JOURNAL OF MAGNETIC RESONANCE IMAGING 41:1570–1580 (2015)

Original Research

Knee Implant Imaging at 3 Tesla Using High-Bandwidth Radiofrequency Pulses Theresa J. Bachschmidt, Dipl-Phys,1,2* Reto Sutter, MD,3,4 Peter M. Jakob, PhD,1 Christian W.A. Pfirrmann, MD, MBA,3,4 and Mathias Nittka, PhD2 Key Words: magnetic resonance imaging; metal artifact; 3.0 Tesla; total knee arthroplasty J. Magn. Reson. Imaging 2015;41:1570–1580. C 2014 Wiley Periodicals, Inc. V

Background: To investigate the impact of highbandwidth radiofrequency (RF) pulses used in turbo spin echo (TSE) sequences or combined with slice encoding for metal artifact correction (SEMAC) on artifact reduction at 3 Tesla in the knee in the presence of metal. Methods: Local transmit/receive coils feature increased maximum B1 amplitude, reduced SAR exposition and thus enable the application of high-bandwidth RF pulses. Susceptibility-induced through-plane distortion scales inversely with the RF bandwidth and the view angle, hence blurring, increases for higher RF bandwidths, when SEMAC is used. These effects were assessed for a phantom containing a total knee arthroplasty. TSE and SEMAC sequences with conventional and high RF bandwidths and different contrasts were tested on eight patients with different types of implants. To realize scan times of 7 to 9 min, SEMAC was always applied with eight slice-encoding steps and distortion was rated by two radiologists. Results: A local transmit/receive knee coil enables the use of an RF bandwidth of 4 kHz compared with 850 Hz in conventional sequences. Phantom scans confirm the relation of RF bandwidth and through-plane distortion, which can be reduced up to 79%, and demonstrate the increased blurring for high-bandwidth RF pulses. In average, artifacts in this RF mode are rated hardly visible for patients with joint arthroplasties, when eight SEMAC slice-encoding steps are applied, and for patients with titanium fixtures, when TSE is used. Conclusion: The application of high-bandwidth RF pulses by local transmit coils substantially reduces through-plane distortion artifacts at 3 Tesla.

1 Department of Experimental Physics 5, University of Wuerzburg, Wuerzburg, Germany. 2 Magnetic Resonance, Siemens AG, Erlangen, Germany. 3 Department of Radiology, Orthopedic University Hospital Balgrist, Zurich, Switzerland. 4 University of Zurich, Faculty of Medicine, Zurich, Switzerland. Contract grant sponsor: Siemens AG. *Address reprint requests to: T.J.B., Healthcare Sector, Siemens AG, Allee am Roethelheimpark 2, D-91052 Erlangen, Germany. E-mail: [email protected] Received April 15, 2014; Accepted August 4, 2014. DOI 10.1002/jmri.24729 View this article online at wileyonlinelibrary.com. C 2014 Wiley Periodicals, Inc. V

FRACTURES OF THE femur and tibia are commonly treated with metal plate osteosynthesis (1), and degenerative joint diseases of the knee are addressed by total knee arthroplasty (TKA). From 1990 through 2002, the annual number of primary total knee arthroplasties in the United States nearly tripled to 381,000 (2) and an annual amount of 3.48 million is expected for 2030. Until then, a revision rate in the range of 7.2% to 7.8% is predicted (3), and a significant proportion of patients with metal fixtures in the knee or full joint arthroplasties require follow-up imaging studies. Magnetic resonance imaging has become the preferred imaging modality for evaluating the musculoskeletal system (4). Advantages of 3 Tesla (T) over 1.5T MRI have been reported for various orthopedic issues (5–7) and 3T is becoming more established in clinical routine (8). The application of magnetic resonance imaging in the evaluation of joint arthroplasties and metal fixtures has been limited in the past by severe artifacts, but it remains the preferred imaging modality due to its outstanding soft-tissue contrast (9). The main source of artifacts is induced by differences in susceptibility between metal hardware and surrounding tissue, and these artifacts increase with higher field strengths (10). Many studies focusing on the reduction of artifacts for knee implants were limited to 1.5T (4,11–13). Historically, metal-artifact reduction focused on in-plane artifacts, i.e., the misregistration of signal along the frequency-encoding direction, and comprised following suggestions: the use of turbo spin echo (TSE) sequences with short effective echo times (TE) and small echo spacing instead of spin echo or gradient echo sequences; the application of three-dimensional (3D) sequences; decreasing voxel size and slice thickness; increasing slice-select and frequency-encoding gradients; using short tau inversion recovery (STIR) technique for fat

