Can New Ultrasound Methods Detect Breast Cancer?
Nov. 29, 1999 (Chicago) -- Researchers showed an array of cutting-edge ultrasound techniques that may help doctors determine whether a suspicious breast lesion is cancerous or benign.
"This is really a work in progress," said Ellen Mendelson, MD, chief of women's imaging and director of the Breast Center at Western Pennsylvania Hospital in Pittsburgh. "What I saw here today was an excellent manifestation of increasing interest in ultrasound and what it might and might not be able to do," said Mendelson, the moderator of the session at the Radiological Society of North America meeting that scrutinized various studies. However, Mendelson said that none of the studies reviewed, by themselves, provided conclusive evidence that the techniques could be applied to clinical practice without a lot more investigation.
Among the studies: Joseph Lo, PhD, an assistant professor of radiology at Duke University Medical Center in Durham, N.C., demonstrated that a computer program called "The Artificial Neural Network" (ANN), which could determine within a few moments whether a ultrasonic image of a breast lesion was malignant or benign -- sometimes with better results than those achieved by human radiologists.
For example, one radiologist who was looking at an ultrasound image rated the lesion as highly likely to be malignant. The ANN disagreed, reporting a less than 20% chance that the tumor was malignant. The verdict after biopsy: benign.
Lo fed reports of 192 ultrasound images into the ANN and reported that the computer program was able to identify all 71 malignancies. However, it was able to exclude from biopsy 48 of 121 benign tumors. Lo said the program could be adjusted to lower sensitivity and increase the number of biopsies avoided, but that was at the risk of missing some cancers. He will continue adding cases to the ANN, which he said has the capability of learning as it gathers information.
In another study, Hye-Young Choi, MD, of the department of diagnostic radiology, Ewha Women's University, Seoul, South Korea, showed that Doppler images, a type of ultrasound, could dramatically reveal features of tumors, especially branching aspects of malignant lesions. She examined lesions from 129 consecutive cases of suspicious lesions, 75 of which were benign and 54 malignant. Her analysis indicated that lesions on the periphery of the breast were more likely to be benign, while branching tumors highlighted by the color Doppler were more likely to be malignant.
Although the pictures were impressive, Mendelson said she expected that the most important elements in determining status of the lesions would be "margins and shape." Whether color Doppler would fit into diagnostic use remains to be sorted out, she said.
Also, researchers from Germany reported use of the experimental ultrasound contrast agent Levovist to help doctors distinguish types of breast tumors. Contrast agents help highlight areas of different density in the body. Jens Uwe Blohmer, MD, from the Frauenklinik Charite in Berlin, administered Levovist by vein to 22 patients with breast cancer and to 11 women with benign breast lesions.
By looking at the intensity of the signal between healthy tissue and the lesion in question, Blohmer said doctors might be able to determine whether the lesion is malignant. In his experiments, the greater the time to peak brightness, the more likely the tumor was benign. The average time to peak brightness in benign lesions was 42 seconds, compared with 26 seconds for malignant tumors.