Comparative efficacy of 3DQuorum 6-mm tomosynthesis slab technology (SmartSlices) vs standard of care 1-mm tomosynthesis slices: results from a multireader multicase study.

Authors

Abstract

This study aimed to compare the diagnostic accuracy and interpretation time of commercially available slabbing technology (6-mm slabs) generated from reconstructed tomosynthesis slices (3DQ) vs the standard of care (SOC) imaging mode (1 mm thickness). Tomosynthesis images (N = 391) were retrospectively collected from female subjects undergoing mammography examinations as part of an institutional review board-approved study conducted in the United States. The subjects were selected in a stratified and randomised manner for imaging dates between January 2016 and April 2017. The study population included 160 negative cases, 27 recalls, 98 biopsy-proven benign cases, and 106 biopsy-proven cancers. Fifteen radiologists assessed the clinical performance of each case in a fully crossed (all readers evaluate all cases) multireader, multicase study. The hypothesis was that the area under the receiver operating characteristic curve (AUC) for 3DQ is noninferior to SOC. Sensitivity and reading time were also compared. The difference in the AUC (3DQ vs SOC) was +0.020 (95% CI: 0.002-0.037; p value = .027), confirming noninferiority in diagnostic accuracy of 3DQ vs SOC. Mixed regression demonstrated a +0.047 recall rate difference between 3DQ and SOC (95% CI: 0.005-0.088; p value = .027) in calcification-only cancers (N = 43). Mixed regression revealed a 3DQ mean read time of 54.5 seconds, reducing the average read time by 13% (SOC mean read time = 61.9 seconds). The secondary end points confirmed noninferior recall rates and significantly reduced read time. No evidence of a differential effect of 3DQ vs SOC performance by breast density was observed. Artificial intelligence-enhanced 3DQ technology reduced interpretation time while maintaining reader-level diagnostic performance relative to SOC.

Tags

Topics