Visual researchers at Brigham and Women’s Hospital (BWH), MD Anderson Cancer Center in Texas, and the University of York and Leeds in England presented mammograms to experienced radiologists for half a second and found they could detect abnormalities in breast tissue at better than chance levels.
The researchers then performed a series of experiments to test this ability to explore what signal may alert radiologists to the presence of a possible abnormality in mammograms in the hopes of using these insights to improve breast cancer screening and early detection.
The study, “A half-second glimpse often lets radiologists identify breast cancer cases even when viewing the mammogram of the opposite breast,” was published in the Proceedings of the National Academy of Sciences.
“Radiologists can have ‘hunches’ after a first look at a mammogram. We found that these hunches are based on something real in the images. It’s really striking that in the blink of an eye, an expert can pick up on something about that mammogram that indicates abnormality,” Jeremy Wolfe, PhD, senior author of the study and director of the Visual Attention Laboratory at BWH, said in a news release. “Not only that, but they can detect something abnormal in the other breast, the breast that does not contain a lesion.”
In North America, screening mammography has a false negative rate of 20 to 30 percent and a recall rate of about 10 percent. With a disease prevalence of about 0.3 percent, the vast majority of those recalled will not have cancer.
Being able to rapidly pick up the substance of an image is an important factor in routine visual perception that allows us to allocate our time and attention intelligently when confronted with new visual information — for example, can I find food here? Is there danger here?
Wolfe and his colleagues have shown that experienced radiologists can distinguish normal from abnormal mammograms at above-chance levels in as little as half a second, whereas nonexpert radiologists cannot.
Here, researchers conducted a series of experiments to assess if image size, breast tissue symmetry, density, resolution, and other characteristics were contributing factors for radiologists’ exceptional detection. They discovered that their rapid-image detection success was based not so much on breast density or symmetry, but rather on the fine details of breast tissue texture.
Researchers also found that experienced radiologists could do better than chance in discerning normal tissue from breast cancer even when the images of abnormal breast tissue did not capture a malignant lesion or when those images were captured from the breast on the other side of the body.
“These results suggest that there may be something in the nominally normal breast that looks abnormal and is detectable,” Wolfe said. “Together, these results suggest that radiologists may be picking up on some sort of early, global signal of abnormality that is unknown to us at this point.”
Defining the signal that radiologists are detecting could help investigators refine and improve computer-aided detection (CAD) methods that can help medical screening and could be included in medical training to improve clinicians’ detection skills.
The researchers are working on whether other medical image specialists such as pathologists and dermatologists can proficiently use similar signals.