Comparison of Three-dimensional Isotropic T1-weighted Fast Spin-Echo MR Arthrography with Two-dimensional MR Arthrography of the Shoulder 1

Transcription

1 Note: This copy is for your personal non-commercial use only. To order presentation-ready copies for distribution to your colleagues or clients, contact us at Hye Jung Choo, MD Sun Joo Lee, MD, PhD Ok-Hwoa Kim, MD, PhD Seung Suk Seo, MD, PhD Jung Han Kim, MD Comparison of Three-dimensional Isotropic T1-weighted Fast Spin-Echo MR Arthrography with Two-dimensional MR Arthrography of the Shoulder 1 Purpose: To determine the accuracy of a three-dimensional (3D) isotropic T1-weighted fast spin-echo (FSE) magnetic resonance (MR) sequence as compared with a conventional two-dimensional (2D) sequence in the diagnosis of rotator cuff tears and labral lesions. Original Research n Musculoskeletal Imaging 1 From the Departments of Diagnostic Radiology (H.J.C., S.J.L.) and Orthopedic Surgery (S.S.S., J.H.K.), College of Medicine, Inje University Pusan Paik Hospital, Gaegeumdong Jin-gu, Busan , Republic of Korea; and Department of Diagnostic Radiology, College of Medicine, Inje University Haeundae Paik Hospital, Busan, Korea (O.H.K.). Received June 16, 2011; revision requested July 23; revision received August 23; accepted August 30; final version accepted September 21. Address correspondence to H.J.C. ( nayaa_neo@hanmail.net). q RSNA, 2012 Materials and Methods: Results: Conclusion: Institutional review board approval was obtained, and the informed consent requirement was waived. Forty-nine patients who had undergone direct or indirect shoulder MR arthrography with the 2D T1-weighted FSE sequence and the 3D isotropic T1-weighted FSE sequence and subsequent arthroscopy were included. Each MR imaging sequence was independently scored by two readers retrospectively for the presence of full- or partial-thickness tears of the supraspinatus (SST) and infraspinatus (IST) tendons and the subscapularis tendon (SCT) and labral lesions. Diagnostic performance based on each sequence type was compared by using the area under the receiver operating characteristic curve (AUC). Arthroscopic findings enabled confirmation of the presence of 17 full-thickness SST-IST tears, 18 partial-thickness SST-IST tears, four full-thickness SCT tears, 17 partial-thickness SCT tears, and 17 labral lesions. The AUCs for the readers using the 3D T1-weighted FSE sequence versus those obtained with the 2D sequence were versus for reader A and versus for reader B in the detection of rotator cuff tears and versus for reader A and versus for reader B in the detection of labral lesions. The mean AUCs between the 2D and 3D sequences were not significantly different, with the exception of partial-thickness SCT tears for one reader. The accuracy of 3D isotropic FSE MR arthrography may be comparable with that of conventional 2D MR arthrography in the diagnosis of rotator cuff tears and labral lesions with a shorter imaging time. q RSNA, 2012 Radiology: Volume 262: Number 3 March 2012 n radiology.rsna.org 921

2 Three-dimensional (3D) isotropic magnetic resonance (MR) imaging has intrinsic advantages in terms of shortened imaging time and multiplanar reformation. It has been used to image joints, such as the knee, ankle, and shoulder (1 7). In multiple performance studies of the knee performed to assess the diagnostic accuracy and reliability of 3D isotropic fast spin-echo (FSE) imaging versus those of two-dimensional (2D) conventional imaging, 3D isotropic FSE MR imaging has been proved to be competitive (1,5 7). However, in the shoulder joint, preliminary work has been performed with the 3D isotropic fast gradient sequence, and the results of a study in which researchers used the 3D FSE sequence, which is a fairly new technique, have not been reported (8,9). Thus, we evaluated the diagnostic performance of 3D isotropic T1-weighted FSE MR shoulder arthrography with regard to detection of rotator cuff tears and labral lesions and compared these results with conventional 2D MR arthrography findings. Materials and Methods This retrospective study was approved by the institutional review board for human research of Inje University Paik Hospital. Informed consent was waived. Subjects A total of 167 consecutive patients with shoulder discomfort underwent direct or indirect shoulder MR arthrography performed with 3D isotropic T1-weighted FSE and conventional 2D sequences Advance in Knowledge nn Conventional two-dimensional (2D) MR arthrography and three-dimensional (3D) isotropic T1-weighted fast spin-echo (FSE) MR arthrography of the shoulder yield similar diagnostic accuracy for rotator cuff tears and labral lesions, and the imaging time for 3D isotropic T1-weighted FSE MR arthrography is shorter than that for three planes of conventional 2D MR arthrography. between August 2010 and April Among these patients, 52 underwent shoulder arthroscopic surgery, of which three were excluded because the interval between MR examination and arthroscopic surgery was longer than 60 days. As a result, 49 imaging studies performed with