Imaging of Soft Tissues Adjacent to Orthopedic Hardware: Comparison of 3-T and 1.5-T MRI

Transcription

1 Musculoskeletal Imaging Original Research Farrelly et al. MRI of Orthopedic Hardware Musculoskeletal Imaging Original Research Cormac Farrelly 1 Amir Davarpanah 2 Stephen Brennan 3 Mathew Sampson 4 Stephen J. Eustace 1 Farrelly C, Davarpanah A, Brennan S, Sampson M, Eustace SJ Keywords: 1.5 T, 3 T, artifact reduction, MRI, susceptibility DOI: /AJR Received August 27, 2008; accepted after revision July 6, Department of Radiology, Cappagh National Orthopaedic Hospital, Finglas, Dublin 11, Republic of Ireland. Address correspondence to C. Farrelly (farrellycormac@ gmail.com). 2 Department of Radiology, Northwestern University Hospital, Feinberg School of Medicine, Chicago, IL. 3 Department of Orthopedic Surgery, Cappagh National Orthopaedic Hospital, Finglas, Dublin, Republic of Ireland. 4 Department of Radiology, Sports Surgery Clinic, Santry Demesne, Dublin, Republic of Ireland. WEB This is a Web exclusive article. AJR 2010; 194:W60 W X/10/1941 W60 American Roentgen Ray Society Imaging of Soft Tissues Adjacent to Orthopedic Hardware: Comparison of 3-T and 1.5-T MRI OBJECTIVE. The purpose of this study was to compare metal artifact reduction techniques at 1.5-T and 3-T MRI. MATERIALS AND METHODS. A titanium plate with steel screws was placed in a freshly harvested pig leg. The leg was imaged with 1.5-T and 3-T MRI. A T2-weighted turbo spin-echo sequence was used with echo-train lengths of 8, 16, 32, and 64 and a constant readout bandwidth of 31.2 khz. The images were compared qualitatively, and the optimal echotrain length was selected. Images were acquired at the optimal echo-train length with four different readout bandwidths. Artifact was measured quantitatively, and image quality was ranked qualitatively. The qualitatively best image acquired at 1.5 T was compared with the qualitatively highest-ranked image acquired at 3 T. RESULTS. At both 1.5 T and 3 T, optimal images of equal quality were produced at echotrain lengths of 8 and 16. At higher readout bandwidths, there was quantitatively less artifact. The qualitatively best images were acquired at a readout bandwidth of 31.2 khz at 1.5 T and 62.5 khz at 3 T (Cronbach s α = 1.00). The optimal image at 3 T was qualitatively superior to that at 1.5 T. CONCLUSION. Optimizing image acquisition parameters in this phantom model resulted in similar quantitative susceptibility artifact at 3 T and 1.5 T and better qualitative images at 3 T than at 1.5 T. J oint replacements such as total hip arthroplasty are among the most commonly performed operations worldwide. With an aging population and as the clinical threshold for surgery decreases, this number will continue to rise [1]. The rate of complications requiring surgical revision after total hip arthroplasty has been reported to be as high as 17.5% in certain centers [2]. MRI of these complications can be difficult because metallic substances in a magnetic field induce magnetic field inhomogeneity, which induces a number of artifacts. These artifacts are often collectively referred to as susceptibility artifact. Gradient-recalled echo sequences are particularly prone to metallic artifact because of intravoxel dephasing [3] (Fig. 1). In regions of steep changes in the local magnetic field, such as those adjacent to metallic prostheses, the magnetic field that acts on protons within a single voxel can vary strongly. Unlike spin-echo sequences, gradient-recalled echo acquisitions do not entail use of a 180 refocusing pulse and therefore do not cor- rect for static field inhomogeneities. The result is marked dephasing of transverse spins and local signal loss (Fig. 1) that is more pronounced at higher field strengths because local distortions are directly proportional to the strength of the main magnetic field. Standard spin-echo and turbo spin-echo (TSE) sequences are less prone to artifact from static magnetic field inhomogeneity because of their refocusing 180 pulses. They are, however, prone to artifact due to variable magnetic field inhomogeneity. This includes signal void in the vicinity of the prosthesis caused by dephasing from randomly diffusing proton spins and characteristic lines of signal void and signal increase due to spatial misregistration artifact. The use of TSE with smaller interecho spacing, in which the frequency-encoding direction is oriented away from the tissues of interest and the readout bandwidth is increased, can reduce the effect of these artifacts [4 18]. With these methods, 1.5-T MRI can be clinically useful in evaluating complications after spinal surgery [4, 5] and total hip arthroplasty [6 12]. The purpose of W60 AJR:194, January 2010

