Breast cancer research seeks to tie imaging with predictions

University of Chicago professor of radiology and medical physics Maryellen Giger was one of the plenary speakers at the SPIE optical science annual conference Aug. 28 through Sept. 1 at the San Diego Convention Center, and her Aug. 30 talk "CAD and Radiomics in Breast Cancer Imaging" addressed the search for relationships between mammographic imaging results and predictive risks.

Radiomics is the high-throughput conversion of images to measurable data.

"We want to use all the information to develop new predictive methods," said Giger. "We want to ask questions about the relationships between features seen in the medical images. We want to find for each patient the right treatment at the right time."

Computer-aided design includes computer-aided detection and computer-aided diagnosis. Computer-aided detection identifies areas of interest.

"In computer-aided detection we're looking to aid the radiologist," said Giger.

Computer-aided diagnosis involves the image being fed to a computer with the results shown to the radiologist. Giger was on a University of Chicago team that developed a prototype system in 1994 in which the radiologist would first look at the image without the computer input and then re-interpret the image once the computer data was added.

A better determination of whether a tumor is benign or malignant has allowed the radiologist to define a management plan for the patient. The University of Chicago researchers are working on a computer program to determine the chance of malignancy and to evaluate the success of the radiologists' determinations. Image-based predictions can utilize factors such as lesion size, texture, and the shape of the tumor.

"We want to ask questions about the relationships between features seen in the medical images and the history of cancer," said Giger.

The density of a breast is suspected to be a factor in the risk of breast cancer, as in many cancer patients the parenchymal pattern is often coarse and low in contrast. Genetic research indicates that the UGT2b gene regulates mammographic density.

"The UGT2b gene variation may contribute to variation in mammographic parenchymal patterns and breast density," said Giger. "We're trying to relate the imaging features to the genomic features."

The work includes utilizing four-dimensional images, or three-dimensional images over time, to analyze tumors.

"What we're trying to do is harness the data with imaging radiomics," said Giger. "We can come up with these breast patterns, but it would be nice to try to relate them."