shoulder MR arthrography were included in this study. The study population comprised 27 women (age range, years; mean age, 57.9 years) and 22 men (age range, years; mean age, 52.7 years). The mean interval between MR imaging and arthroscopic surgery was 14 days (range, 1 60 days). Two patients had a history of previous shoulder arthroscopic surgery due to rotator cuff tears: One had undergone surgery 4 years prior to the study, while the other had undergone surgery 4 months before the study. MR Imaging Twenty patients underwent direct MR arthrography, and 29 underwent indirect MR arthrography. For direct MR arthrography, an intraarticular contrast medium (Magnevist or Gadovist; Bayer Schering Pharma, Berlin, Germany) was injected with fluoroscopic guidance via an anterior approach. A 22-gauge spinal needle was used to inject ml of 0.5% diluted MR contrast material into the glenohumeral joint. MR imaging was initiated within 60 minutes after intraarticular injection. For indirect MR arthrography, MR contrast material (0.1 mmol per kilogram of body weight) was injected intravenously. Patients were instructed to smoothly exercise their shoulders for 15 minutes, after which MR images were obtained with a 3-T MR unit (Achieva 3.0 T TX; Philips Medical Systems, Best, the Implication for Patient Care nn Three-dimensional isotropic T1-weighted FSE MR arthrography of the shoulder shows similar accuracy as more conventional imaging in the evaluation of the rotator cuff tears and labral lesions of the shoulder, with a faster imaging time. Netherlands) and use of an eight-element phased-array shoulder coil. Conventional 2D MR imaging was performed first and was followed by 3D isotropic T1-weighted FSE imaging. Conventional 2D MR imaging consisted of axial, oblique coronal, and oblique sagittal fat-suppressed T1-weighted FSE sequences and oblique coronal and oblique sagittal T2-weighted FSE sequences. We performed the 3D isotropic fat-suppressed T1-weighted FSE (VISTA [volumetric isotropic turbo spin-echo acquisition]; Philips Medical Systems) sequence with 0.5-mm thickness in the oblique coronal plane, and the source data were reformatted into axial and oblique sagittal planes with 1-mm thickness. Postprocessing was performed by a technologist at the imaging workstation immediately after MR imaging; the time required for image reformation was about 1 minute. The imaging parameters of these sequences are summarized in Table 1. Image Analysis MR images were retrospectively and independently reviewed by two musculoskeletal radiologists (O.H.K., H.J.C.; 9 and 6 years of experience in musculoskeletal radiology, respectively) to Published online before print /radiol Content code: Radiology 2012; 262: Abbreviations: AUC = area under receiver operating characteristic curve CI = confidence interval FSE = fast spin echo IST = infraspinatus tendon SCT = subscapularis tendon SST = supraspinatus tendon 3D = three-dimensional 2D = two-dimensional Author contributions: Guarantor of integrity of entire study, H.J.C.; study concepts/study design or data acquisition or data analysis/ interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; literature research, H.J.C., S.J.L., O.H.K.; clinical studies, all authors; statistical analysis, H.J.C.; and manuscript editing, all authors Potential conflicts of interest are listed at the end of this article. 922 radiology.rsna.org n Radiology: Volume 262: Number 3 March 2012

3 Table 1 Parameters of MR Imaging Sequences Parameter Axial Fat-suppressed T1-weighted MR Imaging Coronal Fat-suppressed T1-weighted MR Imaging Coronal T2-weighted MR Imaging Sagittal Fat-suppressed T1-weighted MR Imaging Sagittal T2-weighted MR Imaging Coronal 3D Isotropic Fatsuppressed T1-weighted FSE MR Imaging Repetition time (msec) Echo time (msec) Matrix size 256 x x x x x x 300 Field of view (cm) Section thickness (mm) Intersection gap (mm) Echo train length No. of signals acquired Imaging time 2 min 34 sec 2 min 38 sec 3 min 15 sec 2 min 16 sec 2 min 40 sec 6 min 10 sec detect rotator cuff tears, including supraspinatus tendon (SST) and infraspinatus tendon (IST) tears, subscapularis tendon (SCT) tears, and labral lesions. The radiologists were blinded to all clinical information about the patients. To prevent recall bias, the radiologists reviewed the MR images in random order and in separate settings (2D images vs 3D images) at least 2 weeks apart. The presence of SST-IST tears and SCT tears was evaluated and followed by investigation into the presence of fullor partial-thickness tears in each SST- IST and SCT. Partial-thickness SST-IST tears were classified as either articular or bursal surface tears. Intrasubstance tears were not assessed because they could not be correlated with arthroscopic results (10). The MR criterion for diagnosis of a full-thickness tear was visualization of a complete defect from the articular to the bursal surface of the tendon. The criterion for diagnosis of a partial-thickness tear was visualization of a partial defect in the tendon along the articular or bursal surface (8). To detect an SCT