Optimization of T2-weighted imaging for shoulder magnetic resonance arthrography by synthetic magnetic resonance imaging

Abstract

Background

Synthetic magnetic resonance imaging (MRI) allows reformatting of various synthetic images by adjustment of scanning parameters such as repetition time (TR) and echo time (TE). Optimized MR images can be reformatted from T1, T2, and proton density (PD) values to achieve maximum tissue contrast between joint fluid and adjacent soft tissue.

Purpose

To demonstrate the method for optimization of TR and TE by synthetic MRI and to validate the optimized images by comparison with conventional shoulder MR arthrography (MRA) images.

Material and Methods

Thirty-seven shoulder MRA images acquired by synthetic MRI were retrospectively evaluated for PD, T1, and T2 values at the joint fluid and glenoid labrum. Differences in signal intensity between the fluid and labrum were observed between TR of 500–6000 ms and TE of 80–300 ms in T2-weighted (T2W) images. Conventional T2W and synthetic images were analyzed for diagnostic agreement of supraspinatus tendon abnormalities (kappa statistics) and image quality scores (one-way analysis of variance with post-hoc analysis).

Results

Optimized mean values of TR and TE were 2724.7 ± 1634.7 and 80.1 ± 0.4, respectively. Diagnostic agreement for supraspinatus tendon abnormalities between conventional and synthetic MR images was excellent (κ = 0.882). The mean image quality score of the joint space in optimized synthetic images was significantly higher compared with those in conventional and synthetic images (2.861 ± 0.351 vs. 2.556 ± 0.607 vs. 2.750 ± 0.439; P < 0.05).

Conclusion

Synthetic MRI with optimized TR and TE for shoulder MRA enables optimization of soft-tissue contrast.

Keywords

Magnetic resonance arthrography (MRA)synthetic magnetic resonance imaging (MRI)contrast optimization shoulder rotator cuff tear

Introduction

Rotator cuff tear (RCT) of the shoulder is the most common clinical problem encountered in orthopedic practice. Magnetic resonance imaging (MRI) after intraarticular arthrography, i.e. direct MR arthrography (MRA), is an important diagnostic tool for accurate diagnosis of RCT as well as evaluation of labral pathology. Gadolinium-based contrast agents with optimized concentrations of gadolinium should be administrated intra-articularly before MRA. Intra-articularly injected gadolinium-based contrast agents increase soft-tissue contrast and distend the joint capsule, thus enhancing the resolution of internal structure of joints (1,2). For this reason, MRA should be performed with an optimized concentration of contrast media using optimized protocols (3,4). Previous studies on evaluation of optimal concentration of contrast media have demonstrated highest signal intensities of injected joint fluid at 1.25–2.0 mmol/L contrast media at 1.5-T as well as 3-T (3,5). However, despite MRA with optimized protocols, optimized concentrations of contrast medium are sometimes inadequate for shoulder imaging, which can result in inadequate contrast between joint fluid and adjacent soft tissue, e.g. intermediate or low signal intensity on T2-weighted (T2W) images after intraarticular gadolinium injection. This phenomenon could be due to pre-existing joint effusion, synovitis, or patient-related factors that affect the signal intensity of the contrast medium-combined joint fluid. Additionally, signal intensity is also affected by patient size and location of radiofrequency coil.

In conventional MRI, image contrast is influenced by two parameters—repetition time (TR) and echo time (TE)—which are determined in advance. Synthetic MRI was recently introduced to enable rapid acquisition and accurate quantification of MR images (6 –10), which makes it possible to reformat various synthetic images by adjusting scanning parameters such as TR and TE (11) using the following formula (10,12):

\begin{matrix} Signalintensity, S = PD \cdot exp (- TE / T 2) \\ \cdot (1 - exp (- TR / T 1)) \end{matrix}

Given that the values of T1, T2, and proton density (PD) of joint fluid and adjacent soft tissues vary among patients, we hypothesized that patient-specific optimized MR images can be reformatted by adjusting TR and TE to achieve maximum tissue contrast between joint fluid and adjacent soft tissue. However, no study till date has addressed the optimization of contrast agent concentration for synthetic MRI. The purposes of this study were: (1) to demonstrate the method for optimization of TR and TE by synthetic MRI; and (2) to validate the optimized TR and TE values by comparison with conventional shoulder MRA images.

