (SHapley Additive exPlanations), which allows interpreting the contribution of each variable to the predictions made. Incorporating these models into clinical practice can support decision-making and personalization of patient treatments.