tear, ancillary signs, such as the medial location of the long biceps tendon or atrophy of the subscapularis muscle, were considered (9,11). If there were two or more types of SST-IST tears that did not communicate with one another, each type of SST-IST tear was scored individually. The criteria for diagnosis of a labral lesion were lateral orientation of high signal intensity on coronal images and abnormal morphology or signal intensity of the labrum Table 2 Grading of Diagnostic Image Quality Score Diagnostic Quality Description 4 Excellent Sharp depiction of shoulder anatomy 3 Adequate Minor artifact that did not interfere with image interpretation 2 Questionable Impairment of depiction of shoulder anatomy 1 Nondiagnostic Insufficient image quality (12). We did not divide labral lesions according to direction (anterior, superior, or posterior) because of the lack of posterior or anterior labral lesions in our study. If a labral lesion was detected in any direction, it was interpreted as positive. Labral fraying, which may be considered a degenerative process, was scored as negative (13). The presence of abnormalities in the rotator cuff and labrum was scored by using a six-point confidence scale based on two sets of conventional 2D and 3D sequences (0, definitely absent; 1, probably absent; 2, possibly absent; 3, possibly present; 4, probably present; 5, definitely present). A score of 0 2 was considered negative, and a score of 3 5 was deemed positive. False-positive and false-negative cases on both sets of the sequences were reviewed again by the two readers who originally scored the abnormalities until consensus was reached. Image Quality During the review for rotator cuff tears and labral lesions, the two blinded radiologists also rated image quality. The two reviewers independently assigned an overall image quality score according to a four-point grading system (Table 2). Arthroscopic Surgery Arthroscopic findings were used as the reference standard. Surgery was performed by two orthopedic surgeons (S.S.S., J.H.K) with 20 years of experience in shoulder arthroscopy. Surgeons were not blinded to the diagnostic reports derived from the preoperative MR arthrographic images. The surgeons recorded the abnormalities of the rotator cuff and labrum in detail. Statistical Analysis Data were analyzed with the receiver operating characteristic method. Each reader s performance in assessing the rotator cuff and labrum was evaluated by using the area under the receiver operating characteristic curve (AUC). The difference between AUC values obtained with the 3D method and those obtained with the 2D method was evaluated by using a univariate z test. Radiology: Volume 262: Number 3 March 2012 n radiology.rsna.org 923

4 We considered a score of 3 or more to indicate a positive MR imaging result. Sensitivity and specificity for all types of rotator cuff tears and labral lesions were calculated. Image quality attained with the two methods was compared by using the Wilcoxon signed-rank test. A weighted k test was used to assess interobserver agreement. The degree of agreement was categorized as follows: a k value of less than 0 indicated poor agreement; a k value of , slight agreement; a k value of , fair agreement; a k value of , moderate agreement; a k value of , substantial agreement; and a k value of , near perfect agreement. Data were divided into two groups: direct MR arthrographic data and indirect MR arthrographic data. In each group, AUC values for performance with 2D methods were compared with those for performance with 3D methods for all kinds of rotator cuff tears and labral lesions. Statistical significance was assumed for P,.05. In addition, 95% confidence intervals (CIs) for differences of AUCs and image quality scores between the 2D and 3D methods were given instead of a post hoc power (14,15). All statistical analyses were performed by using statistical software (MedCalc, version , MedCalc, Mariakerke, Belgium; SAS, version 9.2 for Windows, SAS Institute, Cary, NC). Results Diagnostic Performance SST-IST tears. Arthroscopic surgery revealed 17 full-thickness tears and 18 partial-thickness tears (12 articularsided tears, six bursal-sided tears) of the SST and IST (Figs 1, 2). The AUCs reflecting diagnostic performance of readers A and B using 3D T1-weighted FSE sequences versus that of readers A and B using 2D T1-weighted FSE sequences were as follows: (a) and (P =.20; 95% CI: , 0.174), respectively, versus and (P =.32; 95% CI: , 0.069), respectively, in the detection of full-thickness tears; (b) Figure 1 Figure 1: MR images obtained with direct arthrography show arthroscopically proved articular surface partial-thickness tear of the SST-IST in a 56-year-old woman. Oblique coronal (a) fatsuppressed T1-weighted FSE and (b) 3D isotropic fat-suppressed T1-weighted FSE images show the defect (arrows) at the articular surface of the SST-IST. Both readers highly scored this lesion as an articular-surface partial-thickness tear of the SST-IST on a and b. and (P =.09; 95% CI: , 0.228), respectively, versus and (P =.32; 95% CI: 