Mammography database of African American women addresses breast cancer disparities

The rate of decline in breast cancer rates over the past decade has been smaller among African American women than the general population, despite improved screening techniques.

African American women are more likely to develop cancer before age 35 and more likely to die from it at any age. Some studies have suggested there may be a difference in the histology of tumors, leading to a tendency for the cancer to be more aggressive in this population.

Howard University has taken steps to establish the first major database of African American breast cancer cases (J Digit Imaging 2008;21(1):18-26).

"We want to develop a database that provides a large variety of mammograms to provide several examples of breast cancer and benign cases from African American women," said Shani E. Ross, a Ph.D. candidate now in the department of biomedical engineering at the University of Michigan.

Mohamed Chouikha, Ph.D., chair of electrical and computer engineering at Howard, is the principal investigator.

The hope is that this database will be used as an information source and training tool to help in early detection and treatment for this demographic group, Ross said. It could help close the gap in breast cancer mortality.

So far, the project has digitized more than 5000 mammograms of 260 black patients. Additional entries are planned.

The database is to be used primarily by radiologists, radiology students, and other medical or technical personnel involved in breast cancer diagnosis, mammography, and breast cancer imaging research.

"Aspects of the database will also be open to the general public for viewing sample images and obtaining information on breast cancer, mammograms, and other screening procedures and on general breast health for African American women," Ross said.

The Howard database comes with its own viewer.

The advantage of the Howard image viewing software that accompanies the database is that it can open and manipulate images up to 60 MB in near-real-time, Ross said. Annotations can be saved as necessary. The viewer is incorporated into the database so that users can interactively query aspects of the image.

"For example, users might draw a circle around a region of interest and ask the database to retrieve all cases that have similar pathologies," Ross said.