Second Opinions Help Radiologists

Computer-aided diagnosis improves cancer detection rates

Not all lumps, nodules and masses are cancer. To determine whether a lump is malignant or benign, radiologists study its texture, border and shape when they review the CT scan. Researchers at the U-M Comprehensive Cancer Center are developing computer-aided diagnosis methods to make that assessment easier.

Berkman Sahiner, Ph.D., associate professor of radiology, Lubomir Hadjiyski and Heang-Ping Chan, Ph.D., professor of radiology Photo: Martin Vloet

“Our system is designed to help the radiologist,” says Lubomir Hadjiyski, Ph.D., a research assistant professor of radiology in the Medical School. “From our experiences in evaluating computer-aided diagnosis for breast cancer, we know that radiologists with computers are able to detect more cancers than radiologists by themselves. We expect that computer-aided diagnosis for lung cancer will produce similar results.”

In Hadjiyski’s diagnostic system, current and previous scans are fed through a computer program designed by U-M researchers to evaluate the size, texture, density and change of nodules over time. The computer analyzes the images with computer-vision techniques specifically designed for a given type of cancer or disease. Based on all the information, the computer determines how likely it is that the nodule is cancerous.

At the same time, the radiologist examines the images and evaluates the likelihood of cancer. The radiologist then compares the two results and makes a final decision.

“The radiologist is not perfect and the computer is not perfect, but working together they detect more cancers,” Hadjiyski says.

Hadjiyski’s computer-aided diagnosis programs for lung and breast cancer need Food and Drug Administration approval before they can be offered clinically.

—NF

Read an expanded version of the story at: www.med.umich.edu/opm/newspage/2004/computer.htm

For patient information about lung cancer:
www.cancer.med.umich.edu/learn/lung.htm