Methods for Systematic Characterization and Monitoring of Radiotherapy Administration to Breast Cancer Patients

Date of Award

Spring 2017

Degree Name

Doctor of Philosophy


Industrial and Manufacturing Engineering

First Advisor

Dr. Diana Prieto

Second Advisor

Dr. Steven Butt

Third Advisor

Dr. Azim Houshyar

Fourth Advisor

Dr. Elise deDoncker


Breast cancer, Bayesian, monitoring, analytics, radiotherapy, breast cancer patients


This research aims to propose methods to identify and monitor the clinical and non-clinical variables that influence the receipt and adherence to radiotherapy in breast cancer patients. Building upon Bayesian network models for mining cancer registry data two methods with different purposes are developed.

The identification and characterization method, presents the characterization of the clinical and non-clinical variables for patients that received breast-conservation surgery with and without radiotherapy, by means of a Bayesian network and the proposed Zoom-in heuristic that systematically explores the model. Cancer registry data is used for training and testing of the model. The data is obtained from the SEER Program with 11736 patients from the Detroit area from 2007 to 2012. Using the Zoom-In heuristic it was established that patients under stage 0, grade undetermined, histology lobular carcinoma in situ, and age between 26-50 were more likely to receive breast-conservation surgery without radiotherapy. For early invasive breast cancer, patients under stages I, IIA, or IIB, age between 51-75, and grade II were more likely to receive breast-conservation surgery with radiotherapy.

The monitoring method, presents a quality control framework for the monitoring of patients to predict radiation treatment receipt based on the associations among clinical and non-clinical variables. This framework was developed using information from a regional cancer institution with a sample of 1922 patients from years 2009-2014. The method is based on a Bayesian network approach and is formally divided into three modules i) classification, ii) surveillance and iii) quality control. As result of the implementation, a total of 48 patients were flagged for year 2014. The quality control framework flagged patients with a metastatic disease as condition for an aggressive treatment in the cancer center. Physicians from the review panel denoted that radiation treatment is offered as a palliative treatment for patients in a metastatic phase to mitigate side effects and pain. Also, it was found that a group of patients declined radiation treatment and for other cases radiation was received in other institutions or given in different years to the ones registered in the information.

Access Setting

Dissertation-Abstract Only

This document is currently not available here.