Sarah Taft
Pronouns: she/her
Research Mentor(s): Cindy Hsu
Research Mentor School/College/Department: Emergency Medicine / Medicine
Program:
Authors: Sarah Taft, Finn Burke, Traci Cramer, Zachary Sharpe, Nick Greer, Alexis Davis, Courtney Dennis, Lisbeth Hernandez, M. Hakam Tiba, Cindy H. Hsu
Session: Session 1: 9:00 am – 9:50 am
Poster: 12
Abstract
Sudden cardiac arrest is one of the leading causes of death in the United States, with an overall mortality rate of 90%. However, previous cardiac arrest studies have failed to translate to clinical trials because of the lack of clinically relevant large-animal cardiac arrest models. In order to navigate these issues, this study uses data from the ongoing development of a swine model of neuroprotection in cardiac arrest to create a tool for the analysis of systolic and diastolic pressures during mechanical cardiopulmonary resuscitation (CPR). In particular, diastolic pressure is an important measurement in CPR because of its relation to coronary perfusion pressure. During mechanical CPR, devices, such as the LUCAS, create negative pressure in the vessels. This causes current pressure monitors to detect the local minimum value, the negative pressure, as the diastolic pressure instead of the true value. The main focus of this study is to correct this error through the development of a neural network that accurately measures diastolic pressure during mechanical CPR. This is achieved through the creation of a short-term swine cardiac arrest model. To begin, female and male swine were randomized by block: split into groups of male and female, and randomly assigned to 0 minutes, 5 minutes, or 10 minutes of ventricular fibrillation. Then, they were subjected to anesthesia, intubation, instrumentation, and baseline sampling. Ventricular fibrillation was then induced for the previously assigned duration. Then, they received mechanical CPR interrupted by defibrillations every 2 minutes until the subject achieved return of spontaneous circulation (ROSC). During CPR and the following post-ROSC 24-hour monitoring period, pressure measurements were collected in order to compile data for the development of the algorithm. Following the monitoring period, the data is manually labeled and used to train the model. Currently, the study is in the development phase, which means data is actively being collected. Preliminary results show minimal differences in performance between model architectures for a given task: compression diastolic and systolic pressure values or spontaneous diastolic and systolic pressure values. Overall, in the future, the study hopes to further develop and validate a tool for the analysis of arterial pressure values and mechanical CPR quality.
