Skip to content Skip to main navigation Report an accessibility issue

Noise Contaminated Maintenance Decision Processes

Dr. Mahboubeh Madadi
Louisiana Tech University
Assistant Professor of IE
Tuesday, February 13, 2018  2:00-3:00pm
JDT 500 Conference Room


In a stochastic dynamic system where the system’s true state is not directly observable and needs to be estimated using noise contaminated data, extracting right information to make optimal decisions is critical. This talk will present two such systems in healthcare and condition-based maintenance optimization. The first problem concerns the estimation of overdiagnosis risk in breast cancer screening while incorporating uncertainty in patients’ adherence behaviors. Overdiagnosis is known to be the most salient risk of screening and is defined as the diagnosis of an asymptotic cancer that would not have presented clinically in a patient’s lifetime in the absence of screening. Quantifying overdiagnosis is challenging since it is impossible to distinguish between a cancer that would cause symptoms in the patient’s lifetime and the ones that would not.  In this study, two partially observed Markov chains (POMCs) are developed to model the natural history of breast cancer and dynamics in a patient’s adherence behavior. The developed POMCs estimate the risk of breast cancer overdiagnosis when there is uncertainty in patients’ adherence, disease progression and mammography tests accuracy. The second problem focuses on the derivation of optimal dynamic sensor fusion and maintenance policies for a deteriorating system whose status is monitored using a set of heterogeneous and degrading sensors. Sensor fusion is a process by which the data derived from disparate sensors can be combined to obtain improved estimates about the system as a whole. However, when the sensors are heterogeneous and deteriorating over time, selection of the best subset of sensors for system monitoring is challenging.  A partially observable Markov decision process framework is developed to derive optimal sensor fusion and replacement policies for such systems.


Mahboubeh Madadi is an Assistant Professor of Industrial Engineering at Louisiana Tech University. She received her B.S. in applied mathematics from Shahid Beheshti University, her M.S. in industrial and systems engineering from the University of Tehran, and her Ph.D. in industrial engineering from the University of Arkansas. Her primary research interest is in sequential decision making under uncertainty, with applications in medical decision making and maintenance optimization. She is also interested in statistical analysis of complex data primarily within the area of healthcare. She is a member of INFORMS, IISE, and POMS.