Material and Methods

Study population

This retrospective study included 39 patients with shoulder pain, who underwent shoulder MRA, including MRA with synthetic MRI sequence, between March and July 2016. Of the 39 patients, nine had undergone rotator cuff repair. Two patients, one with synthetic images of suboptimal quality for quantification and another with insufficient joint fluid at the glenohumeral joint, were excluded. Thus, 37 patients were finally included in the analysis. The mean age ±standard deviation of the 37 patients was 55.4 ± 10.9 years (age range = 28–78 years; 15 men, 22 women). This study was reviewed and approved by the institutional review board. Informed consent for MRA was obtained from all patients.

Magnetic resonance imaging protocol

All patients underwent shoulder MRA. Diluted dimeglumine gadopentetate solution (15–18 mL) comprising 0.08 mL Magnevist® (Bayer Schering Pharma AG, Berlin, Germany), 18 mL normal saline, and 2 mL iodine contrast was intra-articularly injected into the glenohumeral joint under fluoroscopic guidance before MRA.

MRI was performed with a 3.0-T MR system (Discovery 750w, GE Healthcare, Waukesha, WI, USA) with a 16-channel GEM Flex medium flexible coil (NeoCoil, Pewaukee, WI, USA). Synthetic MR images were acquired using the following imaging parameters: oblique coronal synthetic MRI sequence: TR = 4589 ms; TE = 26.592 and 98.784 ms; inversion time (IT) = 12.208 ms; field of view (FOV) =140 × 140 mm; acquisition matrix = 384 × 224; slice thickness = 3 mm (interslice gap = 0.3 mm); flip angle = 90° and 110°; and echo train length = 14. Image acquisition time for the synthetic MRI sequence was 6 min 20 s. For conventional T2W fast spin-echo MRI, the following parameters were used: TR =4500–5500 ms; TE = 75–85 ms; FOV = 140 × 140 mm; acquisition matrix = 384 × 256; slice thickness = 3 mm (interslice gap = 0.3 mm); flip angle = 80°; and echo train length = 16. Image acquisition time for the conventional T2W fast spin-echo sequence was 3 min 9 s.

Image post-processing and analysis

The values of PD, T1, and T2 at the glenohumeral joint fluid and glenoid labrum were measured by drawing a region of interest (ROI) with the synthetic MRI plugin (Magnetic Resonance Image Compilation, MAGiC®; GE Healthcare) installed in the MRI console. These measurements were repeated three times, and the average values were recorded. Signal intensities were calculated with the following formulae:

\begin{matrix} Signalintensityofjointfluid, \\ S_{jo int} = {PD}_{jo int} \cdot exp (- TE / T 2_{jo int}) \\ \cdot (1 - exp (- TR / T 1_{jo int})) \\ Signalintensityoflabrum, \\ S_{labrum} = {PD}_{labrum} \cdot exp (- TE / T 2_{labrum}) \\ \cdot (1 - exp (- TR / T 1_{labrum})) \end{matrix}

To determine the point with maximum joint fluid–labrum contrast in T2W images, differences in signal intensity between joint fluid and labrum were simulated by varying TR and TE within ranges of 500–6000 ms and 80–300 ms, respectively, using a dedicated software written by Interactive Data Language (IDL) (ExelisVIS, Visual Information Solutions, Boulder, CO, USA).

Analysis of conventional and synthetic MR images

Two sets of synthetic images were reformatted for analysis using the synthetic MRI plugin (MAGiC®; GE Healthcare): synthetic images reformatted with the same TR and TE (synthetic T2W images) as conventional T2W images; and synthetic images reformatted with optimized TR and TE to demonstrate the maximum joint fluid–labrum contrast (optimized synthetic T2W images). Image sets of (i) conventional, (ii) synthetic, and (iii) optimized synthetic T2W images were analyzed in consensus by two musculoskeletal radiologists with two and ten years of clinical experience, who were blinded to the medical records of the patients.

Diagnostic agreement for supraspinatus tendon lesions among the three image sets was evaluated in consensus in patients with no history of surgical repair (n = 28). Tendon lesions were categorized into three types: normal/tendinopathy; partial thickness tears; and full thickness tears. Diagnostic agreement for superior labral lesion among the three image sets was also evaluated in consensus in patients with no history of surgical repair nor biceps tenotomy/tenodesis (n = 30). Superior labral lesions were categorized into two types: normal and wearing/tear.