2 MRI of Orthopedic Hardware Fig. 1 Pig leg with titanium plate prosthesis and two stainless steel nails in situ. Gradient-echo 3-T MR image (TR/TE, 375/10; flip angle, 20 ) shows marked susceptibility artifact due to constant and variable distortions in local magnetic field induced by metallic orthopedic prosthesis. this phantom study was to image orthopedic hardware in an animal cadaver leg to compare quantitative measurements of susceptibility artifact and qualitative measurements of overall image quality of 1.5-T and 3-T MRI after optimization of sequence parameters for both field strengths. Materials and Methods A titanium plate was surgically attached with two steel screws to the tibia of a freshly harvested pig leg. This leg was imaged with 1.5-T MRI (Signa HDx system, GE Healthcare) followed immediately by 3-T MRI (Signa HDx system). A standard transmit receive knee coil was used for each MRI unit. Ethical approval for this animal cadaver study was not required at our institution. Imaging parameters were kept as near identical as possible for the 1.5-T and 3-T MRI units. A matrix, slice thickness of 4 mm, and spacing of 1 mm were used. The field of view was selected to fit the size and orientation of the pig leg in the coronal plane. T2-weighted TSE (TR/TE, 2,540 13,940/70) and a frequency-encoding direction oriented away from the tissues of interest were selected. Each image was evaluated at standard window settings of center 6,500 and width 18,000. The pig leg was weighed (8.5 kg), and the specific absorption rate (SAR) limitations were entered into the MRI acquisition protocols. An estimated SAR of 3 W/kg, average coil SAR of 3 W/ kg, and peak SAR of 6 W/kg were selected. Echo-train length (ETL) and readout bandwidth were varied at 1.5 T and 3 T. Readers were blinded to magnetic field strength and sequence. The leg was initially imaged at ETLs of 8, 16, 32, and 64 with a constant readout bandwidth of 31.2 khz. The images were qualitatively ranked from 1 to 4. The ETL producing the optimal image quality (rank 1) was selected. The ETL then was kept constant at the optimal length (rank 1), and the leg was imaged at readout bandwidths of 15.6, 31.2, 62.5, and 125 khz. Objective Image Quality Parameters Quantitative measurements were made by two experienced radiologists using standard software. The readers assessed the images quantitatively by measuring the thickness of the artifact caused by the titanium plate at a predefined location at one of the screw orifices, measuring the maximal marrow thickness of the tibia in the transverse plane, and calculating the ratio between the two values (Fig. 2). The ratio measurements and calculations were repeated for all sequences at the optimal ETL in identical slice positions. The mean of the measurements recorded by the readers for each sequence was the final measurement for the results. The ratio of artifact to marrow thickness was directly comparable between the sequences imaged on the same unit because the phantom was not moved between sequences. One blinded reader repeated all measurements 5 months later, and intraobserver variability was recorded. Subjective Image Quality Parameters The images were ranked qualitatively from 1 to 4. Two readers with at least 5 years of radiology experience evaluated the images. Qualitative assessment involved comparing the image quality of the four sequences with differing ETLs and the four sequences with differing readout bandwidths at both 1.5 T and 3 T. The images from each sequence were loaded into a workstation and directly compared side by side. For each group of four sequences, the readers were asked to take into account three specific factors: amount of identifiable susceptibility artifact caused by the prosthesis, appreciable image noise or mottle, and lack of definition or sharpness in the periprosthetic tissues. Fig. 2 Pig leg with orthopedic hardware in situ. Example of quantitative measurement. Turbo spinecho MR image (TR/TE, 5,120/70; ETL, 16; readout bandwidth, 15.6 khz) shows increased susceptibility artifact involving periprosthetic soft tissues compared with higher readout bandwidth used for Figure 4B. Thickness of artifact due to titanium plate was measured at predefined location (distal screw orifice, short line), and marrow thickness was measured in transverse plane at thickest level (long line). These two measurements were