After suffering a traumatic brain injury, patients are often placed in a coma to give the brain time to heal and allow dangerous swelling to dissipate. These comas, which are induced with anesthesia drugs, can last for days. During that time, nurses must closely monitor patients to make sure their brains are at the right level of sedation — a process that MIT’s Emery Brown describes as “totally inefficient.”
“Someone has to be constantly coming back and checking on the patient, so that you can hold the brain in a fixed state. Why not build a controller to do that?” says Brown, the Edward Hood Taplin Professor of Medical Engineering in MIT’s Institute for Medical Engineering and Science, who is also an anesthesiologist at Massachusetts General Hospital (MGH) and a professor of health sciences and technology at MIT.
Brown and colleagues at MGH have now developed a computerized system that can track patients’ brain activity and automatically adjust drug dosages to maintain the correct state. They have tested the system — which could also help patients who suffer from severe epileptic seizures — in rats and are now planning to begin human trials.
Maryam Shanechi, a former MIT grad student who is now an assistant professor at Cornell University, is the lead author of the paper describing the computerized system in the XXX issue of the journal PLoS Computational Biology.
Tracking the brain
Brown and his colleagues have previously analyzed the electrical waves produced by the brain in different states of activity. Each state — awake, asleep, sedated, anesthetized and so on — has a distinctive electroencephalogram (EEG) pattern.
When patients are in a medically induced coma, the brain is quiet for up to several seconds at a time, punctuated by short bursts of activity. This pattern, known as burst suppression, allows the brain to conserve vital energy during times of trauma.
As a patient enters an induced coma, the doctor or nurse controlling the infusion of anesthesia drugs tries to aim for a particular number of “bursts per screen” as the EEG pattern streams across the monitor. This pattern has to be maintained for hours or days at a time.
“If ever there were a time to try to build an autopilot, this is the perfect time,” says Brown, who is a professor in MIT’s Department of Brain and Cognitive Sciences. “Imagine that you’re going to fly for two days and I’m going to give you a very specific course to maintain over long periods of time, but I still want you to keep your hand on the stick to fly the plane. It just wouldn’t make sense.”
To achieve automated control, Brown and colleagues built a brain-machine interface — a direct communication pathway between the brain and an external device that typically assists human cognitive, sensory or motor functions. In this case, the device — an EEG system, a drug-infusion pump, a computer and a control algorithm — uses the anesthesia drug propofol to maintain the brain at a target level of burst suppression.
The system is a feedback loop that adjusts the drug dosage in real time based on EEG burst-suppression patterns. The control algorithm interprets the rat’s EEG, calculates how much drug is in the brain, and adjusts the amount of propofol infused into the animal second-by-second.
The controller can increase the depth of a coma almost instantaneously, which would be impossible for a human to do accurately by hand. The system could also be programmed to bring a patient out of an induced coma periodically so doctors could perform neurological tests, Brown says.
This type of system could take much of the guesswork out of patient care, says Sydney Cash, an associate professor of neurology at Harvard Medical School.
“Much of what we do in medicine is making educated guesses as to what’s best for the patient at any given time,” says Cash, who was not part of the research team. “This approach introduces a methodology where doctors and nurses don’t need to guess, but can rely on a computer to figure out — in much more detail and in a time-efficient fashion — how much drug to give.”
Brown believes that this approach could easily be extended to control other brain states, including general anesthesia, because each level of brain activity has its own distinctive EEG signature.
“If you can quantitatively analyze each state’s signature in real time and you have some notion of how the drug moves through the brain to generate those states, then you can build a controller,” he says.
There are currently no devices approved by the U.S. Food and Drug Administration (FDA) to control general anesthesia or induced coma, but there is a device available in Europe and South America, based on an algorithm that uses the patient’s EEG to compute an index on a 100-point scale. However, that system keeps the patient’s brain activity within a very wide range and does not allow for precise control, Brown says.
The MIT and MGH researchers are now preparing applications to the FDA to test the controller in humans.
The research was funded by the National Institutes of Health through a Pioneer Award and a Transformative Research Award.