How GPUs Help Eye Surgeons See 20/20 in the Operating Room

Editor’s note: This is one in a series of profiles of five finalists for NVIDIA’s 2016 Global Impact Award, which provides $150,000 to researchers using NVIDIA technology for groundbreaking work that addresses social, humanitarian and environmental problems.

Performing ocular microsurgery is about as hard as it sounds — and, until recently, eye surgeons had practically been flying blind in the operating room.

Doctors use surgical microscopes suspended over a patient’s eyes to correct conditions in the cornea and retina that lead to blindness. These have limited depth perception, however, which forces surgeons to rely on indirect lighting cues to discern the position of their tools relative to sensitive eye tissue.

But Joseph Izatt, an engineering professor at Duke University, and his team of graduate students are changing that. They’re using NVIDIA technology to give surgeons a 3D, stereoscopic live feed while they operate.

“This is some of the most challenging surgery there is because the tissues that they’re operating on are very delicate, and particularly valuable to their owners,” said Izatt.

Duke is one of five finalists for NVIDIA’s 2016 Global Impact Award. This $150,000 grant is awarded each year to researchers using NVIDIA technology for groundbreaking work that addresses social, humanitarian and environmental problems.

Comparison of conventional rendering (left) and enhanced ray casting with denoising (right) of the anterior segment.
Comparison of conventional rendering (left) and enhanced ray casting with denoising (right) of the anterior segment.

Two Steps Beyond Standard Practice

Standard practice for optical microsurgery is to send the patient for a pre-operation scan. This generates images that the surgeon uses to map out the disease and plan surgery. Post-operation, the patient’s eye is scanned again to make sure the operation was a success.

State-of-the-art microscopes go one step further. They use optical coherence tomography (OCT), an advanced imaging technique that produces 3D images in five to six seconds. Izatt’s work goes another step beyond that by taking complete 3D volumetric images, updated every tenth of a second and rendered from two different angles, resulting in a real-time stereoscopic display into both microscope eyepieces.

“I’ve always been very interested in seeing how technology can be applied to improving people’s lives,” said Izatt, who has been working on OCT for over 20 years.

His team is using our GeForce GTX TITAN Black GPU, CUDA programming libraries and 3D Vision technology to power their solution. Rather than having to do pre- and post-operation images to gauge their success, surgeons can have immediate feedback as they operate.

3D Images at Micrometer Resolution

Resolution of abnormal iris adhesion in full thickness corneal transplant. The top row shows the abnormal iris adhesion (red arrow) in the normal en-face surgical view seen through the operating microscope (left), volumetric OCT (middle), and cross sectional scan (left). The bottom row shows the result of the surgeon injecting a viscoelastic material to resolve the abnormal adhesion (green arrow).
Resolution of abnormal iris adhesion in full thickness corneal transplant. The top row shows the abnormal iris adhesion (red arrow) in the normal en-face surgical view seen through the operating microscope (left), volumetric OCT (middle), and cross sectional scan (left). The bottom row shows the result of the surgeon injecting a viscoelastic material to resolve the abnormal adhesion (green arrow).

A single TITAN GPU takes the stream of raw OCT data, processes it, and renders 3D volumetric images. These images, at a resolution of a few micrometers, are projected into the microscope eyepieces. CUDA’s cuFFT library and special function units provide the computational performance needed to process, de-noise, and render images in real time. With NVIDIA 3D Vision-ready monitors and 3D glasses, the live stereoscopic data can be viewed by both the surgeon using the microscope and a group observing the operation as it occurs—a useful training and demonstration tool.

“The current generation of OCT imaging instruments used to get this type of data before and after surgery typically takes about five or six seconds to render a single volumetric image,” said Izatt. “We’re now getting those same images in about a tenth of a second — so it is literally a fiftyfold increase in speed.”

Thus far, Izatt’s solution has been used in more than 90 surgeries at the Duke Eye Center and the Cleveland Clinic Cole Eye Institute. Out in the medical market, companies are still competing to commercialize real-time 2D displays. Izatt estimates his team’s 3D solution will be ready for commercial use in a couple years.

“The most complex surgeries right now are done in these big centers, but some patients have to travel hundreds or thousands of miles to go to the best centers,” said Izatt. “With this sort of tool, we’re hoping that would instead be more widely available.”

The winner of the 2016 Global Impact Award will be announced at the GPU Technology Conference, April 4-7, in Silicon Valley.

More Global Impact Award 2016 nominees

GPUs Help Monitor Rising Sea Levels with Pinpoint Accuracy

How Haiti’s Earthquake Inspired New Ways to Map Structural Safety Using GPUs

How Imperial College Uses GPUs to Spot Brain Damage

Check out the work of last year’s NVIDIA Global Impact Award winner.

Similar Stories