Human vision the model for video image analyst
From tanks to blood cells, applications of Acton’s work limitless

Photo by Andrew Shurtleff

A rewarding pursuit is how Scott Acton characterizes the research done in his Virginia Image and Video analysis lab.

By Dan Heuchert

It was collaboration at first sight. Biomedical engineering professor Klaus Ley was attending Scott Acton’s introductory seminar three years ago. He watched with interest as the new associate professor of electrical and computer engineering showed off some of the video image-analysis technology he had developed at Oklahoma State University. On the screen, a computer tracked and outlined Tour de France cyclists filmed from above — even as they passed under trees and other obstacles.

“If he can track cyclists from a helicopter,” Ley recalls thinking, “he can track white blood cells.” He introduced himself.

Two years later, in July 2002 — with funding from the National Institutes of Health and the Whitaker Foundation — Acton and Ley delivered what is believed to be the first system to automatically track the movement of white blood cells in footage taken from inside living, breathing blood vessels. The cells, also known as leukocytes, typically help remove damaged tissue and fight infections, but can sometimes turn on the body and attack healthy tissue, leading to arthritis or heart disease.

The tracking system “has big implications for drug testing, studying inflammatory disease and any study of cell motility,” Acton said.
Though the system is still undergoing refinements, Ley and Acton say they field requests almost weekly from companies interested in developing the technology.

The leukocyte-tracking system is typical of the kind of work done by Acton, 37, who grew up in Northern Virginia and earned an undergraduate degree in electrical engineering from Virginia Tech. He was in graduate school at the University of Texas, studying computer design, when he took a class on “computer vision,” which used human vision as a model for doing computer image analysis. “I became hooked,” he recalled.

After working in private industry for AT&T, the MITRE Corporation and Motorola, he joined the faculty at Oklahoma State before U.Va. lured Acton and his lab — including three doctoral students, all of whom have now received their degrees — back to his home state in 2000. Renamed “Virginia Image and Video Analysis,” or VIVA, his lab currently employs two professionals and eight students.

Their work has found an academic niche, he said. “There are a lot of people building [imaging] machines, and some doing image processing. We sort of have a nice niche in signal analysis.”

The commercial potential of VIVA’s work seems limitless. Acton is currently working with cardiologists on a new project to identify areas where blockages have caused damage to heart function, called “wall motion defects.” By tracking the heart’s motion in MRI video and comparing it to a database of normal heart function, it will recognize and “light up” damaged areas — a skill that can take years for cardiologists to develop.

“This technology is going to be useful technology,” said Frederick H. Epstein, an associate professor of radiology, who is leading the project.
There are other, non-medical applications of VIVA’s work. Acton receives some funding from the military to work on automatically tracking targets, and even specific parts of targets — the hood of a military vehicle as it moves through an urban landscape, for instance — via video analysis.

Acton works with all kinds of footage — regular video, ultrasound and MRIs. Tracking tanks and blood cells uses “a lot of the same principles,” he notes.

Image analysis actually breaks down into two parts. “Identifying the cell to be tracked is a different problem than tracking it,” he said. Speaking of the leukocyte program, he said, “The first system made humans identify the cells, and then tracked them. Now we’re trying to make the computer identify them.”

Tracking the identified objects can be tricky, as well. Cells can change shape, or be obscured by other cells; military targets can move, or the camera can shift perspective, altering an object’s two-dimensional profile. Some applications seek to observe how the tracked object interacts with its environs, while others seek to isolate it and remove the clutter around it.

It’s a rewarding pursuit, Acton said. “The most exciting part is probably the biomedical work.”

Another medical project VIVA is working on enhances ultrasound images and more clearly defines the boundaries of various objects. One potential use: allowing surgeons to “see” the boundaries of a cancerous prostate gland as they implant radioactive seeds to destroy a tumor.

Currently, imaging is done before the implantations, and surgeons hope the target doesn’t move in the meantime. A radioactive seed mistakenly left in the nearby rectum can cause serious complications, Acton said.
Down the road, Acton sees another possible medical application for his work. Each year, doctors worldwide order 4 billion complete blood counts (“CBCs” for you “ER” fans), a lab test that requires drawing blood and analyzing it under a microscope.

Acton foresees a time when a machine will trans-illuminate a blood vessel close to the surface of the skin — perhaps at the bottom of the tongue, or under a fingernail — and examine it with a microscope hooked up to a video processor. His technology would then examine the video and recognize and “count” the various elements of the bloodstream as they flowed by — taking a CBC without drawing blood.

Acton will likely remain in great demand as a collaborator, said Tom Skalak, chairman of the biomedical engineering department.

Acton’s work with Ley “is a great example of a real merging of the minds, to address a real problem in medicine that had not been addressed before,” he said.