Send to

Choose Destination
See comment in PubMed Commons below
Rofo. 2006 Dec;178(12):1224-34.

[Characterization of sonographically detected breast lesions using three-dimensional data sets].

[Article in German]

Author information

Institut für Radiologie, Charité Universitätsmedizin Berlin.



The diagnostic value of 3D ultrasound (US) was evaluated in comparison to 2D US on the basis of a set of pre-defined criteria for breast lesions. The individual criteria were correlated with the histological findings and they were ranked according to their significance for lesion characterization in both 3D US and 2D US.


A total of 92 breast lesions were examined of which 61 were malignant and 31 were benign, as confirmed by histology of core biopsy specimens (Aplio 80, Toshiba, Otawara, Japan). The 2D and 3D data sets were stored digitally. The vascular tree was stored separately using 3D fusion. In addition, a power Doppler (PD) examination was performed. A total of 644 individual images were evaluated by two independent readers. The lesions were evaluated using established criteria (BI-RADS classification). For statistical evaluation, the individual US measurements were classified according to the following two groups: those suggesting malignant lesions and those suggestive of benign lesions. Pearsons's chi-square test was performed for both groups. The 2D and 3D techniques were compared directly using kappa (kappa) statistics. Odds ratios (OR) were determined for the purpose of weighting the individual criteria. The sensitivities and specificities of 2D and 3D US were calculated.


The diagnostic value of 3D US was comparable to that of 2D US, with only slight differences observed for specific measurements. Analysis of 2D US data indicated that there was a strong correlation between malignant tumor growth and two specific 2D criteria: lesion margin and effect of the lesion on adjacent breast tissue (r (2) = 0.632, p < 0.001). Analysis of 3D images indicated that the significant measurements for tumor malignancy were shape and lesion margin in sagittal, transverse (r (2) = 0741, p < 0.001), and frontal views (r (2) = 0.592, p < 0.001) as well as retraction and lesion demarcation in the frontal views (r (2) = 0.649, p < 0.001). The comparison of 2D and 3D yielded primarily moderate (kappa = 0.412) to good (kappa = 0.789) agreement between the individual criteria. Crucial criteria for lesion characterization were the effect of the lesion on adjacent breast tissue (OR 47.8 [16.0-143.3]) in 2D US and the lesion margin (OR 36.4 [15.3-86.6]) and the 3 additional criteria of the frontal plane (OR 21.9-32.9) in 3D US. The two US techniques had comparable sensitivities (2D/3D 95-97 %/92-95 %) and specificities (2D/3D 54-58 %/61-67 %). The reconstructed 3D view of the vascular tree was subjectively considered to be helpful by both readers. A higher degree of vascularization was observed with 3D US.


The diagnostic role of 3D ultrasound in the diagnostic evaluation of breast cancer is comparable to that of 2D US. Our findings illustrate the significance of the evaluation of frontal views in lesion characterization.

[Indexed for MEDLINE]
PubMed Commons home

PubMed Commons

How to join PubMed Commons

    Supplemental Content

    Full text links

    Icon for Georg Thieme Verlag Stuttgart, New York
    Loading ...
    Support Center