The study, led by Karen Drukker, Ph.D., a research associate and assistant professor in the diagnostic radiology department at the University of Chicago in Illinois, retrospectively analyzed the diagnostic ultrasound scans of 50 women with suspected breast cancer. All patients had lymph nodes that appeared normal in studies, suggesting that their cancers had not metastasized.
All 50 women later underwent surgery to remove their cancers and axillary lymph nodes; tissue biopsies of the lymph nodes revealed that 20 had metastatic cancer and 30 had cancer that remained localized at the time of surgery, according to the researchers.
The Chicago research team then applied an internally developed computerized quantitative image analysis (QIA) system, which uses artificial intelligence to analyze features such as lesion and parenchymal characteristics. The program was able to predict with promising accuracy which patients had metastases and which did not, according to Drukker.
A linear discriminate analysis classifier yielded an area under the receiver operator characteristics (ROC) curve value of 0.80 (standard error, 0.06) for the task of distinguishing between metastasized breast cancer index lesions and localized breast cancers, according to the authors.
"The QIA scheme obtained promising performance in this preliminary study and shows potential to diagnose metastatic disease," the researchers concluded. "It is important to note that this study was performed using images only of patients for whom presurgery imaging studies suggested that the breast cancer had remained localized."
In the next phase of the research, the Chicago study team plans to perform an observer study, in which several radiologists will use the program to see if it enhances their ability to diagnose metastasis.
Breast ultrasound CAD performance varies in ethnic populations, September 5, 2008
Breast CAD improves classification accuracy, cuts ultrasound biopsies, February 29, 2008
Copyright © 2009 AuntMinnie.com