How Machine Learning Can Personalize, Streamline the Clinical Setting
Machine learning may provide the improved time efficiency and feedback clinicians need to bolster personalized treatment planning in the clinical setting.
Source: Getty Images
By - Many healthcare organizations are moving to adopt data analytics and artificial intelligence (AI) tools, but when it comes to more advanced computing and analytics technologies like machine learning (ML) they struggle to distinguish between current practical applications and the hype around potential future uses in medical research and clinical care.
Amanda Randles, PhD, the Alfred Winborne and Victoria Stover Mordecai assistant professor of biomedical sciences at Duke University, works in biomedical simulation and high-performance computing and focuses on the development of new computational tools to explore the localization and development of human diseases, specifically 3D blood flow modeling and simulation.
On Healthcare Strategies, Randles provided insights into how these and other ML approaches can be used to improve treatment planning and allow real-time feedback for clinicians.
Listen to the full podcast to hear more details. And don’t forget to subscribe on iTunes, Spotify, or Google Podcasts.
“We're trying to understand how different treatment options may affect the hemodynamics of a specific patient,” Randles explained to Healthcare Strategies. “The crux is creating almost a digital twin. It's a personalized view of the 3D model of someone's anatomy. We often look at adult coronary patients to try to determine if they need a stent or if they need intervention. We look at pediatric cardiology patients to determine what's the best treatment option for those patients.”
These high resolution, 3D models of the arteries that the team uses during blood flow simulations provide a wealth of data to inform treatment planning, but they are computationally intense and require significant computational resources, meaning that they can take anywhere from a few minutes to 12 hours to run a simulation. ML can help boost efficiency during the process, Randles stated.
“We're starting to use machine learning to augment that need for the large-scale computation, to speed it up and work as a proxy to allow us to test [multiple treatment options], instead of just trying one,” she said. “Can we predict the results of 10, 15 different treatment options or even five treatment options to allow the doctor, before they ever go into the operating room, to test out different stents that they may place or different locations you may place the stent or different interventions?”
Similarly, the researchers also leverage the tech to evaluate how a particular patient may respond to treatment under different physiological states. These insights, delivered in real-time in many cases, can then provide clinicians with feedback and clinical decision support about a potential treatment plan while seeing the patient in-clinic or before entering the operating room, which could help clinicians facing burnout.
“In our work, we're trying to see this as a way of giving the clinicians [more clinical decision information] and guiding the clinicians, not replacing the clinicians,” Randles said. “It's just giving them more information so they can go in and make a better decision and be more informed in that treatment planning process. I think it's very important to see it as [we’re] aiding the clinician’s decision. We are in no way trying to replace and dictate to the clinician what they should be doing.”
