Modelagem via redes neurais de dados de sobrevivência de longa duração com dispersão não observada

Abstract

Traditional models in survival analysis assume that every subject will eventually experience the event of interest in the study, such as death or disease recurrence, so the survival function is said to be proper. Cure rate model, which was first proposed seven decades ago, has since been used to account for the presence of cure fraction, this means that a certain fraction of the individuals will never experience the occurrence of an event of interest for which they can be treated as immune or cured subjects in the context of cancer treatment. In the literature, various cure rate models have been widely studied and commonly applied to structured data with small quantities of covariates. The use of convolutional neural network, a powerful deep learning technique for image processing problem, has become increasingly more common in the medical field in recent years. Medical images such as histological slides and magnetic resonance images (MRIs) are directly related to a patient's prognostic factors, therefore, it is reasonable to introduce these images as predictors in cure model. In this work, we extend upon the article of Xie and Yu (2021b) in which a neural network was used to model the unstructured predictor's effect in the promotion time cure model's setting to the cases of overdispersed data. We will call our extension as integrated negative binomial cure rate model, and its parameters will be estimated through the Expectation-Maximization algorithm.