repeated for all sequences with echo-train length of 16 at identical slice position. Ratio of artifact to marrow thickness (0.8 in this case) was directly comparable between sequences imaged with same MRI unit because phantom was not moved between sequences. After taking these three factors into account, the readers decided on which images they would prefer to report and ascribed a ranking score of 1 4 to each group, 1 representing superior image quality with the best visualization of periprosthetic soft tissue and 4 being the worst in that group. For each group the observers commented on which one of the three artifacts predominated as the respective scores worsened. If they were found to be of equal quality for visualization of the periprosthetic soft tissues, the images were ranked with the same number. Finally, the highest-ranked qualitative images (rank 1) at 1.5 T and 3 T were compared and ranked qualitatively in the same manner. Statistical Analysis Analysis was performed with statistical software (SPSS version 17.0, SPSS). Agreement on qualitative image ranking and quantitative artifact measurement was determined with reliability statistics including intraclass correlation coefficient based on the Cronbach alpha model. Interreader and intrareader agreement was calculated for the quantitative measurements. Agreement of more than 0.41, 0.61, and 0.81 was considered moderate, good, and very good, respectively. Results In the qualitative analysis, varying ETL produced images of equal quality at ETLs of 8 and 16 (Table 1). This was true at 1.5 T and 3 T. When the ETL was increased to 32 and 64, the readers commented that bone and softtissue definition was progressively worse (subjectively more blurred) at 1.5 T and 3 T (Fig. 3). Agreement between the readers was very good (α = 1.00). At both 1.5 T and 3 T, an ETL of 16 (rank 1) was selected. As the readout bandwidth was increased, there was quantitatively less artifact from the titanium plate (Table 2) (Fig. 2). Both interreader (α = 0.991) and intrareader (α = 0.995) agreement on quantitative measurements were very good. AJR:194, January 2010 W61

3 Farrelly et al. TABLE 1: Qualitative Ranks of Images at Varied Echo-Train Lengths and Constant Readout Bandwidth of 31.2 khz Echo-Train Length Rank Echo-Train Length Rank Note Readers comments stated that definition of periprosthetic tissues worsened progressively with more blurring as echo-train length increased. TABLE 2: Quantitative Measurements of Images With Varied Readout Bandwidths and Constant Echo-Train Length of 16 Readout Bandwidth (khz) Ratio Readout Bandwidth (khz) Ratio Note At both field strengths, the thickness of the susceptibility artifact decreased in relation to marrow thickness of the tibia as the readout bandwidth increased. These measurements are directly comparable between sequences on the same MRI unit but not between sequences at different field strengths (Fig. 2). A According to the comments by the readers regarding the qualitative ranking, noise increased subjectively as readout bandwidth was increased. The optimal compromise between artifact reduction and signal-to-noise ratio (SNR), determined with the qualitative ranking scores, was readout bandwidths of 31.2 khz (1.5 T) and 62.5 khz (3 T). There was no difference in qualitative scores between the two readers (α = 1.00) (Table 3). Both readers ranked the optimal image obtained at 3 T (ETL, 16; readout bandwidth, 62.5 khz) qualitatively superior to the optimal image obtained at 1.5 T (ETL, 16; readout bandwidth, 31.2 khz) (Fig. 4), commenting that there was less image noise and mottle and better definition of the periprosthetic soft tissues at 3 T. Discussion MRI at 1.5 T is clinically useful for evaluating periprosthetic tissue [6, 7]. White et Fig. 3 Pig leg with titanium plate prosthesis and two stainless steel nails in situ. A, Turbo spin-echo 3-T MR image (TR/TE, 6,050/70; echo-train length, 32; readout bandwidth, 31.2 khz) shows more blurring (less sharpness) of periprosthetic soft tissues than in B. B, Turbo spin-echo 3-T MR image (4,400/70; echo-train length, 16; readout bandwidth, 31.2 khz ) shows less blurring than A. B al. [6] found complications in 11 of 14 patients with pain after total hip arthroplasty. Johnston et al. [7] imaged 28 patients, all of whom had undergone previous imaging with other techniques and had persistent unexplained pain after total hip arthroplasty. In 15 cases, the cause of the pain was explained by MRI findings, and in 11 of the 15 cases, the complications were managed locally