Millions of people are admitted to hospital intensive care units following traumatic brain injuries, strokes, brain bleeds and other conditions. Because the initial injury can cause other changes in the brain that lead to further damage, patients frequently are monitored through the use of an external ventricular drain — a device that measures the pressure in brain tissue and its surrounding fluid. If the pressure spikes, physicians can quickly treat the rising pressure by draining fluid or giving medicines or other treatments. 

Being sensitive to this timing can dramatically alter patient outcomes, explains Rohan Mathur, M.D., a neurocritical care physician at Johns Hopkins. 

“Patients can come to the hospital with a small injury to their brain, but because of the subsequent swelling and a number of other mechanisms, there can be a vicious cycle that leads to further injury,” Mathur explains. “If you don’t act to control those secondary mechanisms, this person could have one of two vastly different outcomes: they could have a devastating brain injury or they could walk out of the hospital with minimal impact. It’s very high stakes, and a lot of our decisions are extremely time-sensitive.”

Precision Medicine at Work

Mathur and colleagues, including Jose Suarez, M.D., director of the Division of Neurosciences Critical Care; Peter Dziedzic, director of the Center of mHealth and Innovations; and Lin Cheng, a data scientist, are developing a way to predict spikes in brain pressure before they occur, to intervene proactively. Working with collaborators, they created a machine learning algorithm called CIC-L (“cycle”) that screens data collected by external ventricular drains to make it usable for building predictive models. The model was trained and tested using data from about 60 patients treated at The Johns Hopkins Hospital.

Their research was presented at the Neurocritical Care Society international annual conference and won the Best Scientific Abstract award. The team is now working with applied mathematicians at the Johns Hopkins Whiting School of Engineering to create predictive models of intracranial pressure crisis using machine learning and artificial intelligence tools. 

“What we’re trying to do is create early warning systems for these crisis periods when it’s so vital that you act quickly,” Mathur says. “Our overall vision is that these pursuits will result in multiple early warning systems to revolutionize how care is delivered in all ICUs and lead to new breakthroughs in monitoring and treatment.”