On Predicting the Outcomes of Chemotherapy Treatments in Breast Cancer

摘要

Chemotherapy is the main treatment commonly used for treating cancer patients. However, chemotherapy usually causes side effects some of which can be severe. The effects depend on a variety of factors including the type of drugs used, dosage, length of treatment and patient characteristics. In this paper, we use a data extraction from an oncology department in Scotland with information on treatment cycles, recorded toxicity level, and various observations concerning breast cancer patients for three years. The objective of our paper is to compare several different techniques applied to the same data set to predict the toxicity outcome of the treatment. We use a Markov model, Hidden Markov model, Random Forest and Recurrent Neural Network in our comparison. Through analysis and evaluation of the performance of these techniques, we can determine which method is more suitable in different situations to assist the medical oncologist in real-time clinical practice. We discuss the context of our work more generally and further work.