Original researchOpen Access

AbstractTop

Research question: Short volumetric interpolated breath hold exam (VIBE) 3D gradient echo sequences are comparable to T1 spin echo (SE) sequences for evaluating the shoulder labrum using arthroscopy as a gold standard. Methods: 24 patients with suspected labral tears had magnetic resonance (MR) arthrography of the shoulder using 1.5T MR scanner. Coronal oblique 3D VIBE FS sequences were performed and reconstructed in axial and sagittal oblique planes and compared with standard T1 SE sequences. Presence and location of labral tear and rotator cuff tears were noted. Findings were correlated with surgery where available. Time to obtain T1 SE sequences versus 3D VIBE acquisition was recorded. Results: Of 24 patients, 54% had arthroscopy. Sensitivity and positive predictive value (PPV) of both T1 SE and VIBE sequences for labral tears was 92%. 100% agreement for diagnosis of labral tear by readers between both sequences was noted. Specific labral tear location was accurate in 75% for both sequences. The mean time to obtain T1 SE sequences was 13.5 mins and 7.5 minutes for 3D VIBE sequences. Conclusion: Fast MR arthrography using 3D VIBE sequences shows excellent concordance with standard T1-FS sequence for diagnosis of labral tears. With shorter acquisition time and high diagnostic sensitivity 3D VIBE sequences maybe used as an alternative to standard sequences. Level of evidence: Level III retrospective diagnostic study.

Keywords: MRI arthrography; shoulder; gradient echo sequences; labrum; 3D VIBE sequences

IntroductionTop

Magnetic resonance (MR) arthrography with intraarticular contrast is the gold standard imaging technique for evaluation of shoulder joint instability. It is of particular importance in the evaluation of labroligamentous and cartilage abnormalities [1]. Standard shoulder magnetic resonance imaging (MRI) is limited in the evaluation of the glenoid labrum because of the lack of inherent joint distension. In comparison, direct MR arthrography results in distension of the joint capsule and improved visualisation of the normal anatomy and pathology. The contrast outlines intraarticular structures and delineates labral and capsular tears. The most common indication for MR arthrography is joint instability and pain and the advantages are particularly evident in the evaluation of acute pathology in young active patients.

Standard accepted sequences used in MR arthrography include T1- weighted spin echo and fast spin echo sequences with fat suppression (T1-FS) in multiple planes. Three plane T1 imaging has been advocated as the standard shoulder arthrogram protocol [2-4]. The duration of these sequences is on average 13.5 minutes. The length of these sequences can prove difficult for patients who have experienced recent trauma of the shoulder and find it difficult to keep still for the entire duration of the study. Patients with claustrophobia also find time in the scanner challenging and sometimes intolerable.

Gradient echo sequences such as isometric volumetric interpolated breath hold examination (VIBE) are of much shorter duration and are therefore ideal for patients that find it difficult to lie still for prolonged periods. Few studies have compared the efficacy of such sequences, but initial studies show them to be as sensitive as T1W sequences in the diagnosis of rotator cuff tears [5] and labral tears [6].

The purpose of this study was to evaluate if isometric 3D VIBE sequences are comparable or better than T1-FS standard sequences in the diagnosis of shoulder labral tears using arthroscopy as gold standard.

Materials and methodsTop

This was a prospective study carried out in a single tertiary referral centre. Over a 12 month period, 24 consecutive patients with clinical instability and suspected labral tears were enrolled in the study. Patients were identified by their orthopaedic surgeon and referred for MR arthrography. Ethical approval was obtained by the local ethical committee of University Hospital Galway.

Prior to MR a dilute gadolinium solution was routinely injected into each shoulder joint. With patient consent and under fluoroscopic guidance, the glenohumeral joint was injected using an anterior approach, at the junction of the middle and lower one third of the humeral head cortex using a 22G needle. To confirm adequate position in the joint capsule 0.5ml of iodinated contrast was injected. Subsequently 8 to 12ml of 0.1ml gadolinium in 20ml of saline dilution was injected intra-articularly. The patients were subsequently transferred for MRI using a 1.5 Telsa (Siemens Espree TIM 32 x 8) MR scanner. All MR studies were performed within thirty minutes of intraarticular injection. A dedicated shoulder transmit receiver coil was used for imaging. Patients were positioned in a neutral position with the thumb pointing upwards. MR arthrography was performed using coronal oblique, axial T1-FS SE and sagittal T1 SE sequences. Coronal oblique 3D VIBE FS sequences were performed at 0.7mm slice thickness and reconstructed in the axial and sagittal oblique planes with a slice thickness of 2mm. The voxels for the VIBE acquisition were isotropic thereby allowing multiplanar reconstruction in all planes. Protocols for both sequences are summarised in Table 1.