1570

Knee Metal Implant Imaging at 3T

suppression; choosing the direction for phase encoding with the knowledge of least artifacts in this direction and positioning the long axis of the metal implant in parallel with B0 (13–15). View angle tilting (VAT) can correct in-plane distortion, but in turn may increase blurring (16,17). The effect of readout bandwidth in the presence of metal implants was analyzed in detail (11,18–20). A comparison of imaging in the presence of orthopedic hardware at 1.5T and 3T considered readout bandwidth and echo train length in a TSE sequence only and concluded that a doubled readout bandwidth can address the increased level of artifacts at 3T and thus preserve image quality (21). Through-plane artifacts, i.e., the distortion of the excited slice profiles, were not addressed until the advent of new approaches opened new ways to tackle them: slice encoding for metal artifact correction (SEMAC) (22) and multi-acquisition with variable resonance image combination (MAVRIC) (23). Offresonance suppression (ORS) (24) can be used in addition to VAT or SEMAC to limit through-plane artifacts. The use of thin slices and high-bandwidth RF pulses for excitation have been suggested as amendments to these options (22,25). The effect of high-bandwidth RF pulses was tested on a small scale (26) and optimization of the bandwidth of the RF inversion pulses was reported for a STIR sequence (27), both at 1.5T and without the use of through-plane correction methods. RF pulse bandwidths in the range of 1.25 to 2 kHz have been reported at 1.5T (12,22,27–29) and 1 to 1.6 kHz at 3T (30), using the body coil for transmission. To encounter the enhanced level of artifacts for total joint arthroplasties at 3T, the application of SEMAC with a sufficient amount of additional encoding is necessary. However, this leads to long scan times which are problematic in terms of limited patient comfort, increased likeliness of motion artifacts and finally economic considerations. Scan times may be further prolonged by the increased SAR levels at 3T as compared to 1.5T field strength. The purpose of this study is to fully exploit state-ofthe-art MRI hardware to enhance MRI at 3T in the presence of plate osteosynthesis in the knee and total knee arthroplasties: local transmit coils for the knee are widely-used among MSK-imaging centers on most current scanner platforms, featuring increased RF transmit field (B1) amplitudes, reduced SAR levels and multiple receive coil elements for parallel imaging acceleration. These properties are highly attractive for implant imaging, because high B1 amplitudes enable RF pulses of increased bandwidth, which cannot be obtained with a whole body-transmit coil both due to RF power and SAR limitations. The validation of high-bandwidth RF pulses with respect to through-plane distortion, image quality and scan time reduction for SEMAC and TSE sequences in presence of metal is the objective of this work.

MATERIALS AND METHODS Effects of RF Bandwidth The bandwidth of the RF pulse fBW impacts the distortion of an off-resonant spin both in- and

1571

through-plane. A spin is excited, if its off-resonance df complies with df/fBW ¼ dz/zExc, while dz represents its distance to the center of the desired slice and zExc the nominal slice thickness. Hence, increased RF bandwidths combined with constant slice thickness reduce through-plane distortion, but also allow higher off-resonances to contribute to the final signal, i.e., in-plane distortion can be increased. Thinner slices combined with constant RF bandwidth also decrease through-plane distortion, although more slices are required to cover the anatomy, which may increase scan time. SEMAC is a technique to correct throughplane distortion: Phase encoding in slice direction resolves through-plane distortion for each slice separately. To avoid signal pile-ups due to aliasing, the number of required slice-encoding steps (SESr) to cover the off-resonance is determined by SESr ¼ [max(df)min(df)]/fBW. All described forms of distortion and the associated signal voids and pile-ups in the apparent image are summarized as artifacts. Because the extent of through-plane distortion scales inversely with the excitation bandwidth, an obvious approach to reduce artifacts is to implement RF pulses with increased bandwidth. The bandwidth of the RF pulse is limited by the maximum transmit B1 field provided by the scanner hardware and by SAR restrictions. In this work, RF pulses are selected whose bandwidths scale inversely with their durations t; hence, the energy deposition in the patient is affected by the bandwidth of the RF pulse as follows: Rt SAR / B02 fBW ( B1 dt)2. RThe squared dependence of t SAR on the flip angle a ¼ B1 dt makes the flip angle an important parameter: its reduction allows either a lower SAR deposition or an increased RF pulse bandwidth for a given maximum B1 field. If this parameter is kept constant, SAR scales linearly with the bandwidth of the RF pulse. RF Pulses The main scope of this work is to compare imaging sequences equipped with increased-bandwidth RF pulses versus RF pulses similar to those used by the default sequences provided by the manufacturer. Modifications were implemented only for use in this research study. In general, the bandwidth is limited by refocusing pulses, because aRef > aExc and fBW / t1 for an RF pulse of defined shape, i.e., the selection of the pulse is a tradeoff between duration t, maximum B1 and the quality of the slice profile. As a reference, the RF pulses with maximum bandwidth provided by the standard product TSE sequence are taken and the pulse duration of the refocusing pulses is minimized under the condition to achieve a flip angle of 180 for any load. The duration of the excitation pulse is adjusted to achieve equal RF bandwidth as the refocusing pulse. The reference excitation pulse is an SLR pulse of time-bandwidth product (TBP) 2.2 with a bandwidth of 850 Hz. Windowed SINC-pulses (TBP ¼ 2) are used for refocusing and their duration is selected such that the bandwidth equals 850 Hz for SEMAC and 758 Hz for TSE for ORS. These reference pulses are referred to as conventional mode in the

