Can you see what I see? Maybe not, but the computer can

by Tom Ulrich on July 2, 2012

What if you could just look at someone’s face and tell how fast his or her heart is beating?

The question isn’t as far-fetched as you might think. The movement of our beating heart inside our chest can in fact reveal itself on the surface of the skin, albeit too faintly for our eyes to see. But as you can see in this video, it’s not too faint for a computer (fast forward to 1:25 and 3:18):

Donna Brezinski, MD, of Boston Children’s Hospital’s Division of Newborn Medicine and Neonatal Intensive Care Unit (NICU), recently described the system used to make that video at one of the hospital’s Innovators’ Forums (a series of monthly talks hosted by Boston Children’s Innovation Acceleration Program). It uses computer-based video processing to make a pulse look like it’s bulging on a person’s wrist, or to amplify changes in skin color as freshly oxygenated blood gets pumped through the body.

As Brezinksi told the audience, the system—developed with collaborators at MIT—relies on what is called Eulerian video magnification. It works like this: A computer takes a live video feed and breaks it down into both space- and time-based components. It then looks for things that change over time—skin color or motions of breathing, for example—and puts the video back together, amplifying those subtle changes so that we humans can see them.

“The computer lets us exaggerate movements or changes in the video that we want to see and filters out the background,” Brezinski explains. “It works in real time, and can use video supplied through the camera of an average laptop, though the sensitivity is better with better cameras.”

Why would one want a system like this? For one thing, it allows for touch-free monitoring. “Babies who come to the NICU, especially premature babies with very low birth weights, can have very sensitive skin,” Brezinski says. “We often use tape to hold the leads of our heart rate and other monitors in place, but that tape can severely irritate their skin. Also, if the babies roll around, there’s a risk that they could get caught in the wires that go from the leads to the monitors.”

With computer-aided vision, we might one day be able to see a person's vital signs by magnifying changes to their skin. (meddygarnet/Flickr)

To assess the system, Brezinski and her team set up video cameras over the cribs of 10 babies in a NICU staffed by Boston Children’s staff at Winchester Hospital. “When we compared the system’s estimate of babies’ heart rates based on skin color changes with what was actually being displayed on their monitors, we found that the system was pretty close, within a few beats per minute,” Brezinski says. “We found that we could also get good amplification of breathing movements, which could help us find signs of apnea or obstruction.”

The system, which was featured in a recent paper in the computing journal SIGGRAPH, is not without its limits, though. “Large movements by the baby and dim lighting both pose challenges for the system,” Brezinski says. “It’s something we need to work on.”

As Brezinski pointed out in her presentation, this system illustrates one of two pathways for innovation and technology development in medicine: technological push, a concept-driven approach where innovators take an existing technology (in this case, video-based monitoring of sick newborns, something Brezinski has been refining for a while) and find new applications for it. The other is clinical pull, where an innovation comes about because innovators see a specific need and develop a technology to fill it — exemplified by the effort by her colleague Farhad Imam, MD, PhD, to develop a way to make PICC lines to shine through the skin.)

“We’d been working for some time on ways to use video to monitor babies on their way to the NICU in an ambulance and thought, ‘If we can capture and transmit video now, what else can we do?’” Brezinski says. “And while we’re testing video enhancement in the context of the NICU, there could be applications in sleep medicine, cardiology, and other areas of medicine where you would want to track changes of some parameter over time.”

Leave a comment

Previous post:

Next post: