Computer Science Colloquia
Friday, May 31, 2013
Advisors: John Lach and John A. Stankovic
Attending Faculty: Joanne Bechta Dugan (Chair), Gabriel Robins and Stephen Patek
3:00 PM, Rice Hall, Rm. 242
Ph.D. Proposal Presentation
Enabling Patient Safety Analysis of Sense-Only Wearable Medical Body Sensor Networks
Sense-only body sensor networks (SOBSNs) are an emerging technology that have the potential to impact the healthcare industry (and other industries as well). In the healthcare setting, SOBSNs can be used in mobile monitoring scenarios to provide otherwise-unobtainable information about patients to clinicians to aid in research as well as clinical practice. SOBSNs are part of a medical system and hence have patient safety implications. Patient safety is obvious for parts of the medical system that delivers therapies to the patient. The issue is subtle when the aim of the system (in the SOBSN case) is to provide information to inform decision making. This work is aimed at enabling designers and regulators to examine SOBSNs with respect to their patient safety implications.
This work is part of a long-term effort to equip SOBSN designers and regulators with the insights necessary for producing SOBSNs that meet their intended (safety) goals. It lays the foundation by addressing formally three fundamental questions required to enable analysis of SOBSNs for patient safety: 1) what are the patient safety implications for SOBSNs 2) how does the nature of SOBSNs (and its operational environment) contribute to these implications, and 3) how can SOBSNs be analyzed with respect to the identified safety implications.
The thesis of this work is that by modeling the SOBSN and its operational environment as a set of interacting processes, we can develop a notion of safety consistent with the fundamentals of system safety engineering (while addressing the limitations of scope of application of current techniques designed mostly for non-SOBSN/medical domains), identify a general class of hazards for SOBSNs, provide insight into causal factors for these hazards, and develop analysis techniques that allow us to check at design time whether a SOBSN with particular properties will meet the particular patient safety constraints.
Models at three different levels of abstraction will be used in this work. The highest level will be used to develop a precise notion of safety and to develop algorithms for analyzing (this model of) SOBSNs based on this notion of safety. The second level of abstraction will be used to capture SOBSN components, their interaction with each other and the external environment, to reveal how these interactions contribute to patient safety issues. Lastly, two lower-level models will be developed to show how model-based techniques can aid in explorations at design time to reveal issues related to patient safety, and to show the validity of the other two models. In addition, the techniques developed in this work will be applied to continuous glucose monitoring (and other examples as is feasible) to demonstrate their efficacy and validity.