Artificial intelligence (AI) could potentially play a crucial role in the accurate detection of breast cancer, matching the diagnostic accuracy of human physicians. Recent studies have shown that AI exhibits similar performance to healthcare experts in the interpretation of mammogram screenings.
The study, conducted in partnership with the National Health Service’s Breast Cancer Screening Program (NHSBSP), demonstrated that the AI algorithm exhibited a slight advantage over human healthcare experts when it came to the detection of breast cancer in 120 mammogram screenings. Researchers hold a positive outlook that ongoing investigations will pave the way for the integration of AI as a routine tool in breast cancer screening to complement physicians’ diagnostic efforts.
Usually two readers interpret mammograms to reduce false positives, but a shortage of radiologists makes this difficult. A British study in the RSNA Radiology journal compared an AI algorithm to NHS human evaluators in mammogram interpretation.
The breast cancer screening process includes taking X-rays of both breasts by a mammographer to find small cancer signs. In the NHS Breast Screening Program, women are invited for their first mammogram between ages 50-53, and they get screened every three years until they are 70.
However, this approach is not foolproof in detecting all breast cancer cases, potentially resulting in false-positive results that lead to unnecessary additional tests like imaging or biopsies. A study conducted at the University of Nottingham suggests that reviewing mammogram results twice can increase cancer detection rates by 6% to 15% while keeping recall rates low. Nevertheless implementing this strategy is challenging due to a global shortage of qualified readers.
In the study led by Professor Yan Chen, the researchers used the Personal Performance in Mammographic Screening (PERFORMS) quality assurance assessment, a tool used by the NHSBSP, to conduct tests. Each PERFORMS test included 60 mammogram exams from the NHSBSP, showing different results. The study compared the scores of NHS readers with those of an AI system, using data from two sets of PERFORMS tests consisting of 120 mammograms in total.