An artificial intelligence (AI) system can detect breast cancer using data from mammography examinations, according to research published in Radiology.
Researchers evaluated the accuracy of a commercially available AI system for breast cancer diagnosis, compared with independent double readings by radiologists.
The study included data from 47,877 patients and 122,969 screening examinations.
Overall, there were 752 screen-detected cancers (a rate of 6.1 per 1000 examinations) and 205 cancers diagnosed in the interval between scheduled screenings (a rate of 1.7 per 1000 examinations).
The AI system ranked examinations between 1 and 10, where 1 represented the lowest risk of breast cancer and 10 represented the highest risk.
The researchers used the AI system as a binary decision tool with 3 thresholds for selecting examinations as “suspicious” or “not suspicious.”
“Threshold 1 was set at an AI score of 10, threshold 2 was set to yield a selection rate similar to the consensus rate (8.8%), and threshold 3 was set to yield a selection rate similar to an average individual radiologist (5.8%),” the researchers explained.
At threshold 1, the AI system selected 86.8% of the screen-detected cancers and 44.9% of interval cancers. At threshold 2, the system selected 85.1% and 41.5%, respectively. At threshold 3, the system selected 80.1% and 30.7%, respectively.
“Screen-detected cancers with AI scores not selected using the predefined thresholds had favorable histopathologic characteristics; opposite results were observed for interval cancer,” the researchers noted.
They concluded that the proportion of screen-detected cancers not selected by the AI system at the 3 thresholds was less than 20%, so the tool is “promising” for cancer detection.
Disclosures: Some study authors declared affiliations with biotech, pharmaceutical, and/or device companies. Please see the original reference for a full list of authors’ disclosures.
Larsen M, Aglen CF, Lee CI, et al. Artificial intelligence evaluation of 122969 mammography examinations from a population-based screening program. Radiology. Published online March 29, 2022. doi:10.1148/radiol.212381