Background: Artificial intelligence (AI) systems performing at radiologist-like levels in the evaluation of digital mammography (DM) would improve breast cancer screening accuracy and efficiency. We aimed to compare the stand-alone performance of an AI system to that of radiologists in detecting breast cancer in DM. Methods: Nine multi-reader, multi-case study datasets previously used for different research purposes in seven countries were collected. Each dataset consisted of DM exams acquired with systems from four different vendors, multiple radiologists’ assessments per exam, and ground truth verified by histopathological analysis or follow-up, yielding a total of 2652 exams (653 malignant) and interpretations by 101 radiologists (28 296 independent interpretations). An AI system analyzed these exams yielding a level of suspicion of cancer present between 1 and 10. The detection performance between the radiologists and the AI system was compared using a noninferiority null hypothesis at a margin of 0.05. Results: The performance of the AI system was statistically noninferior to that of the average of the 101 radiologists. The AI system had a 0.840 (95% confidence interval [CI] = 0.820 to 0.860) area under the ROC curve and the average of the radiologists was 0.814 (95% CI = 0.787 to 0.841) (difference 95% CI = −0.003 to 0.055). The AI system had an AUC higher than 61.4% of the radiologists. Conclusions: The evaluated AI system achieved a cancer detection accuracy comparable to an average breast radiologist in this retrospective setting. Although promising, the performance and impact of such a system in a screening setting needs further investigation.

Purpose To study the feasibility of automatically identifying normal digital mammography (DM) exams with artificial intelligence (AI) to reduce the breast cancer screening reading workload. Methods and materials A total of 2652 DM exams (653 cancer) and interpretations by 101 radiologists were gathered from nine previously performed multi-reader multi-case receiver operating characteristic (MRMC ROC) studies. An AI system was used to obtain a score between 1 and 10 for each exam, representing the likelihood of cancer present. Using all AI scores between 1 and 9 as possible thresholds, the exams were divided into groups of low- and high likelihood of cancer present. It was assumed that, under the pre-selection scenario, only the high-likelihood group would be read by radiologists, while all low-likelihood exams would be reported as normal. The area under the reader-averaged ROC curve (AUC) was calculated for the original evaluations and for the pre-selection scenarios and compared using a non-inferiority hypothesis. Results Setting the low/high-likelihood threshold at an AI score of 5 (high likelihood > 5) results in a trade-off of approximately halving (− 47%) the workload to be read by radiologists while excluding 7% of true-positive exams. Using an AI score of 2 as threshold yields a workload reduction of 17% while only excluding 1% of true-positive exams. Pre-selection did not change the average AUC of radiologists (inferior 95% CI > − 0.05) for any threshold except at the extreme AI score of 9. Conclusion It is possible to automatically pre-select exams using AI to significantly reduce the breast cancer screening reading workload. Key Points • There is potential to use artificial intelligence to automatically reduce the breast cancer screening reading workload by excluding exams with a low likelihood of cancer. • The exclusion of exams with the lowest likelihood of cancer in screening might not change radiologists’ breast cancer detection performance. • When excluding exams with the lowest likelihood of cancer, the decrease in true-positive recalls would be balanced by a simultaneous reduction in false-positive recalls.

The radiation dose associated with contrast-enhanced mammography (CEM) has been investigated only by single-center studies. In this retrospective study, we aimed to compare the radiation dose between two centers performing CEM within two prospective studies, using the same type of equipment. The CEM mean glandular dose (MGD) was computed for low energy (LE) and high energy (HE) images and their sum was calculated for each view. MGD and related parameters (entrance dose, breast thickness, compression, and density) were compared between the two centers using the Mann–Whitney test. Finally, per-patient MGD was calculated by pooling the two datasets and determining the contribution of LE and HE images. A total of 348 CEM examinations were analyzed (228 from Center 1 and 120 from Center 2). The median total MGD per view was 2.33 mGy (interquartile range 2.19–2.51 mGy) at Center 1 and 2.46 mGy (interquartile range 2.32–2.70 mGy) at Center 2, with a 0.15 mGy median difference (p < 0.001) equal to 6.2%. LE-images contributed between 64% and 77% to the total patient dose in CEM, with the remaining 23–36% being associated with HE images. The mean radiation dose for a two-view bilateral CEM exam was 4.90 mGy, about 30% higher than for digital mammography.

Early manifestation of breast cancer is often very subtle and is displayed in a complex and variable pattern of normal anatomy that may obscure the disease. The use of dual-energy techniques, that can remove the structural noise, and contrast media, that enhance the region surrounding the tumour, could help us to improve the detectability of the lesions. The aim of this work is to investigate the use of an iodine-based contrast medium in mammography with two different double exposure techniques: K-edge subtraction mammography and temporal subtraction mammography. Both techniques have been investigated by using an ideal source, like monochromatic beams produced at a synchrotron radiation facility and a clinical digital mammography system. A dedicated three-component phantom containing cavities filled with different iodine concentrations has been developed and used for measurements. For each technique, information about the minimum iodine concentration, which provides a significant enhancement of the detectability of the pathology by minimizing the risk due to high dose and high concentration of contrast medium, has been obtained. In particular, for cavities of 5 and 8 mm in diameter filled with iodine solutions, the minimum concentration needed to obtain a contrast-to-noise ratio of 5 with a mean glandular dose of 2 mGy has been calculated. The minimum concentrations estimated with monochromatic beams and K-edge subtraction mammography are 0.9 mg ml(-1) and 1.34 mg ml(-1) for the biggest and smallest details, respectively, while for temporal subtraction mammography they are 0.84 mg ml(-1) and 1.31 mg ml(-1). With the conventional clinical system the minimum concentrations for the K-edge subtraction mammography are 4.13 mg ml(-1) (8 mm diameter) and 5.75 mg ml(-1) (5 mm diameter), while for the temporal subtraction mammography they are 1.01 mg ml(-1) (8 mm diameter) and 1.57 mg ml(-1) (5 mm diameter).