News | July 28, 2008

Automated Computer Analysis May Diagnose Breast Cancer

July 29, 2008 - Ductal carcinoma in situ (DCIS), the development of cancer cells within the milk ducts of breast tissue, is thought to be a possible precursor of invasive cancer, prompting research to understand its underlying biology-and detect it early.

Now medical physics graduate student Neha Bhooshan of the University of Chicago, her advisor, professor Maryellen Giger, and their colleagues have developed an automated computer image analysis technique to ultimately characterize and diagnose DCIS and other breast carcinomas.

The method is similar to the computer-aided detection techniques currently used to identify suspicious features on mammograms for further study by radiologists. It makes use of differences in the morphology of DCIS and other malignant and benign breast lesions, and in their response to the contrast agents used in magnetic resonance imaging (MRI) scans. For example, malignant and benign breast lesions vary in the rates at which they take in and eliminate MRI contrast agents; malignant lesions rapidly take in and wash out the contrast because they have a greater proliferation of blood vessels, while benign lesions have a slow and persistent uptake. The computer program compares seven such features in breast MRI scans taken before and after the administration of contrast, and calculates a numerical value that characterizes the tumor subtype.

To test the program's validity, the researchers used it to analyze MRI scans of 131 benign and 203 malignant breast lesions, including 79 lesions that had been pathologically diagnosed as DCIS and 124 as invasive ductal carcinoma (IDC). The system was able to differentiate benign and malignant lesions, and to distinguish DCIS and IDC lesions. Bhooshan believes computer-aided diagnosis can be applied to the image analysis of other types of cancer and may become more common in the clinical setting.

For more information: www.aapm.org

Source: American Association of Physicists in Medicine

Overlay Init