Convex Optimization Methods for System Identification with Applications to Noninvasive Intracranial Pressure Estimation

Speaker: Cameron Allan Gunn
Affiliation: Advisor: Prof. Lieven Vandenberghe

Abstract:  This talk presents new convex optimization methods for data-driven signal estimation and analysis, motivated by applications to noninvasive intracranial pressure (NICP) estimation, a problem that arises in critical care.

Traumatic brain injuries account for 275,000 hospitalizations in the United States each year. It is important for some of these patients’ intracranial pressure—the pressure of fluid in the skull—to be monitored while they are in intensive care, to assess their condition. However, this measurement process requires an invasive surgical procedure, an impediment that has prompted research on NICP estimation. This is a hard, unsolved problem that is made challenging by both the unreliability of corrupted signal data, and a large inter-patient variability across the population that limits the ability to compare new patients to past patients.

The methods in this talk leverage ideas from linear dynamical systems and system identification, and recent advances in trace norm minimization and large-scale convex optimization methods. Three key contributions to the field will be discussed. The first is a method for artifact rejection and missing data imputation, which mitigates the effect of corruptions in physiological signals. The second is a method to perform clustering on linear dynamical systems, so that patients can be grouped by their signal dynamics. The third is a novel approach to NICP estimation that combines a patient’s noninvasive measurements with a dictionary of retrospective patient models. The methods in this talk have a broad applicability beyond NICP estimation and biomedical signal processing.

Biography: Cameron Allan Gunn is a Ph.D. candidate in the UCLA Electrical and Computer Engineering Department under the mentorship of Professor Lieven Vandenberghe. He is interested in the application of convex optimization and dynamical systems to real-world applications, with experience in medical devices, advertising, and electronic trading. Cameron is the recipient of a Fulbright scholarship, and holds an MS (Electrical Engineering) from UCLA, and a BE (Mechatronics Engineering) with First Class Honours from the University of Canterbury in Christchurch, New Zealand.

For more information, contact Prof. Lieven Vandenberghe ()

Date(s) - May 30, 2018
1:00 pm - 3:00 pm

E-IV Maxwell Room #57-124
420 Westwood Plaza - 5th Flr. , Los Angeles CA 90095