without open revision surgery. The clinical use of 3-T MRI is increasing. Doubling the field strength offers a potential doubling of SNR. There is concern, however, that imaging of orthopedic hardware at 3 T can increase susceptibility artifact [19, 20]. Our study showed that methods similar to those used to reduce susceptibility artifact at 1.5 T, such as using TSE sequences, can be used at 3 T. In addition, although there is potentially more susceptibility artifact from surgical hardware at higher field strengths, this artifact can be offset by the use of higher readout bandwidth. Metallic hardware in a magnetic field acquires magnetism proportional to the strength of the main magnetic field. Adjacent protons and water molecules then precess under the influence of the local magnetic field and either have a different frequency or rapidly dephase. Signal is localized in space by frequency and phase. The phase-encoding direction is less prone to produce geometric distortions induced by magnetic field inhomogeneity. Susceptibility artifact represents a combination of intravoxel dephasing, diffusion-related signal loss, and spatial misregistration. Therefore, susceptibility artifact acquires an oval configuration elongated in the frequency plane [8, 9]. To compare artifacts at 1.5 T and 3 T, an imaging plane (coronal) that depicts the entire length of a prosthesis was selected. All other parameters were selected to minimize artifact on both MRI units. Previous reports [4 18] have described numerous methods of decreasing susceptibility artifact at 1.5 T. These methods include use of TSE sequences, orienting the frequency-encoding direction away from the tissue of interest, increasing readout gradient strength, decreasing field of view and voxel size, and using parallel imaging techniques. We used a matrix size in this study. The frequency-encoding direction was oriented along the long axis of the tibia to allow optimal visualization of the periprosthetic soft tissue, and the field of view was determined by the size of the prosthesis. Parallel imaging technique was not used. W62 AJR:194, January 2010

4 MRI of Orthopedic Hardware TABLE 3: Qualitative Ranking of Images With Varied Readout Bandwidths and Constant Echo-Train Length of 16 Readout Bandwidth (khz) Rank Readout Bandwidth (khz) Rank Note Observers commented that at lower readout bandwidth, artifact from the prosthesis predominated but that at higher bandwidths, image noise and mottle were more prominent. Use of a TSE sequence, which has a long, closely spaced train of refocusing 180 pulses, reduces susceptibility artifact compared with use of a standard spin-echo sequence [4 18]. Randomly diffusing proton spins adjacent to metallic devices cause dephasing of spinning protons. This effect is inversely proportional to the distance from the prosthesis and is more pronounced at T2-weighted imaging because longer TEs allow increased dephasing. Use of TSE helps to reduce diffusion-related signal loss relative to use of standard spin-echo technique because multiple 180 refocusing pulses are applied within a single TR. The resulting shorter effective TE (echo spacing) allows less time for dephasing before the spins are refocused. With longer ETLs, however, the echoes at the end of the echo train are more susceptible to dephasing. High-frequency data responsible for image detail are acquired at the beginning and at the end of the echo train, if conventional linear filling of the k-space is used. Therefore, more blurring (less image sharpness) is expected with longer ETLs because of progressive dephasing. At 3 T more signal is available but spins dephase more quickly. Although one might expect increased blurring at 3 T compared with 1.5 T MRI, this study showed, at similar ETLs, no qualitatively appreciable blurring at 1.5-T or 3 T with ETLs up to 16. Increasing the ETL above this number led to progressively worse image quality at both magnetic field strengths. Receiver bandwidth is the range of frequencies used to sample an MR signal. It is the rate at which the MR signal is converted from analog to digital by the digital converter. As receiver bandwidth is increased, the time needed to sample the echo is decreased. The adjustment is made by increasing the amplitude of the frequency-encoding gradient and keeping the field of view constant. A steeper frequency-encoding gradient increases the speed of rephasing. Thus the echo forms faster, the sampling time decreases, and the sampling rate increases. In this study, increasing the readout bandwidth decreased