Mar 6 2019
A new paper in the Journal of the National Cancer Institute puts forward the notion that artificial intelligence (AI) systems may be able to perform as correctly as radiologists in the assessment of digital mammography in breast cancer screening.
Breast cancer remains the most common cancer in women, in spite of crucial improvements in therapy. It is a major cause for cancer-linked mortality, accounting for about 500,000 annual deaths globally. Breast cancer screening programs using mammography are successful in decreasing breast cancer-linked mortality. But, existing screening programs are very labor intensive because of the high number of women that need to be screened. Taking into consideration the growing scarcity of breast screening radiologists in certain countries, many scientists believe other screening approaches may be worth exploring.
Since the 1990s, computer-aided detection systems have been created to automatically detect and categorize breast lesions in mammograms. However, there are no studies thus far to prove that these systems directly enhance screening performance or cost effectiveness. This has disallowed their use as a technique for screening mammography.
In this research, scientists compare, at a case level, the cancer detection performance of a commercially available AI system to that of 101 radiologists who scored nine diverse cohorts of mammography examinations from four different manufacturers as part of studies formerly done for other purposes.
Each dataset comprised of mammography exams attained with systems from four different manufacturers, many radiologists’ assessments per exam, producing a total of 2,652 exams (653 malignant) and interpretations by 101 radiologists (28,296 independent interpretations).
The AI system’s performance was statistically not-mediocre to that of the average of the 101 radiologists. The assessment system realized a cancer detection accuracy equivalent to an average breast radiologist in this surveying setting.
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. 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.
Ioannis Sechopoulos, Study’s Co-Author, Department of Radiology and Nuclear Medicine, Radboud University Medical Centre