The two radiologists independently evaluated the three image sets. The three images sets were then scored for image quality on the basis of contrast between contrast-injected joint fluid and labrum/joint capsule in all patients (n = 37) using a three-point scale: 3 = good; 2 = moderate; and 1 = poor contrast.

Statistical analysis

Diagnostic agreements for supraspinatus tendon lesions and superior labral lesions between conventional and synthetic T2W images and synthetic and optimized synthetic T2W images were evaluated by Kappa analysis. Comparison of image quality of joint fluid among the three imaging sets was accomplished by one-way analysis of variance with post-hoc analysis using the Bonferroni test. The inter-observer agreements of image qualities between conventional and synthetic T2W images and synthetic and optimized synthetic T2W images were evaluated using intraclass correlation coefficient (ICC) statistics. Statistical analysis was performed with a statistical software (R package 3.1.2; The R Foundation for Statistical Computing, Vienna, Austria). Values of P < 0.05 were considered to indicate statistically signiﬁcant differences.

Results

Optimized TR and TE for maximum contrast between joint fluid and labrum/joint capsule were determined from the simulation (Fig. 1). While tissue contrast increased with increments in TR, it plateaued after a specific TR value, which was determined to be the optimized TR (2724.7 ± 1634.7 ms; range = 6000–1000 ms). The mean optimized TE in 36 of 37 patients was 80.1 ± 0.4 ms; in most cases, optimized TE was determined at a minimum value (i.e. 80 ms). In the remaining one patient, however, the optimized TE was 82 ms.

Fig. 1.

Optimization of the repetition time (TR) and echo time (TE) from the tissue-difference map. Maximal contrast-to-noise ratiois located around a TE of 80 ms and TR of 2300 ms (arrow).

The results of evaluation of diagnostic agreement for supraspinatus tendon abnormality indicated excellent agreement between conventional and synthetic T2W images (0.882; 95% confidence interval [CI] = 0.725–1.039) as well as conventional and optimized synthetic T2W images (0.940; 95% CI = 0.824–1.056; Fig. 2). In 26 of 28 patients, the three image sets exhibited concordance in terms of diagnosis (tendinopathy/normal tendon = 6; partial thickness tear = 14; full thickness tear = 6; Fig. 3). In two patients, conventional T2W images exhibited high signal intensity and suspicious focal discontinuity at the supraspinatus tendon although the two sets of synthetic images did not exhibit definite high signal intensity (Fig. 4). The results of diagnostic agreement for superior labral lesions indicated excellent agreement between conventional and synthetic T2W images (0.851; 95% CI = 0.651–1.051) as well as conventional and optimized synthetic T2W images (0.927; 95% CI = 0.786–1.068).

Fig. 2.

Diagnostic agreement for supraspinatus tendon abnormalities. The results indicated excellent agreement between conventional T2W images and synthetic T2W images (0.882; 95% CI = 0.725–1.039) and conventional T2W images and optimized synthetic T2W images (0.940; 95% CI = 0.824–1.056).

Fig. 3.

Partial thickness tear of the supraspinatus tendon in a 38-year-old man. (a) Conventional T2W images (TR/TE = 5099/76 ms), (b) synthetic T2W images (TR/TE = 5100/76 ms), (c) and optimized synthetic T2W images (TR/TE = 1480/80 ms) all exhibit high signal intensity and focal discontinuity at the supraspinatus tendon (arrows). Synthetic image sequences have the drawback of appearance of artifacts (dashed arrows).

Fig. 4.

Partial thickness tear of the supraspinatus tendon in a 49-year-old man. (a) Conventional T2W images (TR/TE = 5281/83 ms) exhibit high signal intensity and focal discontinuity at the supraspinatus tendon (arrow). (b) However, synthetic T2W images (TR/TE = 5300/83 ms) and optimized synthetic T2W images (TR/TE = 5620/80 ms) images do not exhibit a definite high signal intensity at the supraspinatus tendon. Synthetic image sequences have the drawback of appearance of artifacts (dashed arrows).