the quantitative susceptibility artifact. This effect occurred because increasing the readout bandwidth shortened the interval between echoes in the echo train and therefore less time was available for signal intensity to decline before refocusing occurred. In addition, increasing the readout bandwidth increases the number of frequencies sampled per pixel. Therefore, signal misregistration occurred in fewer pixels. A theoretical, although less marked, effect of increasing readout bandwidth and decreasing effective TE at 3 T is improvement in T2-weighted image contrast. This effect occurs because at higher field strengths, the T2 decay times of various tissues shorten [20 22]. Thus maintaining image contrast requires a decrease in TE. However, increasing readout bandwidth decreases SNR. This A effect occurs because for a given a voxel size and static field strength, the number of excited spins is defined and therefore so is the maximum amount of MR signal available for sampling. The readout of the MR signal can take more or less time, depending on the receiver bandwidth. A broad bandwidth corresponds to fast sampling of the MR signal and a high-intensity readout gradient. Background noise has a constant intensity at all frequencies (white noise). Therefore, the larger the receiver bandwidth, the more noise is sampled relative to the MR signal and the lower is the SNR. In this study, the optimal qualitative readout bandwidth at 1.5 T was 31.2 khz. At readout bandwidths greater than this value, images became progressively more grainy and noisy. At 3 T, it was possible to double the readout bandwidth to 62.5 khz while maintaining a qualitatively adequate SNR. The optimal image obtained at 3 T (ETL, 16; readout bandwidth, 62.5 khz) was qualitatively superior to that obtained at 1.5 T (ETL, 16; readout bandwidth, 31.2 khz). This somewhat counterintuitive finding can be explained because SNR is directly proportional to the main magnetic field strength and SNR is proportional to 1/readout bandwidth. Therefore at 3 T and double the readout bandwidth, in this case 62.5 khz, the SNR is approximately 40% ( 2) greater than at 1.5 T and 31.2 khz. Yet at 3 T and double the readout bandwidth, the susceptibility artifact is approximately the same as at 1.5 T Fig. 4 Pig leg with titanium plate prosthesis and two stainless steel nails in situ. A, MR image obtained at 1.5 T (TR/TE, 4,370/70; echo-train length, 16; readout bandwidth, 31.2 khz) and ranked 1. Thickness of signal void caused by titanium plate is similar to that in B. B, MR image obtained at 3 T (3,680/70; echo-train length, 16; readout bandwidth, 62.4 khz) and ranked 1. Imaging slice position is near identical to A. Susceptibility artifact includes signal void and lines of high signal intensity due to misregistration of protons. Thickness of signal void caused by titanium plate is similar to that in A. Although signal void owing to steel screw in superior aspect is more prominent, high-signal-intensity artifact is decreased. There is improved signal-to-noise ratio, and improved definition of soft tissues adjacent to orthopedic hardware. B AJR:194, January 2010 W63

5 Farrelly et al. because the amount of susceptibility (spatial misregistration and diffusion-related signal loss) approximates linear proportionality to the field strength [5, 23]. Therefore, the SNR increase due to the higher field strength dominates the signal loss due to the increased readout bandwidth. In our experience, imaging a prosthesis, once it has been cemented or surgically fixed with screws, is safe and does not cause movement at 1.5 T or 3 T. It must be stated, however, that the required safety testing should be performed for each piece of hardware at 3 T. The SAR safety limitations were not exceeded despite being set to the weight of the pig leg (8.5 kg). It is therefore likely that in a heavier human patient, the SAR limits also would not be exceeded. This study had limitations. Comparing 1.5- T and 3-T images entails comparing images obtained with different machines. Although both MRI units used in this study were made by the same manufacturer, it is possible that differences in image quality might have been due to differences in hardware design not related to the different field strengths. Second, susceptibility artifact on images is present in two dimensions (oval configuration in the frequency plane), but the measurements used to quantify susceptibility artifact in this study were in one plane only. Although the quantitative measurements used were in the direction of the