1572

Bachschmidt et al.

by the STIR method (25). The bandwidth of the selective adiabatic inversion pulse is selected equal to the excitation bandwidth in both conventional and HiBW mode and corresponding gradient amplitudes are matched (27). Hardware

Figure 1. Schematic of the phantom setup with total knee arthroplasty (TKA) made of CoCr28Mo6 with a polyethylene insert, polypropylene structures on both sides (marked in blue) and a string for fixation during production (a), B0 map of central slice (b), through-plane analysis (c): highest signal intensity of all voxels in each singularly reconstructed SEMAC partition of different slices contributing to the central slice in high-bandwidth (HiBW) and conventionalbandwidth mode (Conv).

In this work, the strength of the B1 field of the local transmit coil shall be exploited to maximize the bandwidth of the RF pulses and hence reduce the number of required SEMAC partitions SESr and the total acquisition time at 3T. A knee coil (Quality Electrodynamics [QED], Mayfield Village, OH) with one channel for transmission and 15 channels for reception was used for all experiments at 3T. Due to its local energy deposition, less restrictive SAR limits for extremities apply. The amount of receive channels in combination with the coil’s customized geometry enable high SNR and high acceleration factors when using parallel imaging techniques like generalized autocalibrating partial parallel acquisition (GRAPPA) (31). Its maximum transmit B1 field is approximately three times the field which can be achieved by the body coil of the whole-body MR scanner (MAGNETOM Skyrafit, Siemens Healthcare, Erlangen, Germany). This wholebody MR scanner in combination with the described knee coil forms the setup for all 3T experiments. For reference scans on a 1.5T imager, a MAGNETOM Espree (Siemens Healthcare, Erlangen, Germany) with a whole-body transmit coil was used in combination with an eight-channel receive-only knee coil (InVivo, Gainesville, FL). Experiments

following. For exploitation of the increased RF power of the local transmit coil, SINC pulses with a TBP of 4 and an RF bandwidth of 4 kHz are used for both excitation and refocusing for the SEMAC sequence. SINC pulses were used, as they can be implemented easily with flexible TBP; effects due to a potentially lower fidelity of the slice profile as compared to numerically optimized SLR pulses were not observed. The refocusing bandwidth is reduced to 3.3 kHz for ORS when using TSE. These pulses are referred to as the highbandwidth (HiBW) mode. In this work, SEMAC comprises VAT, the thickness of SEMAC partitions equals the nominal slice thickness and partitions are combined after distortion correction using the sum-ofsquares method. Hence, each slice and SEMAC partition in HiBW mode covers a range of off-resonances 4.7 times as high as in conventional mode, i.e., for equal off-resonance coverage, SESr in conventional mode must be selected 4.7 times as high as in HiBW mode. Although the conventional-bandwidth mode could be applied by the body coil, both modes are applied by the local transmit coil for a direct comparison focusing solely on the bandwidth as parameter. Because susceptibility-induced off-resonances due to metal exceed the shift in frequency of fat compared with water, spectral fat suppression must be replaced

Both RF modes were compared experimentally in phantom measurements and in a patient study containing eight subjects. Phantom Measurements A total knee arthroplasty (BPK-S, Peter Brehm, Weisendorf, Germany) with both the tibial and femoral part made of CoCr28Mo6 and an insert of polyethylene are immersed in doped agar (1.25g NiSO4x6H2O and 5g NaCl per liter) with polypropylene structures on both sides (Fig. 1a). All phantom experiments were performed on the 3T setup and the readout bandwidth was set to 685 Hz/Pix. A B0 field map was reconstructed from a SEMAC acquisition covering 660 kHz in steps of 2 kHz with the following scan parameters: 60 slices, slice thickness 1.5 mm, 60 SES, resolution 0.4  0.4 mm2, matrix size 312  384, turbo factor (TF) 9, 36 ms echo time (TE), 8.1 s repetition time (TR). All reconstruction algorithms in this work were implemented in MatLab (MathWorks, Natick, MA). Images of the phantom were acquired using the SEMAC sequence in both the conventional and HiBW mode with the following scan parameters: 60 slices, slice thickness 3 mm, resolution 0.4  0.4 mm2, matrix size 300  384, in-plane parallel imaging acceleration of 3 (GRAPPA), TF ¼ 9, TE ¼ 28 ms, and