Routine T1 and 3D VIBE sequences were reviewed on separate sittings by two radiologists with fifteen and two years’ experience in musculoskeletal radiology. The reviewers were blinded to the patient details, clinical histories and the arthroscopic findings.

Interpretation of results was done in a consensus fashion with both radiologists. Initially MRI studies with the T1W images (excluding VIBE images) were reviewed separately and then 4 weeks later the MR studies with VIBE images (and excluding T1 SE weighted images) reviewed. Images at re-review were randomized and in different order than at first review. Any disagreement in results between radiologists was reviewed and a consensus was reached.

Labral abnormalities including tear and detachment were noted. A labral tear was defined as linear or diffuse increased T1 signal in the labrum which contacted the articular surface, amorphous shape of labrum indicating tear with scar and detachment as an increase T1 signal between the labrum and glenoid.

Normal variants were diagnosed and differentiated from a labral tear including a sublabral foramen, located at the 2’o clock position and representing localized detachment of the labrum from the glenoid rim and sublabral recess located at the 12’o clock position and represents a normal recess between the superior labrum and the cartilage of the glenoid cavity.

In order to describe the location of a tear, annotation of labrum was divided into anterior, inferior, posterior and superior regions and a clock face method used to describe the extent of the tear. The anterior labrum was defined from 1 to 5 o’ clock, the inferior from 5 to 7 o’ clock, the posterior from 7 to 11 o’ clock and the superior from 11 to 1 o’ clock. The labrum was evaluated on axial, sagittal oblique and coronal oblique images.

The MR labral findings were correlated with arthroscopic results where available. Arthroscopic results were regarded as the gold standard. Surgery was performed by a single orthopaedic surgeon with twelve years experience who was aware of the MR imaging report.

The separate radiological findings of the 3D VIBE sequences were compared with those from the standard T1-FS sequences. The combined time to obtain T1 SE sequences versus 3D VIBE acquisition, including post processing time was recorded.

Analysis was performed using SPSSv20. Simple descriptive statistics such as mean and standard deviation are used for continuous data, and frequencies and percentages for categorical data. Comparisons of results of surgeries were compared using Fisher’s exact test. A p-value less than 0.05 was deemed statistically significant.

ResultsTop

A total of 24 patients (age range from 18 to 63 years, mean age 30 years) had shoulder MR arthrography and were included in the study. Of this total cohort 13 patients (54%) had subsequent arthroscopy. Only 2 patients were over 50 years. 20 patients were male and 4 were females. 11 MR arthrograms were performed on the right shoulders and 13 on the left shoulder.

At arthroscopy 92% (12 of 13) patients had labral tears at surgery. One patient had a frayed superior labrum which was considered a positive result, contributing to patient symptoms. Of these, 10 (77%) involved the anterior labrum, 5 (38%) the posterior labrum, 6 (46%) the superior labrum and 9 (69%) the inferior labrum. All 12 patients had multiple or extensive tears hence the greater number of tears after subdividing the labrum.

The sensitivity of both T1 SE and VIBE sequences for the presence of labral tears was 92% with 100% agreement between VIBE and T1 SE sequences. There was one false negative for both T1-FS and VIBE sequences which was a SLAP tear at surgery. There were no false positives for either T1 or VIBE sequences.

Sensitivity of both VIBE and T1W sequences for anterior labral tears was 90%, identifying 9 out of 10 tears (Figure 1). Specificity of 33% was similar for both VIBE and T1-FS.

Figure 1 Axial T1-W (a) and VIBE (b) images displaying anteroinferior labral tears (white arrow) and a posterolateral depression of the humeral head indicating a Hills Sachs lesion (black arrow).

Sensitivity for posterior labral tears was 60% for VIBE sequences and 40% for T1 weighted sequences (Figure 2). There were no false positives for posterior tears.

Figure 2 T1-W (a) and VIBE (b) sequences in the same patient showing osseous Bankart with fracture of the anterior glenoid(white arrows) and a posterior labral tear (black arrows) as indicated by the amorphous shape and high signal intensity. Findings were confirmed at surgery.

On evaluation of VIBE sequences 4 of the 6 superior labral tears were identified and confirmed at surgery (sensitivity 67%). On evaluation of T1 sequences 3 of the 6 superior labral tears were accurately identified (sensitivity 50%). Of the 7 superior labra that were normal at surgery both VIBE and T1 sequences depicted 2 tears (specificity 78%) (Figure 3). Sensitivity of VIBE and T1W sequences for inferior labral tears was 100%: identifying all 9 tears (sensitivity 100%). Of the 4 labra that did not have abnormality of the inferior labrum VIBE depicted one tear (specificity 75%) and T1W depicted 2 tears (specificity 50%). A summary of the sensitivities and specificities are presented in Table 2.

Figure 3 Coronal T1-W (a) and Sagittal and coronal VIBE images (b) demonstrating a labral tear with extension into the long head of biceps tendon (arrow).