Leaders from Texas Children’s, University of Alabama at Birmingham, and Johns Hopkins look forward to an immersive session at AIMed’s Global Summit.
AI-based tools built from comprehensive physiologic waveform datasets are key to delivering patient-centric care at scale. At AIMed’s Global Summit, Medical Informatics Corp. and Intel will be delivering an immersive session demonstrating how healthcare leaders are accessing waveform and other time-series data to close the loop from development to deployment of real-time, patient-centric AI.
Attendees will see how complete datasets improve the accuracy of model fitting, get the opportunity to work with sample models and data, and meet directly with world-renowned researchers and physicians to discuss their work through an interactive workshop featuring 5 hands-on stations.
Parag Jain, Associate Professor of Pediatrics at Texas Children’s Hospital, will be talking about a novel arrhythmia detection tool for children with congenital heart diseases that uses multimodal physiological waveforms including ECG and Central Venous Pressures.
“Alarm fatigue is a real life-threatening concern for critically ill patients and often adds to provider burnout in ICUs and ORs,” said Dr Jain. “Multimodal physiology-based alarms can lead to a more sensitive and specific alarm decreasing the false positive alarms and subsequent alarm fatigue.”
Ryan Melvin, Principal Data Scientist and Assistant Professor in the Department of Anesthesiology and Perioperative Medicine at The University of Alabama at Birmingham, will be discussing the necessary skills for data scientists to effectively engage in projects that use high resolution data collection and capture. Special attention will be given to the interaction with stakeholders such as clinician researchers, department leadership, IT resources, and vendor support personnel.
“The goal of my team is to use the infrastructure enabled by IT to address data-based questions through advanced analytical and predictive techniques. I want attendees to take away a checklist of what personnel, skills, software, and infrastructure is necessary to enable and succeed at Artificial Intelligence and Machine Learning work”, said Dr Melvin.
Melvin’s colleague Josh Hagood, Director of Information Technology at The University of Alabama at Birmingham, will be talking about the logistics of applied AI research projects in an operating room environment. In particular, how to work with organizational IT and other stakeholders in the OR in order to install automated, high resolution data acquisition systems without interrupting existing EMRs.
“The goal of my team is to take care of connecting all the things so that clinicians and data scientists can focus on what they do best”, said Hagood. “This session will provide a basic framework for how to go from limited or non-existent AI-ready data collection infrastructure to a system that transparently collects data for future research.”
James Fackler, Director of Safety, Quality, and Logistics at Johns Hopkins Pediatric Intensive Care Unit, will be discussing recent work on the prediction of cardiac arrest in children with sufficient warning to give clinicians time to better prepare for or even prevent cardiopulmonary resuscitation.
“There is substantial information, currently invisible to clinicians, within the routinely used but haphazardly collected critical care monitor data”, said Dr Fackler. “Analyzing the data will produce actionable insights to allow clinicians to mitigate complications.
“Attendees should leave understanding that these analytics are neither ‘exotic’ nor expensive. They can be implemented with what is now ‘off-the-shelf’ software that is linked to ‘every-day’ critical care monitors.”
Thinking about advice he would give to other organizations looking to deploy AI, Dr Jain said:
“Explainable AI is a powerful tool for addressing concerns due to the black-box nature of AI. Engaging in human-centered AI projects will allow easy adaptability from a clinician standpoint. Consider the investment in software that allows seamless integration and analyses of multiple data sources, with potential for rapid research and clinical deployment.”
“Start small”, said Mr Hagood. “Pick a small starting area with engaged investigators and high data density. Make friends. Medical devices have limited physical connections and humans have limited bandwidth for new projects. Bringing others into the conversation early pays big dividends when you run into these kinds of roadblocks. Think in systems. The endpoints won’t be connected on day 1 or even day 100 but have a basic plan of how the system could scale to more patients and how the products of AI could be presented back to clinical decision makers. You want to create an environment where lightning has more opportunity to strike.”
Dr Melvin added:
“Carefully walk the line between ‘too narrow’ and ‘too broad.’ Spend time thinking about what high-impact projects would take roughly one month to pilot and one year to complete. Rely on frontline, IT, AI, and clinical experts to say whether an idea fits this framework, and believe them when they yes or no! Worry about your team before you worry about the goal. To paraphrase Jim Collins (Good to Great), get the right people on the bus before you worry about where the bus should go. Read Data Science for Business by Provost and Fawcett, Scrum by Sutherland and Sutherland, and You Look Like a Thing and I Love You by Shane.”
“Decide what clinical problem is most important to solve”, concluded Dr Fackler. “Start getting grounded, not in the weeds of how the algorithms are created, but in what they can accomplish. Create the infrastructure to be able to acquire the necessary data.”
Session moderator Heather Hitchcock summarized, “Creating an algorithm is a great first step, and developing and deploying it back into clinical care completes the cycle. In order to close the loop you need the right team working with the right data. Then you need to have the right information sent with the right visualization, at the right time, to the right person, in the right location. Our workshop is designed to demonstrate that process from end-to-end and provide a mechanism for creative discussion and collaboration to help pave the way for the new era of data driven medicine and predictive care.
Developing and deploying real-time AI in patient care will take place at 1pm PT on Wednesday May 25, 2022, as part of AIMed’s Global Summit, taking place live and in person in San Francisco. Book your place now!
The post Developing and deploying real-time AI in patient care appeared first on AIMed.
Author: