3.0-T MRI of the supraspinatus tendon

MRI of the shoulder has been found to be highly sensitive and specific for detection of full-thickness supraspinatus tendon tears at 1.5-T or lower field strength compared with arthroscopy. MRI of the shoulder at 3.0 T has not been specifically assessed. This study assesses the sensitivity and specificity of MRI at 3.0 T for supraspinatus tendon tears compared with arthroscopy.

Two experienced musculoskeletal radiologists retrospectively reviewed MR images of the shoulder in 150 consecutive patients who had subsequent arthroscopy. All patients had oblique coronal and sagittal T1-weighted and fat-saturated T2-weighted axial, oblique coronal, and sagittal imaging performed. The radiologists interpreted the MR images by consensus without knowledge of the arthroscopy results. Scans were interpreted for full-thickness or partial-thickness supraspinatus tendon tears. If partial-thickness supraspinatus tendon tears were seen on MRI, the reviewers noted whether the partial-thickness tear was articular or bursal in location. The radiologists also separated the supraspinatus tendon tears into small (< 1 cm retraction from the humeral head) and large (> 1 cm retraction from the humeral head). All 150 patients went on to arthroscopy. After consensus review of the MR images, arthroscopy results were compared with consensus MR interpretations.

Ninety-eight of the 150 patients had full-thickness supraspinatus tendon tears at arthroscopy. Twenty-six of the 150 patients had partial-thickness supraspinatus tendon tears. Seventeen of these 26 partial-thickness tears were along the articular surface and nine were along the bursal surface. Ninety-six of 98 full-thickness tears seen at arthroscopy were seen on consensus MRI interpretation. All 26 partial-thickness tears seen at arthroscopy were seen at consensus MR interpretation; however, two of the partial-thickness articular surface tears seen at arthroscopy were interpreted as full-thickness tears on consensus MRI interpretation. Twenty-eight of the 98 full-thickness supraspinatus tendon tears were small tears (< 1 cm retraction from the humeral head) on arthroscopy. Two of these 28 small tears seen on arthroscopy were not seen on consensus MRI interpretation. Twenty-six patients had intact supraspinatus tendons on both retrospective consensus MRI interpretation and at arthroscopy.

MRI of the shoulder at 3.0 T is highly sensitive and specific compared with arthroscopy in the detection of full-thickness and partial-thickness supraspinatus tendon tears.