0.005, 0.018), respectively, in the detection of articular surface partial-thickness tears; and (c) and (P..99; 95% CI: , 0.202), respectively, versus and (P..99; 95% CI: , 0.309), respectively, in the detection of bursal surface partial thickness tears. AUC values, the difference Figure 2 Figure 2: MR images obtained with indirect arthrography show arthroscopically proved bursal surface partial-thickness tear of the SST-IST in a 69-year-old man. Oblique coronal (a) fatsuppressed T1-weighted FSE and (b) 3D isotropic fat-suppressed T1-weighted FSE images show high signal intensity (arrow) at the bursal surface of the SST-IST. Both readers highly scored this lesion as a bursal surface partial-thickness tear of the SST-IST on a and b. of AUCs between the two sequences, and sensitivity and specificity are summarized in Tables 3 and 4. There was one false-negative case in the detection of SST-IST tears and one false-positive case in the detection of articular partial-thickness tears of the 924 radiology.rsna.org n Radiology: Volume 262: Number 3 March 2012

5 Table 3 AUC Values with 2D Sequences and 3D T1-weighted FSE Sequence in Detection of SST-IST Tears Tear Type and Reader 2D FSE Sequence 3D T1-weighted FSE Sequence Difference* P Value All Reader A (0.923, 1.000) (0.875, 0.998) (20.008, 0.065).15 Reader B (0.875, 0.998) (0.827, 0.986) (0.012, 0.056).32 Reader A (0.854, 0.994) (0.775, 0.965) (20.057, 0.174).20 Reader B (0.832, 0.987) (0.854, 0.994) (20.038, 0.069).32 Partial thickness Reader A (0.827, 0.986) (0.738, 0.946) (20.036, 0.176).23 Reader B (0.742, 0.948) (0.761, 0.958) (20.039, 0.069).63 Articular surface Reader A (0.881, 0.999) (0.748, 0.951) (20.030, 0.228).09 Reader B (0.781, 0.968) (0.767, 0.961) (0.005, 0.018).32 Bursal surface Reader A (0.802, 0.976) (0.802, 0.976) (20.202, 0.202)..99 Reader B (0.699, 0.924) (0.802, 0.976) (20.142, 0.309).32 Note. Unless otherwise indicated, data are AUC scores. Data in parentheses are 95% CIs. * Data are mean differences of AUC scores between the 2D FSE group and the 3D T1-weighted FSE group. Data are all kinds of SST-IST tears, including full- and partial-thickness tears. Data include both articular surface partial-thickness tears and bursal surface partial-thickness SST-IST tears. Table 4 Sensitivity and Specificity of 2D FSE and 3D T1-weighted FSE Sequences in the Detection of SST-IST Tears Sensitivity (%) Specificity (%) Tear Type and Reader 2D FSE Sequence 3D T1-weighted FSE Sequence 2D FSE Sequence 3D T1-weighted FSE Sequence All* Reader A 97 (34/35) 97 (34/35) 93 (13/14) 79 (11/14) Reader B 97 (34/35) 97 (34/35) 93 (13/14) 93 (13/14) Reader A 94 (16/17) 82 (14/17) 97 (31/32) 97 (31/32) Reader B 94 (16/17) 94 (16/17) 94 (30/32) 97 (31/32) Partial thickness Reader A 89 (16/18) 83 (15/18) 87 (27/31) 87 (27/31) Reader B 83 (15/18) 89 (16/18) 90 (28/31) 90 (28/31) Articular surface Reader A 100 (12/12) 83 (10/12) 92 (34/37) 92 (34/37) Reader B 83 (10/12) 83 (10/12) 95 (35/37) 95 (35/37) Bursal surface Reader A 83 (5/6) 83 (5/6) 100 (43/43) 100 (43/43) Reader B 67 (4/6) 83 (5/6) 100 (43/43) 100 (43/43) Note. Data in parentheses are numbers of lesions. * Data are all kinds of SST-IST tears, including full- and partial-thickness tears. Data include both articular and bursal surface partial-thickness SST-IST tears. SST and IST for both techniques by both readers. The former was a 20% depth of bursal surface partial-thickness tear at arthroscopy; at second review of the MR images, a subtle bursal surface irregularity was identified. The latter was a full-thickness tear of the SST and IST, which was an alternatively false-negative case in the detection of full-thickness SST and IST tears. At a second review, the joint distention with contrast media was thought to be inadequate, and a high-grade articular surface partial-thickness tear (16) was identified. Arthroscopic findings Radiology: Volume 262: Number 3 March 2012 n radiology.rsna.org 925

6 revealed a large amount of synovitis coexistent with a full-thickness tear. SCT tears. Arthroscopic surgery revealed four full-thickness tears of the SCT and 17 partial-thickness tears (Fig 3). The AUCs reflecting diagnostic performance of readers A and B using the 3D method versus those reflecting diagnostic performance of readers A and B using the 2D method were as follows: (a) and (P..99; 95% CI: , 0.083), respectively, versus and (P..99; 95% CI: , 0.083), respectively, in the detection of full-thickness tears and (b) and (P =.04; 95% CI: , 0.144), respectively, versus and (P =.28; 95% CI: , 0.120), respectively, in the detection of partial-thickness tears. AUC values, the difference in AUC values between the two methods, and sensitivity and specificity are summarized in Tables 5 and 6. There were eight false-positive cases in the detection of partial-thickness SCT tears with both sequences by both readers. On a second review of these cases, six cases were still thought to be highly suspicious for small partialthickness tears (Fig 4). Labrum lesions. Arthroscopic surgery revealed 17 labral lesions (15 superior lesions, one