Digital mammography screening with photon-counting technique: can a high diagnostic performance be realized at low mean glandular dose?

To assess screening performance of a direct radiography (DR) photon-counting system versus statewide screening units with different digital technologies. The local ethics board approved retrospective study of prospectively acquired data from the North Rhine-Westphalian mammography screening program (2009-2010). Informed consent was waived. Examinations in 13 312 women with a DR photon-counting system and statewide digital screening examinations in 993 822 women were included (37 computed radiography mammography systems and 55 DR systems). Diagnostic performance was assessed with cancer detection rate, recall rate, and proportion of small invasive cancers and ductal carcinoma in situ (DCIS). Mean glandular dose was calculated for DR photon counting and for a conventional DR subgroup. Differences were tested with χ(2) and t tests (P < .05). The cancer detection rate for subsequent screenings was higher for DR photon counting than statewide rates (0.76% [67 of 8842] vs 0.59% [3108 of 527 194], P = .05) at a higher recall rate (5.4% [475 of 8842] vs 3.3% [17 656 of 527 194], P = .001). Detection of invasive cancers up to 10 mm for DR photon counting was high for initial (40% [14 of 35]) and subsequent (42% [19 of 45]) screenings but not significantly different from statewide rates (initial, 31.6% [942 of 2979], P = .50; subsequent, 32.5% [765 of 2353], P = .25). The DCIS subsequent screening rate was higher for DR photon counting than statewide screening (0.23% [20 of 8842] vs 0.12% [616 of 527 194], P = .01) and the conventional DR subgroup (0.23% [20 of 8842] vs 0.12% [65 of 52 813], P = .025). Mean glandular dose for DR photon counting was significantly lower than that for conventional DR (0.60 mGy ± 0.20 vs 1.67 mGy ± 0.47 [craniocaudal views], 0.64 mGy ± 0.23 vs 1.79 mGy ± 0.53 [mediolateral oblique views], both P = .0001). Digital mammography screening with dose-efficient photon counting enables desirable detection rates of small invasive cancers and DCIS. Higher detection rates compared with statewide performance occurred with subsequent screening but had higher recall rates.