Helping computers help women

Someday women seeking a second opinion concerning their mammograms may turn to computers.

But first computers must learn to read shapes.

UNLV electrical engineer Lori Bruce is trying to translate the round shapes of benign lumps and the spiny forms of cancerous tumors from a typical mammogram into language a computer understands.

If she is successful, Bruce hopes her technique will help launch a medical imaging laboratory for research on women's health issues at UNLV.

"It isn't easy," Bruce says from her third-floor office in the Howard Hughes School of Engineering.

The papers on her desk are filled with equations for measuring the three basic tissue shapes: round, nodular and star-like.

Electrical engineers began to use computers to dissect aerial pictures for the U.S. Department of Defense in the mid-1980s, she said.

Bruce began working on the computer-recognition problem for her dissertation five years ago. The University of Alabama at Huntsville awarded Bruce her doctorate in electrical and computer engineering in 1996. She is now an assistant professor in the Department of Electrical and Computer Engineering at UNLV.

Bruce came to UNLV two years ago. She learned to spot patterns from work on aerial photographs taken by the military and the space programs that use remote viewing. Since then she has published 10 journal articles and conference papers related to her medical research.

Besides reading mammograms, Bruce also has worked on monitoring the heartbeat of fetuses.

"Computers are very good at looking at points in detail," Bruce said.

In medicine, however, spotting the difference between diseased and healthy tissues must rise above the level of looking at a cell and instead focus on the entire lump.

Using wavelet transformation, an engineering term that explains how human eyes detect light and human ears detect sound, Bruce is searching for the language that will allow computers to shift their gaze from the microscopic scale to a more global view of a lump's shape.

When Bruce set out to work on the problem, computers recognized about 62 percent of breast lumps accurately. Her work has raised the accuracy level to between 80 and 86 percent. "In the worst case, it's 76 percent," she said.

But even if her research is successful, Bruce said the computers will not replace medical experts. That is why she is working with the University Medical Center and the University of Nevada School of Medicine.

UMC supplies hundreds of mammograms for Bruce to use in her project.

Drs. Daniel Kirgan and Annabel Barber at the Department of Surgery are eager to help Bruce discover a more accurate method for reading mammograms. Barber is also an associate professor with the School of Medicine.

"They help me to read the mammograms so I can tell the truth and program the computers," Bruce said, acknowledging that she is not a medical expert.

archive