Artificial intelligence (AI) in a breast screening program: a retrospective study

The AI case malignancy score (AI-CMS) represents the AI algorithm’s confidence (from 0% to 100%) that a mammography exam is malignant. This work aims to evaluate retrospectively a strategy that integrates AI-CMS into a standard screening scenario to reduce the radiologists' workload.
43,344 consecutive year 2023 screening exams from the Reggio Emilia Breast Screening Program were considered, which included 202 biopsy-proven cancers. In the proposed strategy, Computer Aided Detection (CAD) acts as reader (CR), recalling women with an AI-CMS greater than a predefined threshold (ranging from 5% to 25%). If the first radiologist (HR1) disagrees with CR, the case goes to a second radiologist (HR2) and, in case of human disagreement, to a third radiologist (HR3). This strategy's recall rate, workload reduction, and economic impact were analyzed.
Assuming AI-CMS thresholds of 5%, 8%, 10%, 15%, 20%, and 25%, reader CR’s decision to recall (RD) rates vary significantly. These strategies significantly decrease the total human readings, reducing the human workload between 15.0% and 36.7%. The final recall rate (RR) decreases between 3.95% and 3.75% compared to the actual RR of 4.13%. Up to the threshold of 8%, no true positive cases were missed. Considering CAD payback periods of 6 or 8 years, this strategy leads to savings ranging between about 17,800 and over 590,000 euros.
Integrating AI-CMS support into a standard screening scenario could substantially reduce the screen-reading workload without affecting false negatives and false positives. This approach has also proven to be economically sustainable.