Share this article on:

UP & COMING: M. Brandon Westover, MD, PhD On Automated Predictors of Delayed Stroke After Subarachnoid Hemorrhage

Rukovets, Olga

doi: 10.1097/01.NT.0000419608.00623.1d

M. Brandon Westover, MD, PhD, is mesmerized by the “complex, information-processing” system that is the human brain. An engineer by training, Dr. Westover uses his quantitative background to frame his neurological research. Whether in stroke, epilepsy, or anesthesia delivery, his projects are aimed at reducing human error and increasing accuracy with the use of computational and automated methods and analyses.

In 2006, he completed his MD and PhD in physics at Washington University in St. Louis, MO. Dr. Westover then trained in neurology at Massachusetts General Hospital (MGH) and Brigham and Women's Hospital in Boston (2007–2010). Currently, he is an Epilepsy Fellow in the CashLab, and will become an instructor in neurology at Massachusetts General Hospital (MGH) and Harvard Medical School in Boston.

A recipient of the 2012 American Brain Foundation Clinical Research Training Fellowship, Dr. Westover hopes to use automated EEG signals to detect delayed stroke after subarachnoid hemorrhage.

Back to Top | Article Outline


We are studying patients with aneurysmal subarachnoid hemorrhage. Many patients who have a burst aneurysm die within the first 24 hours. Individuals who survive the hemorrhage and stabilize typically end up staying in the neurological ICU for about two weeks, and the primary risk that they face is having a stroke later. The reasons these delayed strokes happen are not entirely understood — part of it may have to do with the blood vessels becoming irritated and constricting and limiting blood flow. The other major factor may be cortical spreading depressions; these events are also common in patients with migraine auras, but in an injured brain are likely not as benign as they are in otherwise healthy migraine patients. So one of those two things or both, as well as factors we don't know about, cause delayed strokes.

My project is about trying to detect these strokes before it's too late to do much about them, mainly using EEG signatures of an impending stroke that we hope we can extract automatically from the EEG signals. Several small published studies suggest that the EEG often does have some changes beforehand which, if you know how to recognize them, signal that strokes are going to happen — sometimes one day, or even two days before you can detect them by standard means like serial neurological exams. In the ICU we typically wake people up every one to two hours for a basic neurological exam to see if they have any new deficits, and do ultrasounds of the major arteries in the head about once a day to see whether blood velocity has increased, a sign that blood vessels are constricting. Many of us who spend time clinically monitoring EEG over the course of the ICU stay in these patients are convinced that EEG picks up signs of impending stroke before either clinical exams or ultrasound methods can detect them.



Back to Top | Article Outline


Prevention is the ultimate goal. But for my own immediate work, the idea is that we already have things that we can do to reverse or prevent strokes — or at least to minimize their size — but we're limited in how much good those existing interventions can do by the inability to detect strokes at an early stage.

The one “good” thing about this disease — if there's anything good about it — is that these delayed strokes happen in a supervised setting and they seem to occur more slowly than typical ischemic strokes, so we would have time to do something if we had reliable ways to detect them early on.

Back to Top | Article Outline


There are a few reasons: one is that this is among the only situations in which you can study stroke and have someone under monitoring a stroke is happening. From this perspective, it's a very interesting disease, in contrast to strokes that people just come with, after it's too late to meaningfully intervene. Second, delayed strokes are a huge source of disability for patients with subarachnoid hemorrhage — many would be fine if they didn't have these late strokes. So there's an opportunity here to do some real good.

The other draw for me is that I'm an engineer by training and this is an interesting technical challenge. My goal is to try to automate the process in a way that doesn't produce too many false alarms. In graduate school, I worked in a lab that does a lot of signal processing and statistical pattern recognition, so I'll be hopefully to leverage this experience this research.

Back to Top | Article Outline


My primary mentor is Sydney Cash. I've been working with him since I was a resident and he's really great. He's always enthusiastic, and we usually come away from our frequent conversations with 20 extra ideas of neat new things to do. He studies EEG signals recorded from the surface of the brain in patients with epilepsy, trying to understand the mechanisms of seizures.

I've analyzed EEG signals with him, and we actually just completed a project together that we're about to submit for publication looking at a slightly offbeat idea — we found a way to convert the EEG into a set of “motifs”, or basic wave forms. The idea was to see if EEG can be decomposed into repeating elements like words in a language. We ultimately succeeded — it looks like it's possible to decompose EEG waves into just a couple of thousand different wave forms which turn out to mainly obey a statistical law that's characteristic of word frequency in most spoken languages called Zipf's law.

My other primary mentor is Steve Greenberg, who is a mensch — he's a wise guy. He's a very experienced stroke researcher, and I look to him to criticize my ideas, always in a good-humored way. I worked with him on a decision-analysis problem as a resident about whether patients who had bleeding in the brain and who also have heart problems should take statins, since on the one hand the statins may prevent heart attacks, but on the other hand, they may make it more likely that you'll have another bleeding episode in the brain.

So we made a quantitative model of that situation to weigh the risks and benefits, in hopes of understanding what the risk-benefit calculus ultimately favors, and we found out that — at least for people who have had hemorrhages in the hemispheres of the brain rather than in the deeper parts of the brain – that the probably shouldn't be taking statins. We had a nice fifteen minutes of fame after we published that finding, because statins are so widely used, and this finding was a bit of a surprise. I think it's fair to say that Steve and I think changed the practice of neurologists regarding prescription of statins for that particular set of people who have had hemorrhages before, a very satisfying experience.

Back to Top | Article Outline


I would probably have been happy doing anything quantitative within neurology. On the other hand, no offense to people who study stroke (which I suppose includes me!), but in some ways they tend to think of the brain as any other organ, while missing out on enjoying the beautiful electrical rhythms and patterns that make the brain what it is. For me, epilepsy gets closest to what I find most fascinating about the brain.

I am especially interested in seizures that happen in critically ill patients, and other events that can be monitored with EEG, like in this project that I'm doing, where you can do some real good for these patients — and I find that pretty satisfying.

There also happen to be a lot of people with engineering interests in epilepsy. So, I get to rub shoulders with other engineering-minded colleagues, which I enjoy. Syd Cash has influenced me a lot in my career choice. He's doing a lot of the sort of research that I enjoy doing, and that helped draw me in.

Back to Top | Article Outline


With my three kids (ages nine, seven, and four), I have recently been reading a book called Poop: A Natural History of the Unmentionable for bedtime stories. I also speak Mandarin Chinese; I lived in Taiwan from 1992 to 1994.

Back to Top | Article Outline


I am working with colleagues over at MIT on a method to automatically deliver anesthetic drugs to patients that with refractory seizures who need treatment with pharmacological coma. Essentially, we're working on a way to make it so that nurses don't have to be the ones with the burden of handling anesthesia for these patients. We expect that this will allow better control. Typically when we deliver anesthesia and people are in charge, we end up too often underdosing or overdosing patients at least some percentage of the time — both of which are bad for the patients. We have our system working already in an animal model, and are hoping to start testing in human patients soon.

The American Brain Foundation, the foundation of the American Academy of Neurology, is one of the largest providers of neurology research grants in the United States. The Foundation supports the most promising research & education to discover causes, improved treatments, and cures for brain and other nervous system diseases. To apply for the Clinical Research Training Fellowship program, visit and





Listen as M. Brandon Westover, MD, PhD, a 2012 recipient of the American Brain Foundation Clinical Research Training Fellowship, discusses how he hopes to use automated EEG signals to detect delayed stroke after subarachnoid hemorrhage:

© 2012 American Academy of Neurology