Knee Metal Implant Imaging at 3T

1573

Table 1 Implant Types, Acquired Contrasts and Orientations and Applied Methods* Patient #1 #2 #3 #4 #5 #6 #7 #8

Implant type

PD sag

STIR cor

T1 cor

PD sag 1.5T

TKA, CoCr TKA, CoCr TKA, CoCr TKA, CoCr TKA Tibia intramedullary rod, Ti6AL4V Osteosynthesis with plate and screws, Ti Plate osteosynthesis, Ti

S S S S, T – S, T T T

S S S – T T S, T T

S S S S,T S,T T T S, T

– – – S,T – S,T – –

*Each acquisition was performed in conventional and high-bandwidth mode; S ¼ SEMAC ¼ slice encoding metal artifact correction, T ¼ TSE ¼ turbo spin echo, TKA ¼ total knee arthroplasty, PD ¼ proton density, STIR ¼ short tau inversion recovery.

TR ¼ 9.6 s resulted in a total scan time of 2:05 h. The high number of 60 SES per slice was chosen to get reference data with very high off-resonance coverage. For the quantification of through-plane distortion of a selected slice, the maximum signal intensity of all SEMAC partitions being reconstructed to one particular slice is depicted as a function of slice-encoding steps for both RF modes. The maximum signal intensity of each partition is of particular interest, because it allows the evaluation of its impact on the reconstructed image: aliasing or a lack of certain areas of a partition cause signal pile-ups or voids in the final image, whose intensity scale with the signal intensity in this specific area of the original partition. Because the phantom provides homogeneous signal intensity, the graph representing the spatial extent of the HiBW mode is expected to be more restricted to the center than the conventional one, reflecting the ratio of the RF bandwidths. Based on the imaging data with extensive SEMAC encoding, acquisitions with less than 60 SES are simulated by combining partitions contributing to the selected slice appropriately. This is referred to as virtual reconstruction in the following and used for the analysis of the dependence of the level of artifacts on the number of SES. According to the results of the field map, the number of required SES for full offresonance coverage to ensure maximum through-plane correction was determined for the HiBW mode. HiBW and conventional images were reconstructed virtually from this reduced number of SEMAC partitions and image quality was compared. In preparation of an in vivo study and facing the limitations of clinically feasible scan time (less than 9 min for all relevant

contrasts), the maximum number of SES at 3T was set to eight. The level of artifacts for virtual reconstructions from eight SES was assessed for both RF modes. In Vivo Study This study was approved by the institutional review board. All patients signed informed consent. The effects of high-bandwidth RF pulses in both TSE and SEMAC imaging were evaluated for eight patients with metal implants of different types and materials. Five patients had total knee arthroplasties, one had an intramedullary rod of the tibia, and two patients had metal plate osteosyntheses. Different contrasts and orientations were acquired both using the HiBW and the conventional mode (see Table 1). The applied RF pulses are the only difference between both acquisition modes, all other scan parameters were kept identical. Due to scan time limitations, not all patients could be scanned with SEMAC sequences. Expecting stronger artifacts for total knee arthroplasties than for titanium fixtures, SEMAC had priority for those patients. Twenty-five slices of 3 mm thickness were acquired of each knee with an in-plane resolution of 0.4  0.4 mm2 for T1 and proton density (PD) contrasts, and 0.6  0.6 mm2 for the STIR contrast. SEMAC was always applied with eight slice-encoding steps, because larger values would result in unacceptable scan time durations. Further protocol parameters are found in Table 2. Echo time TE may differ up to 3 ms from the given value due to different pulse durations in HiBW and conventional mode. These measurements were performed using the 3T setup. Two

Table 2 Imaging Parameters for Each Contrast and Acquisition Mode

PD sag STIR cor T1 cor PD sag 1.5T

Sequence

TE/TR

TF

aRef

TSE SEMAC TSE SEMAC TSE SEMAC TSE SEMAC

37/4500 36/4500 40/7200 40/7500 16/774 16/800 23/4500 36/4500

9 9 9 10 2 3 9 9

140 135 140 140 140 130 145 145

matrix 346 270 256 205 288 288 269 269

       

