A paper suggests that an artificial intelligence (AI) system may be able to evaluate a digital mammography in breast cancer screenings as accurately as a radiologist.
Today, breast cancer is the most common cancer in women, and although there has been extensive improvements in therapy, it’s still a major cause of death, accounting for approximately 500,000 annual deaths worldwide.
Breast cancer screening programs that use mammograms are often effective in reducing breast cancer-related mortality. On the contrary, current screening programs require an intensive amount of labor because of the amount of women who need to be screened. With this, the amount of breast screening radiologists is scarce in some countries, so researchers believe looking into other methods of screening may be worth investigating.
Although computer-aided detection systems that detect breast lesions in mammograms date back to the 1990s, there are no current studies that suggest these systems improve screening performance or cost effectiveness.
In the current study, researchers compared the cancer detection performance of commercially available AI systems to 101 radiologists who scored nine different cohorts of mammography examinations from four different manufacturers.
Each dataset contained mammography exams that were acquired with AI systems from four different vendors and multiple radiologists’ assessments per exam. In total 2,652 exams were done and interpreted by 101 radiologists with 653 malignant results.
The performance of the AI system was statistically not inferior to the 101 radiologists. The system did achieve a cancer detection accuracy comparable to an average breast radiologist.
“Before we could decide what is the best way for AI systems to be introduced in the realm of breast cancer screening with mammography, we wanted to know how good can these systems really be,” says Ioannis Sechopoulos, one of the paper’s authors. “It was exciting to see that these systems have reached the level of matching the performance of not just radiologists, but of radiologists who spend at least a substantial portion of their time reading screening mammograms.”