Intelligent ICU monitoring for patients in status epilepticus: BurSIn

The BurSIn system, in development, interprets EEG data along several key parameters and accurately identifies burst and suppression patterns.
The BurSIn system, in development, interprets EEG data along several key parameters and accurately identifies burst and suppression patterns.

Status epilepticus, a life-threatening form of persistent seizure activity in the brain, is challenging to treat. It requires hospitalization in an intensive care unit, constant monitoring and meticulous medication adjustment. An automated, intelligent monitoring system developed by clinicians and engineers at Boston Children’s Hospital could transform ICU care for this neurological emergency.

Typically, children in status epilepticus are first given powerful, short-acting seizure medications. If their seizures continue, they may need to be placed in a medically induced coma, using long-acting sedatives or general anesthetics. “The goal,” explains biomedical engineer Christos Papadelis, PhD, “is to supply enough sedating medication to suppress brain activity and protect the brain from damage, while at the same time avoiding over-sedation.”

Walking this fine line requires the brain’s electrical activity to be continuously monitored with an electroencephalogram (EEG), via a set of electrodes attached to the child’s scalp. An ICU physician or nurse periodically reviews the EEG data, looking for a visual pattern of spikes or “bursts” alternating with relatively diminished activity. This burst suppression pattern (BSP) indicates a state of induced coma. Clinicians then adjust sedatives based on changes in the ratio between burst periods and suppression periods.

“The identification of this pattern is right now done manually,” says Papadelis. “The nurses or physicians go into the room and physically examine the EEG output.” However, because this system necessitates frequent monitoring, visual interpretation of EEG data and manual drug titration, it is prone to inefficiency and human error.

When Tobias Loddenkemper, MD, an epilepsy specialist, and Robert Tasker, MB, BS, MD, director of Boston Children’s Neurocritical Care Program, approached him with this problem, Papadelis knew it would lend itself to a technological solution. “The ideal setup,” he says, “is to have a machine that could measure these burst suppression patterns and then titrate the sedating drugs accordingly.”

The team of Boston Children’s clinicians and engineers created intelligent software that takes the first step toward automated monitoring: the Burst Suppression Index (BurSIn) algorithm, named for the brainwave pattern it detects. Papadelis and his team taught the system how to read EEGs and identify burst and suppression periods by showing it numerous examples from a variety of pediatric patients. “Neural network technology is a system that learns by itself, a kind of artificial intelligence system,” explains Papadelis.

BurSin eventually learned to detect burst suppression patterns accurately and reliably in new patients. Importantly, Papadelis notes, this is the only such system that has been specifically tailored for pediatric patients with status epilepticus.

The first phase of the BurSIn project, supported by an Innovestment Grant from Boston Children’s Innovation Acceleration Program, was presented at the American Epilepsy Society Annual Meeting in Seattle this month. Papadelis and his team are now in the process of building BurSin into an active software system that would be able to read patient EEGs in real time and provide supplemental information to clinicians. They hope to have the active system ready for initial testing in the ICU in February 2015. The next step would be a clinical trial to see how effective BurSin is at monitoring compared to clinicians.

Ultimately, Papadelis would like to design a system that not only interprets the EEG but also automatically adjusts medication to maintain an appropriate and safe level of sedation. For now, his team is very encouraged by the early results.

That team also includes Chiran Doshi, BSc, MSc, a technologist in Boston Children’s Hospital’s BabyMEG Lab who developed the software. Other members include Ellen P. Grant, MD, PhD, a pediatric neuroradiologist, Sigride Thome, MD, a pediatric neurologist, and Saba Jafarpour, MD, a clinical research fellow.

“This project was an achievement in teamwork, bringing together needs from multiple departments,” says Papadelis. “Our physicians identified the clinical problem, and as biomedical engineers, we were then able to develop a technological solution.”