[Show abstract][Hide abstract] ABSTRACT: OBJECTIVE: To evaluate the clinical value of combining one-view mammography (cranio-caudal, CC) with the complementary view tomosynthesis (mediolateral-oblique, MLO) in comparison to standard two-view mammography (MX) in terms of both lesion detection and characterization. METHODS: A free-response receiver operating characteristic (FROC) experiment was conducted independently by six breast radiologists, obtaining data from 463 breasts of 250 patients. Differences in mean lesion detection fraction (LDF) and mean lesion characterization fraction (LCF) were analysed by analysis of variance (ANOVA) to compare clinical performance of the combination of techniques to standard two-view digital mammography. RESULTS: The 463 cases (breasts) reviewed included 258 with one to three lesions each, and 205 with no lesions. The 258 cases with lesions included 77 cancers in 68 breasts and 271 benign lesions to give a total of 348 proven lesions. The combination, DBT(MLO)+MX(CC), was superior to MX (CC+MLO) in both lesion detection (LDF) and lesion characterization (LCF) overall and for benign lesions. DBT(MLO)+MX(CC) was non-inferior to two-view MX for malignant lesions. CONCLUSIONS: This study shows that readers' capabilities in detecting and characterizing breast lesions are improved by combining single-view digital breast tomosynthesis and single-view mammography compared to two-view digital mammography. KEY POINTS: • Digital breast tomosynthesis is becoming adopted as an adjunct to mammography (MX) • DBT (MLO) +MX (CC) is superior to MX (CC+MLO) in lesion detection (overall and benign lesions) • DBT (MLO) +MX (CC) is non-inferior to MX (CC+MLO) in cancer detection • DBT (MLO) +MX (CC) is superior to MX (CC+MLO) in lesion characterization (overall and benign lesions) • DBT (MLO) +MX (CC) is non-inferior to MX (CC+MLO) in characterization of malignant lesions.
European Radiology 04/2013; 23:2087-2094. DOI:10.1007/s00330-013-2831-0 · 4.01 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: PURPOSE
To compare lesion detection and characterization by digital breast tomosynthesis (DBT) and mammography (MX) using free-response ROC methods.
METHOD AND MATERIALS
A multireader study was conducted to compare two-view MX to one-view DBT, including 371 breasts of 200 subjects (213 breasts with lesions (70 cancers), 158 without). Six breast radiologists, blinded to any clinical information, evaluated images from both modalities independently. Detection was limited to up to three findings per breast using BI-RADS classification for each modality. Data were analyzed per-lesion, using lesion localization fraction (LLF), the percentage of lesions correctly localized. Lesion detection was determined by calculating the total number of lesions rated above BI-RADS 1; the same was done separately for malignant and benign lesions. Lesion characterization was evaluated by counting malignant lesions rated above BI-RADS 3 and benign lesions scored BI-RADS 2 or 3. Lesion detection and characterization for all lesions, and stratified for malignant and benign lesions, were compared between DBT and MX using analysis of variance (ANOVA), adjusted according to the Obuchowski-Rockette model. P-values <0.05 were considered statistically significant. Uncertainty in estimated LLF differences between MX and DBT was quantified using two-sided, 95% confidence intervals.
The total number of lesions correctly localized and detected (LLF) by DBT was significantly higher than by MX (DBT = 69.1%, MX = 65.0%, p = 0.0001), with equal numbers of cancers (DBT = 79.8%, MX = 80.0%, p = 0.933), but a higher number of benign lesions detected and correctly localized by tomosynthesis (DBT = 65.0%, MX = 59.3%, p = 0.011). Correct lesion characterization was superior with tomosynthesis (DBT = 55.6%, MX = 50.9%, p = 0.01). Difference in cancer characterization was not statistically significant between the two modalities (DBT = 67.9%, MX = 72.1%, p = 0.258), while tomosynthesis was shown to be superior to mammography in characterization of benign lesions (DBT = 51.0%, MX = 42.8%, p = 0.028).
DBT was shown to be superior to MX in both detection and characterization of benign lesions, with comparable cancer detection capability.
The improved lesion detection and characterization of single-view DBT suggest that DBT could reduce screening callback rates and improve specificity compared to two-view digital mammography.
Radiological Society of North America 2011 Scientific Assembly and Annual Meeting; 11/2011
[Show abstract][Hide abstract] ABSTRACT: OBJECTIVE: We show with both a clinical case and measurements using the American College of Radiology mammography phantom that some digital mammography acquisition and image display combinations lead to both marked over- and underestimation of lesion sizes on geometrically magnified images. CONCLUSION: The results of this study indicate that the accuracy of lesion size measurements in all magnification modes should be a routine quality control acceptance test with each acquisition-display system combination in digital mammography.
American Journal of Roentgenology 01/2010; 194(1):W115-8. DOI:10.2214/AJR.09.2927 · 2.73 Impact Factor