AI Detects Breast Cancer Better Than Average Radiologist

Cancer-detecting system commercially available

Computer systems have been developed since the 1980s to help detect and classify breast lesions in mammograms automatically. However, no single system has been proven to have good enough performance or accuracy for clinical use.

The latest study thought shows that this may be changing. The study compared at a case level the cancer detecting performance of a commercially available AI system against 101 radiologists who evaluated digital mammograms (DM) examinations for reasons related to other research.

AI as good as average radiologist

Over 28,000 interpretations were included in the study. The AI was trained using 9,000 mammograms with cancer and another 180,000 mammograms without any abnormalities. The comparison showed that the AI system was statistically similar to that of the trained radiologists.

The system essentially could detect cancer with an accuracy comparable to an average breast radiologist in this retrospective 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," said Ioannis Sechopoulos, one of the paper's authors.

Research continues

"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." 

[see-also]

Breast cancer is the most common form of cancer in women and accounted for more than 500,000 annual deaths worldwide.

Implementation of the AI system could mean more women can get screened per year, particularly in areas where there are difficulties in accessing trained radiologist. 

“Considering the increasing scarcity of radiologists in some countries, including breast screening radiologists, alternative strategies to allow continuation of current screening programs are required,” wrote Alejandro Rodríguez-Ruiz, MSc, Radboud University Medical Center in the Netherlands, and colleagues.

The team noted that “the performance and fashion of implementation of such an AI system in a screening setting” still need to be researched in more detail.