An Analysis of Thyroid Cancer Incidence and Treatment Classification

Abstract

Although surgery is the primary treatment for patients diagnosed with thyroid cancer, the extent of the required surgery remains controversial. This thesis presents a statistical analysis of thyroid cancer incidence and management costs in order to identify commonly-associated risk factors with the development of the disease, and to motivate the necessity for improvements in treatment recommendations. In an at- tempt to determine the optimal treatment given the features of a diagnosed patient, it employs statistical and machine learning techniques, including principal compo- nent analysis, logistic regression, least absolute shrinkage and selection operator, and support vector machines, to create different models that predict treatment recom- mendations from a computational perspective. The models developed are restricted to binary classification of either a total thyroidectomy recommendation or not. The results do not resolve the debate with regards to thyroid cancer treatment, but pro- vide a strong foundation for moving forward in applying computational techniques to the problem.