There were significant differences in image quality scores for joint space among the conventional, synthetic, and optimized T2W images (P = 0.026; Fig. 5). The image quality score was significantly higher for the optimized synthetic T2W images than it was for the conventional T2W images on post-hoc analysis (P = 0.023). In seven of 37 patients, both synthetic and optimized synthetic T2W images exhibited higher scores than conventional T2W images (Fig. 6). In two of 37 patients, optimized synthetic T2W images exhibited higher scores than the other two image sets. In the remaining 28 patients, there were no significant differences in image quality scores among the three sets of images (Fig. 7). Qualitative image quality evaluation showed excellent agreement of in conventional T2W images (0.823; 95% CI = 0.657–0.909) and fair agreements in synthetic T2W images (0.598; 95% CI = 0.219–0.793) and optimized synthetic T2W images (0.554; 95% CI = 0.133–0.770). A summary of image quality scores is presented in Table 1.

Fig. 5.

Mean scores of image quality of joint fluid. Image quality scores of optimized synthetic images were significantly different from those of synthetic and conventional T2W images.

Fig. 6.

Oblique coronal magnetic resonance arthrography images of a 48-year-old man. (a) Conventional T2W images (TR/TE = 4673/84 ms) and (b) synthetic T2W images (TR/TE = 4700/84 ms) exhibit intermediate signal intensity at the joint fluid (arrow). However, optimized synthetic T2W images (TR/TE = 1140/80 ms) exhibit high signal intensity at the joint fluid (asterisk).

Fig. 7.

Oblique coronal MRA images of a 35-year-old woman. (a) Conventional T2W images (TR/TE = 4583/81 ms), (b) synthetic T2W images (TR/TE = 4600/81 ms), and optimized synthetic T2W images (4360/80 ms) all exhibit adequate tissue contrast at the joint.

Table 1.

Comparison of mean scores of image quality of joint fluid (mean ± standard deviation).

Conventional T2W images	Synthetic T2W images		Overall P value
Conventional T2W images	Same TR/TE	Optimized TR/TE	Overall P value
2.556 ± 0.607	2.750 ± 0.439	2.861 ± 0.351*	0.026^†

Significantly different from conventional T2W images (P value by post-hoc analysis = 0.023).

†

Statistically significant (one-way analysis of variance).

T2W, T2-weighted; TR, relaxation time; TE, echo time.

Discussion

For successful MRA evaluation of the shoulder joint, outstanding soft-tissue contrast with optimized concentration of contrast medium is essential. Previous studies on optimization of contrast medium concentration (3,4,13) have suggested methodologies for mixture or optimization of contrast concentration, which are now relatively well established. However, these methods are mainly focused on fat-suppressed T1 or T2W images. Non-fat-saturated T2W images exhibit issues related to joint-fluid contrast, which can result in inadequate image quality (Fig. 6a). The recently introduced synthetic MRI technology allows post-acquisition modulation of TR and TE with just one image (6 –11). Values of TR and TE were optimized to achieve ideal soft-tissue contrast after synthetic MRI of the shoulder joint in the present study.

Diagnostic agreement for supraspinatus tendon lesions between conventional and reformatted synthetic MR T2W images in the present study was excellent, as expected. However, in two patients, synthetic images did not exhibit definite high signal intensity at the supraspinatus tendon although conventional T2W images exhibited high signal intensity and focal discontinuity, suggestive of partial thickness tear. Such discrepancy could be one of the drawbacks of synthetic MRI, and, therefore, the diagnostic performance of synthetic MRI should be evaluated before the technology may be widely employed in the clinical setting. In a meta-analysis of diagnostic accuracy of rotator cuff injuries (14), the sensitivities of MRI and MRA for partial thickness tears were determined to be 0.68 and 0.83, respectively, while the specificities were in the range of 0.93–0.94; the values of sensitivity and specificity of MRA for tendinopathy were 0.79 and 0.94, respectively. Considering the high accuracy of conventional MRA, the accuracy of synthetic MRI for tendon visualization should be thoroughly evaluated. We expect to perform further comparative studies on synthetic MRI.

In the present study, we comparatively evaluated the image qualities of the three image sets based on the high signal intensity of the contrast-injected joint space relative to that of the labrum/joint capsule. Regardless of TR and TE optimization, synthetic MRI provided better image quality with high signal intensity of joint fluid than conventional T2W imaging in nine of 37 patients. However, in two of 37 patients, optimized synthetic MR images exhibited better delineation of the shoulder joint space, with bright signal intensity, than conventional and synthetic T2W images. The joint fluid–labrum contrast was maximized by adjusting the values of TR and TE to determine the optimized values. This concept of achieving maximal tissue contrast can be extended to other tissues as well. Furthermore, in the present study, optimized TR and TE for maximal tissue contrast were found to vary among patients, which indicates that patient-specific MR parameters should be optimized. Future studies should perform patient-oriented evaluation for improvement of image quality.