tissues of interest, that is, into the periprosthetic tissues and not along the length of the bone, this limitation remains. Third, insofar as possible, the observers were blinded to whether the images were obtained with a 1.5-T or a 3-T unit. However, inherent qualitative differences in the acquired images might have allowed the observers to determine which images were obtained with the 1.5-T unit and which with the 3-T unit. This is a potential source of bias. Finally, because of slight differences in imaging planes, quantitative measurements were directly comparable between images obtained on the same MRI unit only. As such, qualitative scoring was used for comparing the highest-ranked image obtained at 1.5 T with the highestranked image obtained at 3 T. We conclude that although more artifact from metal prostheses is present on images obtained at higher field strengths, if all imaging parameters are kept the same, optimizing imaging parameters at 3 T by use of TSE and doubling the readout bandwidth will cause susceptibility artifact similar to that at 1.5-T MRI and improve qualitative measurements of soft-tissue visualization adjacent to orthopedic hardware. References 1. Birrell F, Johnell O, Silman A. Projecting the need for hip replacement over the next three decades: influence of changing demography and threshold for surgery. Ann Rheum Dis 1999; 58: Skutek M, Bourne RB, MacDonald SJ. International epidemiology of revision THR. Curr Orthop 2006; 20: Port JD, Pomper MG. Quantification and minimization of magnetic susceptibility artifacts on GRE images. J Comput Assist Tomogr 2000; 24: Tartaglino LM, Flanders AE, Vinitski S, Friedman DP. Metallic artifact on MR images of the postoperative spine: reduction with fast spin-echo techniques. Radiology 1994; 190: Petersilge CA, Lewin JS, Duerk JL, Yoo JU, Ghaneyem AJ. Optimizing imaging parameters for MR evaluation of the spine with titanium pedicle screws. AJR 1996; 166: White LM, Kim JK, Lee MJ, et al. Complications of total hip arthroplasty: MR imaging initial experience. Radiology 2000; 215: Johnston C, Kerr J, Ford S, O Byrne J, Eustace S. MRI as a problem-solving tool in unexplained failed total hip replacement following conventional assessment. Skeletal Radiol 2007; 36: Suh JS, Jeong EK, Shin KH, et al. Minimizing artifacts caused by metallic implants at MR imaging: experimental and clinical studies. AJR 1998; 171: Eustace S, Jara H, Goldberg R, et al. A comparison of conventional spin-echo and turbo spinecho imaging of soft tissues adjacent to orthopedic hardware. AJR 1998; 170: Eustace S, Shah B, Mason M. Imaging orthopedic hardware with an emphasis on hip prostheses. Orthop Clin North Am 1998; 29: Potter HG, Nestor BJ, Sofka CM, Ho ST, Peters LE, Salvati EA. Magnetic resonance imaging after total hip arthroplasty: evaluation of periprosthetic soft tissue. J Bone Joint Surg Am 2004; 86: Twair A, Ryan M, O Connell M, Powell T, O Byrne J, Eustace S. MRI of failed total hip replacement caused by abductor muscle avulsion. AJR 2003; 181: Naraghi AM, White LM. Magnetic resonance imaging of joint replacements. Semin Musculoskelet Radiol 2006;10: Törmänen J, Tervonen O, Koivula A, Junila J, Suramo I. Image technique optimization in MR imaging of a titanium alloy joint prosthesis. J Magn Reson Imaging 1996; 6: Eustace S, Goldberg R, Williamson D, et al. MR imaging of soft tissues adjacent to orthopaedic hardware: techniques to minimize susceptibility artifact. Clin Radiol 1997; 52: Olsen RV, Munk PL, Lee MJ, et al. Metal artifact reduction sequence: early clinical applications. RadioGraphics 2000; 20: Viano AM, Gronemeyer SA, Haliloglu M, Hoffer FA. Improved MR imaging for patients with metallic implants. Magn Reson Imaging 2000; 18: Buckwalter KA. Optimizing imaging techniques in the postoperative patient. Semin Musculoskelet Radiol 2007; 11: Lee MJ, Kim S, Lee SA, et al. Overcoming artifacts from metallic orthopedic implants at highfield strength MR imaging and multidetector CT. RadioGraphics 2007; 27: Bolog N, Nanz D, Weishaupt D. Muskuloskeletal MR imaging at 3.0 T: current status and future perspectives. Eur Radiol 2006; 16: Gold GE, Han E, Stainsby J, Wright G, Brittain J, Beaulieu C. Musculoskeletal MRI at 3.0 T: relaxation times and image contrast. AJR 2004; 183: Gold GE, Suh B, Sawyer-Glover A, Beaulieu C. Musculoskeletal MRI at 3.0 T: initial clinical experience. AJR 2004; 183: King KF. Signal acquisition and k-space sampling. In: Bernstein MA, ed. Handbook of MRI pulse sequences. Boston, MA: Academic Press, 2004: W64 AJR:194, January 2010