CP 10: Artificial Intelligence for Dynamic, Individualized CPR Guidance: AID CPR

Collaborating Investigator:

Michelle Nasal, The Ohio State University

Funding Status:

K08HL168330

NIH/NHLBI

09/01/23 – 08/31/28

Associated with:

TR&D3

Publications
Significance
Approach & Push Pull Relationship

TR&D3: Translation

* Artificial Intelligence for End Tidal Capnography Guided Resuscitation: A Conceptual Framework

Authors:

Anna L. Trella, Kelly W. Zhang, Inbal Nahum-Shani, Vivek Shetty, Finale Doshi-Velez, Susan A. Murphy

Publication Venue:

Algorithms in Decision Support Systems

Publication Date:

July 22, 2022

Keywords:

reinforcement learning, online learning, mobile health, algorithm design, algorithm evaluation

View Full Paper

Related Project:

CP 2, CP 8

Abstract:

Online reinforcement learning (RL) algorithms are increasingly used to personalize digital interventions in the fields of mobile health and online education. Common challenges in designing and testing an RL algorithm in these settings include ensuring the RL algorithm can learn and run stably under real-time constraints, and accounting for the complexity of the environment, e.g., a lack of accurate mechanistic models for the user dynamics. To guide how one can tackle these challenges, we extend the PCS (predictability, computability, stability) framework, a data science framework that incorporates best practices from machine learning and statistics in supervised learning to the design of RL algorithms for the digital interventions setting. Furthermore, we provide guidelines on how to design simulation environments, a crucial tool for evaluating RL candidate algorithms using the PCS framework. We show how we used the PCS framework to design an RL algorithm for Oralytics, a mobile health study aiming to improve users’ tooth-brushing behaviors through the personalized delivery of intervention messages. Oralytics will go into the field in late 2022.

TL;DR:

Online RL faces challenges like real-time stability and handling complex, unpredictable environments; to address these issues, the PCS framework originally used in supervised learning is extended to guide the design of RL algorithms for such settings, including guidelines for creating simulation environments, as exemplified in the development of an RL algorithm for the mobile health study Oralytics aimed at enhancing tooth-brushing behaviors through personalized intervention messages.

CP10 seeks to develop personalized interventions by integrating End-tidal carbon dioxide (ETCO2) capnography into Out-of-hospital cardiac arrest (OHCA) resuscitation.

Out-of-hospital cardiac arrest (OHCA) is a dynamic process that requires new interventions to improve outcomes. End-tidal carbon dioxide (ETCO2) measurement is a tool that is widely recognized, easy to use, and can potentially provide real-time insights into ongoing resuscitation efforts; however, it has yet to be applied to individualized medicine. Our overall hypothesis is that integrating ETCO2 capnography into OHCA resuscitation will improve outcomes. Using innovative signal processing and machine learning methods, we will identify a wide range of resuscitation quality characteristics over resuscitation, their relation to individual patient characteristics and predictability of OHCA outcomes. These goals will be accomplished via the following aims: Aim 1. Determine the influence of resuscitation interventions on real-time physiologic dynamics and outcomes in OHCA. Aim 2. Establish the influence of individual patient characteristics on the real-time physiologic dynamics and OHCA outcomes. Aim 3. Develop a novel cardiac arrest resuscitation strategy based upon real-time individualized physiologic dynamics. We will create a large repository of cardiopulmonary resuscitation process data encompassing data from over 5300 adult OHCA. This work will define intra-arrest ETCO2 dynamics over resuscitation to allow for the development of guided resuscitation efforts, and the resultant data will provide a solid foundation for future hypothesis-driven research. Dr. Nassal’s training plan encompasses both formal didactics and experiential training with experienced mentors and collaborators that will develop a skillset in both signal processing and equitable artificial intelligent driven algorithms. The team has extensive experience in using machine learning and multimodal signal processing for classification and predictions in resuscitation. This training program will develop a unique skillset in advanced cardiac signal processing; artificial intelligence, including equitable machine learning processing; and expertise in the application of these skills to develop dynamically guided resuscitation strategies that few other physician-scientist possess.

TR&D3 is collaborating with CP10 to develop individualized digital twin model for the OHCA patient allowing real time interpretation of ETCO2. This will allow simulation of alternative resuscitation strategies (e.g. through optimizing the timing of epinephrine dosing).

CP10 challenges TR&D3 to develop digital twin models of cardiovascular system under stress from a unique field dataset of 2,500 patients in cardiac arrest with ECG, ETCO2, Chest Compression signals collected in the field. While digital twin models have been developed before for cardiovascular system in homeostasis, these are not applicable to patients undergoing cardiac arrest.

TR&D3 will help CP10 in assessing resuscitation intervention strategies using real-time physiologic dynamic and patient specific characteristics through a digital twin model.