anterior lesion, and one posterior labral lesion) (Fig 5). The AUCs reflecting diagnostic performance for readers A and B using the 3D technique versus those reflecting diagnostic performance for readers A and B using the 2D technique were as follows: (a) and (P =.43; 95% CI: , 0.107), respectively, versus and (P..99; 95% CI: , 0.070), respectively, in the detection of labral lesions. AUC values, the difference in AUC values between the two techniques, and sensitivity and specificity are summarized in Tables 5 and 6. There was one false-positive case and two false-negative cases in the detection of labral lesion with both sequences by both readers. At the second MR review of the false-positive case, a thick area of high signal intensity between the labrum and the biceps tendon Figure 3 Figure 3: (a, b) MR images obtained with direct arthrography show arthroscopically proved partialthickness tears of the SCT and SST-IST in a 56-year old woman. (a) Oblique sagittal fat-suppressed T1-weighted FSE and (b) reformatted sagittal 3D isotropic fat-suppressed T1-weighted FSE images show high signal intensity at the articular surface of the SCT (thick arrow). The contrast fillings (thin arrows) under the articular surface of the SST-IST are also visible. Both readers highly scored these as partial-thickness tears of the SCT and SST-IST. was identified, which might have been a normal deep sulcus between the origins of the biceps tendon and the labrum. On second review of the two false-negative cases, the labral lesions were not detected on MR images. Image Quality The mean image quality scores with the 3D technique for readers A and B versus those with the 2D technique for readers A and B were 3.9 and 3.8 (P =.05; 95% CI: , ), respectively, versus 3.8 and 3.6 (P =.03; Figure 4 Figure 4: MR images obtained with indirect arthrography show arthroscopically proved articular surface partial-thickness tear of the SST-IST and absence of a partial-thickness tear of the SCT in 53-year old man. (a) Oblique sagittal fat-suppressed T1-weighted FSE and (b) reformatted oblique sagittal 3D isotropic fat-suppressed T1-weighted FSE images show a small high-signal-intensity area (thick arrow) at the articular surface of the subscapularis tendon. Both readers false positively highly scored this lesion as a partial-thickness tear of the SCT. Both readers correctly scored abnormal high-signal-intensity areas (thin arrows) under the SST-IST as a partial-thickness tear on both images. The fluid collection at the subacromial-subdeltoid bursa was also identified on both images. 95% CI: , ), respectively. The image quality of the 3D isotropic T1-weighted FSE sequence in the two patients who had previously undergone surgery was scored highly by both readers. 926 radiology.rsna.org n Radiology: Volume 262: Number 3 March 2012

7 Table 5 AUC Values of 2D FSE and 3D T1-weighted FSE Sequences in the Detection of SCT Tears and Labral Lesions Abnormality and Reader 2D FSE Sequence 3D T1-weighted FSE Sequence Difference* P Value SCT tear All Reader A (0.729, 0.941) (0.671, 0.907) (20.019, 0.117).11 Reader B (0.681, 0.913) (0.650, 0.893) (20.067, 0.121).27 Reader A (0.907, 1.000) (0.907, 1.000) (20.083, 0.083)..99 Reader B (0.907, 1.000) (0.907, 1.000) (20.083, 0.083)..99 Partial thickness Reader A (0.704, 0.927) (0.629, 0.879) (20.012, 0.144).04 Reader B (0.658, 0.898) (0.629, 0.879) (20.069, 0.120).28 Labrum lesion Reader A (0.777, 0.966) (0.762, 0.958) (20.083, 0.107).43 Reader B (0.774, 0.964) (0.774, 0.964) (20.070,0.070)..99 Note. Unless otherwise indicated, data are AUC scores. Data in parentheses are 95% CIs. * Data are mean differences of the AUC score between the 2D FSE group and the 3D T1-weighted FSE group. Data are all kinds of SCT tears, including full- and partial-thickness tears. Table 6 Sensitivity and Specificity of 2D Sequences and 3D T1-weighted FSE Sequence in Detection of SCT Tears and Labral Lesions Sensitivity (%) Specificity (%) Abnormality and Reader 2D FSE Sequence 3D T1-weighted FSE Sequence 2D FSE Sequence 3D T1-weighted FSE Sequence SCT tear All* Reader A 90 (19/21) 90 (19/21) 71 (20/28) 68 (19/28) Reader B 95 (20/21) 100 (21/21) 57 (16/28) 61 (17/28) Reader A 100 (4/4) 100 (4/4) 98 (44/45) 98 (44/45) Reader B 100 (4/4) 100 (4/4) 98 (44/45) 98 (44/45) Partial thickness Reader A 82 (14/17) 82 (14/17) 77 (27/35) 74 (26/35) Reader B 88 (15/17) 94 (16/17) 66 (23/35) 69 (24/35) Labrum lesion Reader A 88 (15/17) 88 (15/17) 84 (27/32) 84 (27/32) Reader B 77 (13/17) 82 (14/17) 81 (26/32) 75 (24/32) Note. Data in parentheses are numbers of lesions. * Data are all kinds of SCT tears, including full- and partial-thickness tears. Interobserver Agreement Interobserver agreement was substantial to perfect for 2D and 3D sequences in the diagnosis of all types of rotator cuff tears and labral lesions (Table 7). Direct and Indirect Arthrography When the data were grouped according to direct and indirect MR arthrography, there were eight full-thickness SST-IST tears, six articular partial-thickness SST-IST tears, one full-thickness SCT tear, nine partial-thickness SCT tears, and 14 labral lesions in the direct MR arthrography group, while there were nine full-thickness SST-IST tears, six articular partial-thickness SST-IST tears, six bursal partial-thickness SST-IST tears, three full-thickness SCT tears, eight partial-thickness SCT tears, and 19 labral lesions in the indirect MR arthrography group. In the direct MR arthrography group, the AUCs for readers A and B using the 3D method versus those for readers A and B using the 2D method ranged from to and from to (P =.06 to P..99), respectively, versus from to and from to (P =.08 to P..99), respectively, in the detection Radiology: Volume 262: Number 3 March 2012 n radiology.rsna.org 927

