Date of Award

Spring 5-2020

Author's School

McKelvey School of Engineering

Author's Department

Electrical & Systems Engineering

Degree Name

Master of Science (MS)

Degree Type

Thesis

Abstract

Burst suppression is a clinical term describing a phenomenon in which the electroencephalogram (EEG) of a sedated patient produces behavior that switches between higher frequency and amplitude bursting to lower frequency and lower amplitude suppression. This phenomenon can be observed during general anesthesia, hypothermia, or in an otherwise induced coma state. In a clinical setting, this phenomenon is typically induced by sedation from a drug such as propofol (2,6-diisopropylphenol). The level of sedation can be quantified by something called the burst suppression ratio (BSR), which is defined as the amount of time that a patient’s EEG is in a suppressed state over the amount of time measured. One can vary this ratio by either increasing or decreasing the propofol infusion rate that the patient is given to bring them to a deeper or lighter state of sedation. By measuring the EEG data, one can form a closed loop feedback system where the EEG data is monitored for signs of burst suppression and the propofol is increased or decreased accordingly. Therefore, it becomes desirable to create models of this closed loop system to simulate the kind of behavior that would be expected from a clinical setting such as the one described. Many methods and experimental paradigms have been developed to address this problem including development of pharmacokinetic (PK) models that describe the dynamics of drug infusion in the body as well as signal processing methods for computing the burst suppression estimation such as the burst suppression probability. Some of these paradigms have been tested in rodent experiments, though human studies remain elusive. In this regard, simulations and detailed physiological modeling and control design can play a key role. This thesis seeks to add on the rich body of work that has been done thus far by incorporating a Schnider PK model with the Wilson-Cowan neural mass model to form a closed loop model which we can use as a basis for more detailed analysis which includes real-time burst suppression estimation as well as uncertainty modeling in both the patient’s physical characteristics (such as weight, height, age and gender) in addition to neurological phenomena such as the recovery and consumption rates of neurons during burst suppression behavior. By creating a conversion from the physiological parameters that describe the PK models to the dimensionless and more abstract parameters which guide the Wilson-Cowan equations, and implementing an actuator and burst suppression ratio estimation algorithm, we have effectively modeled the clinical setting with which the BSR is sought to be controlled. Thus, in this study we wished to show that with PID control, one could control this model at a nominal condition (i.e., the patient and neurological parameters which the gains were designed for) as well as at various uncertainty conditions that include both physical and neurological uncertainty, as described above. Using the Zeigler-Nichols tuning method, we were able to design gains to sufficiently control this system at set points of 0.8, 0.5 and 0.2 BSR over a simulation time of roughly 18 hours in both nominal, patient varied with noise added and with reduced performance when including patient variation and noise as well as neurological uncertainty. This time duration was chosen because it was convenient for the model’s time constants but also because it is representative of the time a patient may be sedated. The BSR ranges were chosen so as to show the closed loop system’s ability to maintain control at multiple levels of sedation. The reduced performance due to neurological uncertainty was due to the BSR estimation algorithm estimating a lower bound that was too high for the system to be controlled at a BSR of 0.2. The minimum BSR the system with added neurological uncertainty could be controlled to was 0.38, which is where the system held at during the portion of the trajectory that a BSR of 0.2 was commanded. During the achievable parts of the envelope, however, the control scheme worked with similar performance to that of the nominal case. This would suggest that an adaptive estimation algorithm needs to be developed to estimate the neurological deviations from the nominal case. Further, this suggests that if variations in the BSR of a patient due to neurological uncertainty is expected, then accurate estimation of these parameters are vital to reaching a robust solution in a real-time system.

Language

English (en)

Chair

Dr. ShiNung Ching, Electrical and Systems Engineering

Committee Members

Dr. James Feher, Dr. Sankalp Bhan

Share

COinS