Can we predict lesion detection rates in second-look ultrasound of MRI-detected breast lesions? A systematic analysis. Issue 113 (April 2019)

• Record Type: Journal Article
• Authors: Bumberger, Alexander; Clauser, Paola; Kolta, Michael; Kapetas, Panagiotis; Bernathova, Maria; Helbich, Thomas H.; Pinker, Katja; Baltzer, Pascal A.
• Is Part Of: European journal of radiology. Issue 113(2019)
• Page Start: 96
• Page End: 100
• Publication Date: 2019-04
• Subjects: Breast -- Breast neoplasms -- Ultrasonography -- Mammary -- Clinical decision making -- Magnetic resonance imaging; Medical radiology -- Periodicals; Radiology -- Periodicals; Radiologie médicale -- Périodiques; Medical radiology; Periodicals; 616.075705
• DOI: 10.1016/j.ejrad.2019.02.008
• Languages: English
• ISSNs: 0720-048X

Abstract

Highlights: Breast size, composition and lesion features predict whether MRI-detected lesions are visible on second-look ultrasound. Combined features can stratify the probability of an MRI-detected lesion to be found by second-look ultrasound as low, intermediate and high. These findings could be used to guide clinical decisions regarding whether to do second-look ultrasound after MRI. Abstract: Purpose: To predict detection rates (DR) in second-look ultrasound of MRI-detected breast lesions by systematically combining clinical and anthropomorphic features. Methods: A total of 104 suspicious breast-lesions, that were initially detected on breast MRI and underwent subsequent SLU from January 2013 through December 2013, were evaluated in this retrospective analysis. All images were reviewed by an experienced radiologist for this study. Both anthropomorphic, spatial and BI-RADS lesion features were recorded. Uni- and multivariate Classification and Regression Trees (CRT) statistics were used to predict SLU DR by these features. Results: Among 104 MRI-detected lesions, 58 (55.8%) showed a correlate on SLU. In univariate analysis, homogeneous fatty or dense fibro-glandular-tissue-composition (FGT) as assessed by ultrasound, segmental non-mass-distribution pattern and small breast size as assessed by MRI were significantly associated with higher DR on SLU. The remaining BI-RADS features did not significantly affect SLU DR according to our data. The predictive model could stratify the likelihood of SLU correlates as high, intermediate and low according to FGT, lesion type, size and position. Conclusions: By systematically combining the features FGT, lesion type, size and position, we could predict SLU DR of MRI-detected breast lesions. This may help to decide the preferable method for lesion biopsy or follow-up in clinical practice.