The synthetic MRI were reformatted directly in MRI scanner, and the reformat time was less than 30 s in our setting (Discovery 750w, GE Healthcare). The scan time can be greatly reduced if synthetic MRI could replace conventional MRI. However, it is possible to replace the conventional MRI by two preconditions: (i) acceptable image quality; and (ii) equivalent diagnostic performance. These issues should be further evaluated thoroughly.

The present study has a few limitations. First, we were unable to evaluate fat-saturated T1W sequences because the current version of the synthetic MRI module does not provide fat-saturation or water-excitation. We expect future studies to evaluate the feasibility of synthetic MRI with fat-saturation function for shoulder MRA. Second, synthetic MRI produced imaging artifacts, including wraparound artifacts (delineated in Figs. 3 and 4; dashed arrows). In our study, there were the wraparound artifacts in two patients among 37 patients. The artifact might affect the diagnostic performance by hiding the glenoid labrum. This resulted from the lack of no-phase wrap option in the current version of the synthetic MRI module, which is expected to be remedied in subsequent versions.

In conclusion, application of synthetic MRI with optimized TR and TE to MRA is not only feasible, but also enables optimization of soft-tissue contrast. This imaging technique could enable patient-specific improvement of MR image quality.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by the National Research Foundation (NRF) grant funded by the Korean government, Ministry of Science, ICT & Future Planning (MSIP, 2015R1A2A1A05001887).

References

Schulte-Altedorneburg

Gebhard

Wohlgemuth

et al.

MR arthrography: pharmacology, efficacy and safety in clinical trials. Skeletal Radiol 2003; 32: 1–12.

Steinbach

Palmer

Schweitzer

. Special focus session. MR arthrography. Radiographics 2002; 22: 1223–1246.

Masi

Newitt

Sell

et al.

Optimization of gadodiamide concentration for MR arthrography at 3 T. Am J Roentgenol 2005; 184: 1754–1761.

Shang

Zheng

. Direct MR arthrography of the shoulder at 3 Tesla: optimization of gadolinium concentration. J Magn Reson Imaging 2009; 30: 229–235.

Montgomery

Morrison

Schweitzer

et al.

Effects of iodinated contrast and field strength on gadolinium enhancement: implications for direct MR arthrography. J Magn Reson Imaging 2002; 15: 334–343.

Zhu

Hutchinson

Hawnaur

et al.

Magnetic resonance image synthesis using a flexible model. Br J Radiol 1994; 67: 976–982.

Riederer

Suddarth

Bobman

et al.

Automated MR image synthesis: feasibility studies. Radiology 1984; 153: 203–206.

Warntjes

Kihlberg

Engvall

. Rapid T1 quantification based on 3D phase sensitive inversion recovery. BMC Med Imaging 2010; 10: 19–19.

Warntjes

Dahlqvist

Lundberg

. Novel method for rapid, simultaneous T1, T2*, and proton density quantification. Magn Reson Med 2007; 57: 528–537.

10.

Warntjes

Leinhard

West

et al.

Rapid magnetic resonance quantification on the brain: Optimization for clinical usage. Magn Reson Med 2008; 60: 320–329.

11.

Bobman

Riederer

Lee

et al.

Cerebral magnetic resonance image synthesis. Am J Neuroradiol 1985; 6: 265–269.

12.

Meara

Barker

. Evolution of the longitudinal magnetization for pulse sequences using a fast spin-echo readout: application to fluid-attenuated inversion-recovery and double inversion-recovery sequences. Magn Reson Med 2005; 54: 241–245.

13.

Kopka

Funke

Fischer

et al.

MR arthrography of the shoulder with gadopentetate dimeglumine: influence of concentration, iodinated contrast material, and time on signal intensity. Am J Roentgenol 1994; 163: 621–623.

14.

Roy

Braen

Leblond

et al.

Diagnostic accuracy of ultrasonography, MRI and MR arthrography in the characterisation of rotator cuff disorders: a systematic review and meta-analysis. Br J Sports Med 2015; 49: 1316–1328.