8 Figure 5 Table 7 Interobserver Agreement for Evaluation of Rotator Cuff Tears and Labral Lesions Abnormality 2D FSE 3D T1-weighted FSE SST-IST tear All* 0.95 (0.888, 1.000) 0.90 (0.790, 1.000) 0.96 (0.869, 1.000) 0.91 (0.782, 1.000) Partial thickness 0.90 (0.784, 1.000) 0.85 (0.717, 0.982) Articular surface 0.82 (0.655, 0.997) 0.87 (0.718, 1.000) Bursal surface 0.94 (0.805, 1.000) 1.00 (1.000, 1.000) SCT tear All 0.78 (0.654, 0.908) 0.77 (0.641, 0.904) 0.78 (0.480, 1.000) 0.78 (0.480, 1.000) Partial thickness 0.70 (0.546, 0.862) 0.69 (0.530, 0.856) Labrum lesion 0.82 (0.722, 0.919) 0.79 (0.694, 0.881) Note. Data are k values, and data in parentheses are 95% CIs. * Data are all kinds of SST-IST tears, including full- and partial-thickness tears. Data include both articular surface partial-thickness tears and bursal surface partial-thickness tears of the SST-IST. Data are all kinds of SCT tears, including full- and partial-thickness tears. Figure 5: MR images obtained with direct arthrography show an arthroscopically proved superior labral lesion in a 29-year-old man. Oblique coronal (a) fat-suppressed T1-weighted FSE and (b) 3D isotropic fat-suppressed T1-weighted FSE images show high-signal-intensity extension (arrow) into the superior labrum. Both readers highly scored this finding as a labral lesion on both images. of rotator cuff tears and labral lesions. In the indirect MR arthrography group, the AUCs for readers A and B using the 3D technique versus those for readers A and B using the 2D technique ranged to and from to (P =.13 to P..99), respectively, versus from to and from to (P =.30 to P..99), respectively, in the detection of rotator cuff tears and labral lesions. The 95% CIs for differences of AUCs between the two sequences became wider, compared with the data when they were not grouped (Table 8). Discussion Until fairly recently, 3D SE imaging was difficult to implement in clinical practice because of its long imaging time (17,18). However, more recently, 3D FSE imaging has been used for a number of studies (19,20). We used a 3D isotropic FSE sequence in this study. This sequence uses flip angle sweep, short and nonvolume selective refocusing pulse, and sensitivity encoding, which makes it possible to get shorter echo spacing and a better signal-tonoise ratio. It can be used to produce T2-, intermediate-, or T1-weighted images (21). Beyond this, there are similar sequences that have been used to image the knee, ankle, abdomen, and pelvis (1,3,5 7,22 24). In this study, the diagnostic performance of 3D isotropic T1-weighted FSE MR arthrography was similar to that of 2D MR arthrography in several planes in the diagnosis of rotator cuff tears and labral lesions. We found no significant difference between 3D T1-weighted FSE images and conventional 2D images in terms of diagnostic performance as measured with the AUC value in the detection of rotator cuff tears and labral lesions, with the exception of partial-thickness tears of the SCT. For one of the two readers, the AUC value in the diagnosis of a partial tear of the SCT with 3D T1-weighted FSE images was significantly lower than that with 2D sequences. One possible reason for the low AUC value in the diagnosis of partial-thickness SCT tears might be the more prominent motion artifacts on 3D FSE images. Because the acquisition time of 3D T1-weighted FSE images is double that of one sequence in the 2D series and because the 3D FSE sequence was performed after the conventional 2D sequences, motion artifacts on 3D FSE images could be more prominent, possibly affecting the diagnostic accuracy of 3D isotropic images. Another possible reason for this discordance is that although arthroscopy is the best current reference standard, partial-thickness tears and mid- to distal footprint tears of the SCT can be missed because of poor visualization of the tendon footprint at arthroscopy (25,26). In our study, there were eight false-positive cases for detection of partial-thickness SCT tears by both readers and with both sequences. At second review, abnormal defects at the SCT were still definitely identified 928 radiology.rsna.org n Radiology: Volume 262: Number 3 March 2012

