Breast cancer detection accuracy of AI in an entire screening population: a retrospective, multicentre study

Cancer Imaging

Table 1 Clinical characteristics of the study sample

	Study sample (n = 257,671)
Screening site
Aabenraa Esbjerg Odense Vejle	49,641 (19.3) 49,860 (19.4) 104,984 (40.7) 53,186 (20.6)
Age at screening, years
< 50^* 50–59 60–69 70–79 ≥ 80	59.3 (6.0) 25 (< 0.1) 133,223 (51.7) 120,315 (46.7) 4024 (1.6) 84 (< 0.1)
Breast cancer prevalence Screen-detected cancer Interval cancer	2041 (0.8) 1479 (72.5) 562 (27.5)
Breast cancer type
Invasive cancer DCIS	1830 (89.7) 211 (10.3)
Screening outcome^†
Normal Abnormal	250,810 (97.3) 6861 (2.7)
Arbitrations^‡	7434 (2.9)
Agreement between readers^‡
First and second reader First reader and arbitrator Second reader and arbitrator	250,663 (97.3) 3299 (44.4) 4537 (61.0)

Data are n (%) or mean (SD). DCIS = ductal carcinoma in situ
*These women were all 49 years old and were invited to regular biennial screening a few months too early
†Combined reading outcome of the double reading with arbitration
‡There is a small overlap of n = 426 (0.2%) studies in the arbitrations and agreements between first and second readers due to disagreements on subset outcomes with additional initiatives, such as stereotactic breast biopsy, which were eventually classified into the available binary screening outcome

ISSN: 1470-7330