projects

Electroceuticals

There is considerable interest in leveraging electrical stimulation as a medical therapeutic. This has led to an interest in “electroceutical” therapies. While there has been a lot of success and FDA-approved devices (including deep brain stimulation for Parkinson’s disease, vagus nerve stimulation for epilepsy, retinal ganglion stimulation for retinopathy), the underlying mechanisms are not fully understood. Much of the focus today is on the use of rectangular waveforms, but our lab hypothesizes that non-traditional, customized waveforms can produce more energetically-efficient and selective stimulation. We are working on developing novel algorithms to find these more efficient waveform, as well as discovering the mechanisms underlying how these non-traditional waveforms are unlocking different access mechanisms for selective stimulation.

Falls and Gait Stability

Falls in the elderly are associated with significant morbidity and mortality. While numerous fall detection devices incorporate AI and machine learning, many of these algorithms have been trained using simulated falls among young healthy volunteers. Video capture studies among the elderly have shown that falls are more often associated with weight transfer (e.g., during movement transitions). We have been focused on developing a contactless, markerless camera system that can track joint positions of individuals as they move. Our goal is to use this camera system to study kinematic movements and ultimately develop a model of human balance and gait stability. Our hope is that we can develop a predictive algorithm to identify when a person’s risk of falling dramatically increases, allowing for an early warning system preventing falls as opposed to only detecting falls.

Stroke and Emergency Transport

Among emergency transport services, there have been a number of interesting questions arising from the optimal transport strategy for patients suspected of having a stroke. These questions arise due to the fact that time is brain. The longer it takes for a stroke victim to be reperfused, the more brain matter is lost. In partnership with Austin EMS and the Seton hospital network in Austin, we have been examining both from a data-driven approach, as well as developing our own physiologically-driven model, the different transport strategies and their impact in Travis County. Our goal is to be able to provide insights for public health and policy officials regarding how to optimize transport strategies as well as the cost-effectiveness of various proposed improvements to the local stroke care network (including mobile stroke units and upgrading hospital certifications).

Predictive Analytics in the ICU

The vast amounts of time-dependent physiological data being routinely collected in the intensive care unit provides us with an extremely rich data set. We have two projects in this area:

Preterm Infants in the NICU

With preterm infants (born before 37 weeks), their nervous system is not fully developed and thus their cariorespiratory system is not as robust to exogenous stimulation. These infants are known to need respiratory support through both medications (e.g., caffeine) as well as oxygen supplementation. There are a number of interesting research questions exploring how continuous physiological signals can be used to identify when the patient is ready for extubation, being taken off caffeine, and ultimately when they are ready to be discharged home.

Delirium Identification in the ICU

While delirium has been shown to be a major factor regarding poorer health outcomes in the hospital, it is often undetected or detected too late. There have been a number of studies that have suggested that physiological changes may be associated with delirium, and as such we are exploring the use of both in-hospital monitoring units as well as wearable devices to capture these physiological changes, with the goal of developing an automated delirium prediction and detection system.