Speeding up detection

ASEE Prism, Jan 2003 by Rushmore, Jane

ON CAMPUS

By taking the digital-imaging processes used to inspect airplanes and applying it to the human body, two Rowan

University engineering students have come up with a computer program that could speed up the detection of breast cancer. The software analyzes digital mammograms and automatically determines the percentage of breast density- an important component in diagnosing potential disease.

Shreekanth Mandayam, an associate professor in Rowan's department of electrical and computer engineering, knew that researchers at Fox Chase Cancer Center needed a quick way to process mammograms as part of a study on the radiodensity of breast tissue. Mandayam selected electrical and computer-engineering students Lyndsay Burd and Richard Eckert to develop a computer program that analyzes mammograms.

The students work with Fox Chase epidemiologist Marilyn Tseng, who studies the factors that effect breast density. Women with 60 to 75 percent radiodense breast tissue have six times the risk of developing breast cancer. Heredity, in comparison, only doubles the risk. Normally, a trained radiologist would carefully look over each mammogram to determine the amount of radiodense tissue. But by using the new program, the students can speed up the analysis and eliminate sources of human error.

To do the analysis, Burd first scans the mammograms. Then, using the binary template masks Eckert helped develop over the past year, the computer distinguishes between the outside film region and the inside breast tissue region of the mammogram. Finally, the program applies an algorithm to the breast tissue region and the computer instantly calculates the percentage of radiodense breast tissue. The whole process takes only a few seconds.

Burd and Eckert are currently working with Mandayam to test the accuracy of the software. They take a material called BR-12 that has optical characteristics similar to typical breast tissue and obtain a mammogram of it called an "artificial breast phantom." Based on the thickness of the BR-12 they predict the radiodensity of the phantom and then compare it to the software's results.

It will take a few upgrades for the technology to supersede human analysis. Right now the program can only tell if there is radiodense tissue in the breast-not precisely where it is. So for their next challenge, Burd and Eckert will try to find a way for the computer to pinpoint the location of the suspect tissue.