Date(s) - 04/19/2012
10:30 am - 11:30 am
Our research treats a person as an individual process and the objective is to determine the best dynamic model for an individual and exploit this model to improve their health by providing on-line monitoring or control of key variables related to their health. Our approach develops complex dynamic models from multiple non-invasive input variables that impact the levels of the key variables. This talk will discuss three areas of application. The first one is in human thermoregulatory modeling (HTRM) where the key variables are core temperature, skin temperature and sweat rate. The inputs consist of environmental variables as well as physiological ones. The potential impact of HTRM include the design of clothing for use in extreme environments, the control of environmental suits, the prediction and control of diseases related to skin blood flow for example, and control of the environment for individuals with poor thermoregulation such as premature babies.
The second area of application is in diabetic modeling. In the case of type 2 diabetes, this work focuses on the development of continuous-time glucose monitoring of blood glucose level via a virtual sensor. Glucose modeling data come from the subject’s personal meter; inputs come from an activity armband and eating information. From as little as 12 glucose readings over 3 days of data collection, this work demonstrates the development of continuous-time glucose sensors based on 11 inputs for 24 subjects in a clinical study. The type 1 work involves the development of an artificial pancreas based on process control ideas. The development of continuous glucose sensors and the insulin pump are two technologies that have paved the way for feedback control of glucose. However, except under mild changes in disturbances such as during sleep, feedback control alone is not sufficient. This approach involves the modeling of activity, including stress, food and insulin with a view towards feedforward control and presents results from11 subjects in an ongoing clinical study.
The final area is in cancer treatment and this talk will propose a new paradigm in modeling tumor growth from a process systems engineering perspective. The basic idea is to identify noninvasive inputs that can be measured frequently or known fairly continuously that influence the internal processes and chemical concentrations that impact tumor growth. Possible variables include all types of stress (measured by galvanic skin response sensors), skin temperature, chemical and radiation therapies, activity variables, circadian rhythm, diet, rest efficiency, sleep duration, environmental changes, blood glucose levels, heart rate, etc. Having a means to measure tumor growth at a sufficient rate, a multiple input dynamic model, which is likely to be nonlinear, can be identified (i.e., structure determined and parameter estimated). With a representative phenomenological structure for this model, a cause and effect relationship can be found that can be effective in providing information to optimally reduce tumor size. Thus, the objective here is to describe this approach to this application and discuss its potential as an active area of research as it relates to monitoring, virtual sensor development, feedback control, feedforward control, and model predictive control in subject-specific on-line adaptive modeling.