ORNL's Titan supercomputer helps sharpen breast cancer detection

Anne Brock # State

OAK RIDGE, Tenn. (WVLT) - Scientists at Oak Ridge National Laboratory are working to improve breast cancer diagnosis by making mammograms more accurate.

Director Georgia Tourassi at the ORNL Health Data Sciences Institute said mammogram readings are not always accurate.

"This is an extremely difficult diagnostic task. Human error exists,"
Tourassi said.

Tourassi's team at ORNL used the Titan supercomputer to analyze data about how doctors were reading mammograms. They tried to limit the amount of human bias involved in reading those breast images in the search for cancerous tumors.

"(The team's goal is) to help physicians detect cancer, breast cancer in screening mammograms earlier and more accurately,"
Tourassi said.

Analyzing the way radiologists looked at the mammogram images involves placing special devices on their heads and monitoring the way they looked at various points on the images. It took into account that the images they had examined previously might bias their next analysis of a new image.

"So what we are trying to do is to put together the best possible computer with the best possible expert, to develop a far more effective decision support technology,"
Tourassi said.

The ORNL research is published with the National Center for Biomedical Information in the Journal of Medical Imaging.