Answers are grounded in this paper and cite their sources.
Purpose To assess the association between patient-specific breast compression-force reduction and reader-rated image quality in digital breast tomosynthesis (DBT), and explore reductions associated with non-diagnostic images. Approach In this within-subject study, 132 women underwent paired standard- and reduced-compression DBT (one breast per woman; CC and MLO views). Four radiologists independently scored randomized synthetic 2D images on a study-specific 10-point scale, blinded to compression condition. Mean scores < 8 were classified as non-diagnostic. Analyses included paired comparisons, inter-reader reliability, patient-clustered modeling, and exploratory ROC analysis with internal cross-validation. Results Mean force fell from 93.14 to 54.20 N (CC) and 116.86 to 67.82 N (MLO), representing reductions of 40.4% and 40.9% (both P < .001). Mean image-quality scores decreased from 9.55 to 9.03 (CC, P = .004) and 9.50 to 9.01 (MLO, P = .012). System-reported glandular dose increased from 1.41 to 1.46 mGy (CC) and 2.11 to 2.24 mGy (MLO). Non-diagnostic images occurred in 16/132 CC (12.1%) and 18/132 MLO (13.6%) reduced-compression acquisitions; 28/132 women (21.2%) had at least one non-diagnostic view. Mean scores by reduction band (<30%, 30–<40%, 40–<50%, ≥50%) were 9.95, 9.86, 9.23, and 6.75, respectively; no acquisition below 50% reduction was non-diagnostic. Exploratory ROC cutoffs were approximately 52% in both views (apparent AUC 1.000); repeated cross-validation yielded approximately 93% sensitivity. Cutoffs were derived and evaluated within the same cohort. Conclusion Patient-specific force reduction preserved acceptable reader-rated quality in most acquisitions, but quality declined sharply near 50% reduction, and approximately one in five women had a non-diagnostic view. External validation incorporating patient comfort and lesion-detection outcomes is needed before clinical application.
No takes yet. Share an insight, caveat, or question.
El-Sheikh et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: