*********************************
There is now a CONTENT FREEZE for Mercury while we switch to a new platform. It began on Friday, March 10 at 6pm and will end on Wednesday, March 15 at noon. No new content can be created during this time, but all material in the system as of the beginning of the freeze will be migrated to the new platform, including users and groups. Functionally the new site is identical to the old one. webteam@gatech.edu
*********************************
Thesis Title: Medical Decision Making – a Personalized Approach
Advisors: Dr. Eva K. Lee
Committee members:
Dr. Edwin Romeijn
Dr. Yajun Mei
Dr. Andy Sun
Dr. Alexander Quarshie (Morehouse School of Medicine)
Date and Time: Friday, June 29, 2018, 12:00 PM
Location: ISyE Main Poole Board Room
Abstract:
Medical decision making seeks to explain how physicians and patients routinely make decisions and identify both barriers and facilitators of effective decision making. In real practice, patients are different in their personal characteristics, drug response and treatment compliance. Furthermore, physicians are still depending on trial-and-error approach to treat patients. In this dissertation, we focus on developing mathematical foundation and computational tools that guide physicians (and patients) to make good decisions. Specifically, we developed models that estimate the patient’s personalized response to treatment and optimized the treatment to achieve best possible outcome.
In the first chapter, we focus on the management of diabetes mellitus. A pharmacokinetic and pharmacodynamics (PK/PD) drug effect model is developed to characterize the dose response of patients receiving anti-diabetic drug therapy. A linear disease progression model is combined with a drug effect model to characterize the trend of blood glucose level over time. The personalized dose response is estimated using the daily self-monitored blood glucose (SMBG) data recorded by the patient during the first 4 weeks of diabetes treatment. We tested the model on patients with gestational diabetes mellitus (GDM). Compared to the standard autoregression model, our treatment effect model gives better long-term prediction on the trend of blood glucose level. It offers the first predictive treatment-effect model which relates drug dose to drug effect.
In the second chapter, we utilized the individualized treatment effect model and result to design a personalized treatment planning model to optimize the dosing strategy for each patient. A mixed-integer program is developed to optimize the drug prescription for each patient based on his/her own personalized dose response and disease progression. In a retrospective study, we optimized the dose regimen using our model. Compared to the original dose regimen that is used to treat these patients in the clinic, the optimized dose regimen uses smaller or equal amount of drug but achieve better glycemic control.
In the third chapter, we focused on the optimization of external radiation therapy. We first proposed a math programming formulation to the isocenter selection problem and developed a fast heuristic approach to determine the number and location of the isocenters of the radiation beams. Second, we proposed a multi-objective direct aperture and beam-angle model that optimizes each treatment objective based on its clinical priority. The objectives are determined by a patient’s personalized need. To solve these optimization problems, we developed an efficient heuristic column generation algorithm that creates apertures without using any dose information. Finally, we incorporated a mixed-integer beam selection module into the direct aperture optimization to optimally select beam angles from which the radiation is delivered. The test result of three types of patients (lung, prostate, and intracranial cancer) shows that our method can create deliverable clinically acceptable plans within reasonable time. The resulting plans offer better dose distribution than the clinical plans with bean angles that cannot be pre-selected by human planner.
