UF Enhanced Mammograms For Earlier Detection To Undergo Trial

August 14, 1998

GAINESVILLE — A major University of Florida research effort to improve breast exam X-ray images using computers will undergo its first trial this fall, when radiologists determine whether the more detailed mammograms better show early signs of breast cancer.

The trial for the digitally enhanced mammograms comes shortly before new digital mammography systems reach the market. The machines are expected to gradually replace traditional X-ray methods.

“The advent of the digital machines is one more reason to process the images using computers, because when you have everything in digital form already it’s not difficult to do some enhancing,” said Iztok Koren, a postdoctoral associate in electrical and computer engineering and one of several engineers and medical professionals who have worked on the project.

Early detection is vital to successful treatment of breast cancer, the second major cause of cancer death in American women, according to the American Cancer Society (ACS). Some 178,700 cases have been diagnosed this year, with about 43,900 deaths.

Signs of breast cancer usually show up in a mammogram before women or health-care practitioners identify suspicious lumps by touch. But experts say as many as 30 percent of mammograms found negative for potentially cancerous lesions are actually positive.

The problem rests partly with human ability. X-ray images of breasts often appear cloudy and ill-defined, and radiologists have varying degrees of success identifying early tumors or potential tumors based on experience and skill. Also, the eye can distinguish between only 100 shades of gray, far less than the thousands recorded in X-ray detectors.

Medical professionals long have recognized enhancing mammographic images electronically could highlight suspicious abnormalities, but doing so has proved a challenge. Some past attempts, for example, resulted in the introduction of features not present in the original mammogram, leading radiologists to suspect cancer where none existed.

Launched five years ago with a $1.4 million grant from the U.S. Army Breast Cancer Research Program, the UF project uses mathematical tools to separate the mammographic image into distinct components, then puts them back together in a form that highlights irregular or suspicious features. The method achieves this goal through decomposing the mammogram in different orientations and scales, enhancing lesions, masses, calcifications or other features that could prove cancerous, then reconstructing the image.

The enhanced images have shown considerable promise, said Andrew Laine, a former associate professor in the computer and information science and engineering department who spearheaded the UF group before joining Columbia University’s department of biomedical engineering last year.

Laine, who continues to be involved with the project, said tests of the method on a simulated breast used to calibrate mammography machines found it successfully highlighted a mass that was not only invisible in the original mammogram but also was left undiscovered by other enhancement methods.

“Preliminary results are very strong,” he said, adding radiologists who had perused the images informally gave them “rave” reviews.

In the trial this fall, researchers will enhance images from about 450 mammogram cases in a national database, then show them to about 10 general radiologists and six radiologists who specialize in diagnosing breast cancer, Laine said. The goal is to determine whether the enhancements will enable general radiologists to perform as well as specialized radiologists and whether they can improve specialized radiologists’ performance, he said.

Others who have worked on the UF project include Fred Taylor, a UF professor of electrical and computer engineering; Janice Honeyman, an associate professor and director of informatics in Shands’ radiology department; and Dr. Barbara Steinbach, former director of mammography at Shands and now medical director of the ETMC Breast Care Center in Tyler, Texas.