BACKGROUND It is possible to evaluate the size of rotator cuff tears by ultrasonography (US) or magnetic resonance imaging. However, there are only a few reports on the imaging assessment of the configuration of cuff tears, which could provide important preoperative information that assists performing an optimal anatomical repair. The purpose of this study was to determine quantitatively the reproducibility of three-dimensional US in the assessment of rotator cuff tear configuration. METHODS Ten embalmed cadaveric shoulders with rotator cuff tears were examined. After resecting the proximal humerus with the rotator cuff, we put it in water and scanned it using high-resolution US with a three-dimensional linear probe. Actual tear lengths and widths were compared with sonographic measurements (Pearson correlation coefficient). By superimposing the real photographic image on the reconstructed three-dimensional image, we calculated the concordance rate (ratio of the concordance area to the tear area). RESULTS The actual tear length (16.6 +/- 7.1 mm; mean +/- SD) and width (8.4 +/- 4.4 mm) were correlated with the tear length (16.4 +/- 7.5 mm) and width (8.2 +/- 4.4 mm) measured from reconstructed three-dimensional ultrasonograms (r = 0.998 and 0.994, respectively). The mean concordance rate was 91.4%, indicating that almost exactly the same configuration was reconstructed by US. CONCLUSIONS Three-dimensional US is useful for evaluating the configuration of rotator cuff tears. This is the first report to quantify the similarity between the configuration evaluated by US and the actual configuration. Using this method, we can visualize the configuration of rotator cuff tears preoperatively, facilitating optimal repair design.