9 Table 8 AUC Values of Direct MR Arthrography and Indirect MR Arthrography in Each Sequence for Detection of Rotator Cuff Tears and Labral Lesions Direct MR Arthrography Indirect MR Arthrography Abnormality and Reader 2D FSE Sequence 3D T1-weighted FSE Sequence 95% CI of Difference P Value 2D FSE Sequence 3D T1-weighted FSE Sequence 95% CI of Difference P Value IST-SST tear All* Reader A , , Reader B , , Reader A , , Reader B , , Partial thickness Reader A , , Reader B , , Articular surface Reader A , , Reader B , , Bursal surface Reader A NA NA NA NA , Reader B NA NA NA NA , SCT tear All Reader A , , Reader B , , Reader A NA NA NA NA , Reader B NA NA NA NA , Partial thickness Reader A , , Reader B , , Labrum lesion Reader A , , Reader B , , Note. Unless otherwise indicated, data are AUC values. NA = not applicable. * Data are all kinds of SST-IST tears, including full- and partial thickness tears. Data include both articular surface partial-thickness tears and bursal surface partial-thickness tears of SST-IST. Data are all kinds of SCT tears, including full- and partial-thickness tears. in six cases, which was a large number compared with the number of incorrect diagnoses of other types of rotator cuff tears or labral lesions. We hypothesize that some small partial SCT defects that are not visible at arthroscopy could be detected on 3D T1-weighted FSE images in which section thickness is as thin as 1 mm and that the actual diagnostic performance of the 3D T1- weighted FSE sequence for SCT tears could be higher than that found in the present study. In our study, sensitivity and specificity in the detection of rotator cuff tears and labral lesions with 3D isotropic T1-weighted FSE MR imaging were similar to each other and were comparable with sensitivity and specificity reported in previous studies in which researchers used the 3D fast gradient technique (8,9). Recently, Chen et al (27) compared various 3D isotropic MR images in the evaluation of knee cartilage and concluded that 3D FSE images yielded the best signal-to-noise ratio and contrast-to-noise ratio among various 3D techniques. Although this result supports the assumption that the 3D FSE technique could be superior to the 3D fast gradient sequence in the diagnosis of the shoulder abnormality, further evaluation of this assumption is needed. This study had several limitations. First, we analyzed the MR images retrospectively. Second, the total sample size and the number of shoulder disorders were limited. This was reflected by Radiology: Volume 262: Number 3 March 2012 n radiology.rsna.org 929

10 the wide range of the 95% CI for the differences of the AUCs in some variables, especially bursal surface partialthickness SST-IST tears. Third, CI was used instead of post hoc power analysis to complement the insignificant results. Retrospective power analysis is most appropriate when it incorporates the minimum meaningful difference rather than the observed difference (28). However, in this study, it was difficult to obtain the appropriate value for the meaningful difference. Thus, CI which is the most widely advocated alternative to hypothesis testing and power calculation was used, although it cannot be considered a complete resolution (29). Fourth, there might have been a patient selection bias because we studied only patients who had undergone arthroscopic surgery. Fifth, a reader bias may have been introduced because the reviewers who evaluated the images knew that the patients had undergone arthroscopy. Sixth, the radiologists reviewed the 2D MR images first and the 3D MR images at least 2 weeks later to prevent recall bias. However, because all the 3D images were reviewed after the 2D images were reviewed, we cannot exclude the possibility that the lesion detection rate with the 3D sequence might have been affected. Seventh, although arthroscopy is the best current reference standard, it is not perfect because it is an operatordependent modality, it has known interobserver variability, and the precise evaluation of the footprint of the SCT has been deemed difficult, even with arthroscopy (25,26,30). Eighth, we did not obtain diagnostic performance data for labral lesions in the anterior, superior, or posterior directions because of the lack of posterior or anterior labral lesions in the patient group. Ninth, we included all the data for direct and indirect MR arthrography. Tenth, articular cartilage lesions were not accessed in this study, even though they are fairly common and may be a primary source of pain (31). In conclusion, 3D isotropic T1- weighted FSE imaging may yield diagnostic performance comparable with that of conventional 2D MR arthrography in the diagnosis of rotator cuff tears and labral abnormalities with a faster imaging time. A larger randomized study is needed to establish 3D isotropic T1-weighted FSE imaging as a substitutable method in patients suspected of having rotator cuff tears and labral lesions. Acknowledgment: We thank Minkyung Oh, PhD (Department of Pharmacology, College of Medicine, Clinical Trial Center, Inje University Busan Paik Hospital), for providing statistical support. Disclosures of Potential Conflicts of Interest: H.J.C. No potential conflicts of interest to disclose. S.J.L. No potential conflicts of interest to disclose. O.H.K. No