Nov. 4 (UPI) -- An artificial intelligence, or AI, tool improved radiologists' breast cancer detection rate by an average of 5%, according to a study published Wednesday in the journal Radiology: Artificial Intelligence.
The technology, called MammoScreen, also reduced rates of false-positives, or instances in which radiologists misidentify spots on mammograms as cancer, by an average of 6%, the researchers said.
"The results show that MammoScreen may help to improve radiologists' performance in breast cancer detection," study co-author Serena Pacilè, clinical research manager at Therapixel, where the software was developed, said in a statement.
Although mammography screening for breast cancer generally improves prognosis in women with the disease by identifying cases earlier and enabling earlier treatment, doctors miss many cancers with the technology, and suspicious findings often turn out to be benign, according to Pacilè and her colleagues.
Related
A study published by Radiology in 2013 found that, on average, only 10% of women who underwent additional diagnostic work based on suspicious mammogram findings are ultimately found to have cancer.
MammoScreen is designed to identify regions suspicious for breast cancer on two-dimensional digital mammograms, assess their likelihood of malignancy based on a complete set of four views and generate a set of image positions with a related suspicion score.
The U.S. Food and Drug Administration cleared the tool for use in March.
For the study, Pacilè and her colleagues asked 14 radiologists to assess a dataset of 240 digital mammography images depicting different types of abnormalities, cancerous and non-cancerous, collected between 2013 and 2016
Researchers assessed half of the dataset without AI, while the other half was assessed with the help of Mammoscreen during a first session and without it during a second session.
MammoScreen-assisted evaluation successfully identified women with breast cancer 69% of the time, a "slight" improvement over assessment by a radiologist alone, which was accurate 66% of the time, the researchers said.
The tool also helped reduce the rate of false negatives, or findings that look normal even though cancer is present, by about 10%, the data showed.
In addition, the improved diagnostic performance of radiologists in the detection of breast cancer was achieved without prolonging their workflow, allowing them to focus their attention on the more suspicious examinations, the researchers said.
