Seleção de modelos e estimação de parâmetros no tratamento quimioterápico de tumores via inferência bayesiana

Abstract

Cancer is a disease arising from the disordered growth of cells. Commonly, anti-neoplastic chemotherapy is used to treat the most common cancers. In this context, researcheshave turned to mathematical models that describe the growth of tumor cells with an action of achemotherapeutic drug. Faced with a variety of models in the literature for this purpose, a methodfor selecting the most suitable model is necessary. This dissertation studies mathematical modelsof cell growth and applies theApproximate Bayesian Computation(ABC) to select the modelthat best represents the observed data. The ABC algorithm used was deterministic, prioritizingthe model selection. To the selected model, the SIR particle filter was applied, which allowed toimprove the parameter estimates. Tumor growth models were studied using ordinary differentialequations and the parameters to be assumed as constants. The models were structured fromBicompartmental pharmacokinetics, which allow the study of antineoplastic drugs administeredorally. In addition, known tumor growth formulations were used by adding the influence factorof a single dose of chemotherapeutic drug