potential conflicts of interest to disclose. S.S.S. No potential conflicts of interest to disclose. J.H.K. No potential conflicts of interest to disclose. References 1. Gold GE, Busse RF, Beehler C, et al. Isotropic MRI of the knee with 3D fast spin-echo extended echo-train acquisition (XETA): initial experience. AJR Am J Roentgenol 2007;188(5): Yao L, Pitts JT, Thomasson D. Isotropic 3D fast spin-echo with proton-density-like contrast: a comprehensive approach to musculoskeletal MRI. AJR Am J Roentgenol 2007;188(2):W199 W Stevens KJ, Busse RF, Han E, et al. Ankle: isotropic MR imaging with 3D-FSE-cube initial experience in healthy volunteers. Radiology 2008;249(3): Jung JY, Yoon YC, Choi SH, Kwon JW, Yoo J, Choe BK. Three-dimensional isotropic shoulder MR arthrography: comparison with two-dimensional MR arthrography for the diagnosis of labral lesions at 3.0 T. Radiology 2009;250(2): Jung JY, Yoon YC, Kwon JW, Ahn JH, Choe BK. Diagnosis of internal derangement of the knee at 3.0-T MR imaging: 3D isotropic intermediate-weighted versus 2D sequences. Radiology 2009;253(3): Kijowski R, Davis KW, Woods MA, et al. Knee joint: comprehensive assessment with 3D isotropic resolution fast spin-echo MR imaging diagnostic performance compared with that of conventional MR imaging at 3.0 T. Radiology 2009;252(2): Notohamiprodjo M, Horng A, Pietschmann MF, et al. MRI of the knee at 3T: first clinical results with an isotropic PDfsweighted 3D-TSE-sequence. Invest Radiol 2009;44(9): Magee T. Can isotropic fast gradient echo imaging be substituted for conventional T1 weighted sequences in shoulder MR arthrography at 3 Tesla? J Magn Reson Imaging 2007;26(1): Oh DK, Yoon YC, Kwon JW, et al. Comparison of indirect isotropic MR arthrography and conventional MR arthrography of labral lesions and rotator cuff tears: a prospective study. AJR Am J Roentgenol 2009;192(2): Torstensen ET, Hollinshead RM. Comparison of magnetic resonance imaging and arthroscopy in the evaluation of shoulder pathology. J Shoulder Elbow Surg 1999;8(1): Pfirrmann CW, Zanetti M, Weishaupt D, Gerber C, Hodler J. Subscapularis tendon tears: detection and grading at MR arthrography. Radiology 1999;213(3): Chang D, Mohana-Borges A, Borso M, Chung CB. SLAP lesions: anatomy, clinical presentation, MR imaging diagnosis and characterization. Eur J Radiol 2008;68(1): Maffet MW, Gartsman GM, Moseley B. Superior labrum-biceps tendon complex lesions of the shoulder. Am J Sports Med 1995;23(1): Hoenig JM, Heisey DM. The abuse of power. Am Stat 2001;55(1): Colegrave N, Ruxton GD. Confidence intervals are a more useful complement to nonsignificant tests than are power calculations. Behav Ecol 2003;14(3): Ellman H. Diagnosis and treatment of incomplete rotator cuff tears. Clin Orthop Relat Res 1990;(254): Yuan C, Schmiedl UP, Weinberger E, Krueck WR, Rand SD. Three-dimensional fast spin-echo imaging: pulse sequence and in vivo image evaluation. J Magn Reson Imaging 1993;3(6): Zhou X, Cofer GP, Suddarth SA, Johnson GA. High-field MR microscopy using fast spinechoes. Magn Reson Med 1993;30(1): Busse RF, Hariharan H, Vu A, Brittain JH. Fast spin echo sequences with very long echo trains: design of variable refocusing flip angle schedules and generation of clinical T2 contrast. Magn Reson Med 2006;55(5): Wang Z, Fernández-Seara MA. 2D partially parallel imaging with k-space surrounding neighbors-based data reconstruction. Magn Reson Med 2006;56(6): Qiao Y, Steinman DA, Qin Q, et al. Intracranial arterial wall imaging using three-dimensional high isotropic resolution black blood 930 radiology.rsna.org n Radiology: Volume 262: Number 3 March 2012

11 MRI at 3.0 Tesla. J Magn Reson Imaging 2011;34(1): Mugler JP 3rd, Bao S, Mulkern RV, et al. Optimized single-slab three-dimensional spin-echo MR imaging of the brain. Radiology 2000;216(3): Kim H, Lim JS, Choi JY, et al. Rectal cancer: comparison of accuracy of local-regional staging with two- and three-dimensional preoperative 3-T MR imaging. Radiology 2010;254(2): Rosenkrantz AB, Neil J, Kong X, et al. Prostate cancer: Comparison of 3D T2-weighted with conventional 2D T2-weighted imaging for image quality and tumor detection. AJR Am J Roentgenol 2010;194(2): David TS, Bravo H, Scobercea R. Arthroscopic visualization of subscapularis tendon lesions. Orthopedics 2009;32(9). 26. Koo SS, Burkhart SS. Subscapularis tendon tears: identifying mid to distal footprint disruptions. Arthroscopy 2010;26(8): Chen CA, Kijowski R, Shapiro LM, et al. Cartilage morphology at 3.0T: assessment of three-dimensional magnetic resonance imaging techniques. J Magn Reson Imaging 2010;32(1): Eng J. Sample size estimation: how many individuals should be studied? Radiology 2003;227(2): Penington A. Negative results and the limitations of power. ANZ J Surg 2008;78(1-2): Sasyniuk TM, Mohtadi NGH, Hollinshead RM, Russell ML, Fick GH. The inter-rater reliability of shoulder arthroscopy. Arthroscopy 2007;23(9): Polster JM, Schickendantz MS. Shoulder MRI: what do we miss? AJR Am J Roentgenol 2010;195(3): Radiology: Volume 262: Number 3 March 2012 n radiology.rsna.org 931