384 384 256 256 384 384 384 384

resolution 0.4 0.4 0.6 0.6 0.4 0.4 0.5 0.5

       

0.4 0.4 0.6 0.6 0.4 0.4 0.5 0.5

       

3.0 3.0 3.0 3.0 3.0 3.0 4.0 4.0

PAT

TA

off 3 off 3 2 3 off 3

3:02 7:18 3:38 9:09 2:04 8:10 2:21 7:17

TE ¼ echo time [ms]; TR ¼ repetition time [ms]; TF ¼ turbo factor; aRef ¼ flip angle of refocusing pulse [ ], resolution [mm3]; PAT ¼ parallel acquisition technique applied in-plane; TA ¼ total acquisition time [min].

1574

Bachschmidt et al.

Figure 2. SEMAC images of the phantom in Conv (a–c) and HiBW (d–f) mode reconstructed with varying numbers of sliceencoding steps (SES): (a) and (d) show the images reconstructed from the maximum of 60 SES (exploits maximum memory size of reconstruction system), which cover 625 kHz (a) and 6118 kHz (d); long arrows point to areas, where both polypropylene and polyethylene is more distorted in readout direction in (d) compared with (a); short arrows point to areas of ripples, which differ for different bandwidths; considering results of Figure 1b, virtual reconstruction from a reduced number of 12 SES in HiBW mode (e) is sufficient to cover all occurring off-resonances; corresponding Conv image of 12 SES (b); virtual reconstruction with 8 SES in Conv (c) and HiBW (f) mode, indicating the level of artifacts to be expected in SEMAC measurements of TKA in vivo.

patients had reference scans on the 1.5T imager for both SEMAC and TSE sequences to compare image quality. Correspondent sequence parameters are identical with 3T, except for voxel size (0.5  0.5  4 mm3), readout bandwidth (566 Hz/Pix versus 685 Hz/Pix at 3T) and the bandwidth of the RF pulse (2.7 kHz). The increased RF bandwidth can be achieved without a local transmit coil due to reduced SAR restrictions at 1.5T systems and different hardware specifications. To evaluate the effects of the RF pulse bandwidth in vivo, distortion was rated qualitatively by two radiologists independently on a five-point scale (1 ¼ no artifacts; 2 ¼ hardly visible artifacts; 3 ¼ clearly visible artifacts, but without impairment of diagnostic quality; 4 ¼ substantial artifacts with moderate impairment of diagnostic quality; 5 ¼ severe artifacts and nondiagnostic image) (12).

RESULTS Phantom Measurements Figure 1 depicts a schematic of the phantom setup (a) and the corresponding field map of a specific slice

acquired at 3T (b). The acquisition range of the field map is 660 kHz. Maximum off-resonances are detected close to sharp edges (up to 24 kHz) and between the medial and lateral condyle (18 kHz). Through-plane analysis in Figure 1c shows Smax over the SEMAC partitions. Smax represents the highest signal intensity of all voxels in each singularly reconstructed SEMAC partition contributing to the composed slice. This slice is used for further analysis in Figure 2 and it matches the location of the slice of the field map in Figure 1b. The maxima of both graphs representing the two RF modes are scaled to one. The distribution in conventional mode covers a significantly larger amount of SES than in HiBW mode: selecting the first minor peak to the left of the main peak as point of reference (arrows), 12 SES in conventional mode and 3 SES in HiBW mode are required to cover this peak. This diagram is instructive, because it demonstrates the ability of the HiBW approach to cover the same range of off-resonances with much fewer SEMAC encoding steps. Both full SEMAC acquisitions with 60 sliceencoding steps in conventional and HiBW modes (Figs. 2a,d) cover all prevalent off-resonances as indicated by the field map. Comparing the image quality

2.8 6 1.0 2.0 6 0.0

– – – 3-4 – 3.5 6 0.7 2-2 – – 2.0 6 0.0 2-2 2-2 2-2 2-2 2-2 2.0 6 0.0 – – 2-2 2.0 6 0.0

H

T

4-5 4-5 4-3 4-4 4-4 4.1 6 0.6 – – 2-3 2.5 6 0.7

3.8 6 0.8 2.5 6 1.1

H

– – – 2-4 3-4 3.3 6 1.0 1-1 2-3 2-3 2.0 6 0.9

C

S T1 cor

3.2 6 1.0 2.1 6 0.35

– – – 3-5 4-4 4.0 6 0.8 2-2 2-3 3-4 2.7 6 0.8 2-2 2-2 2-2 – – 2.0 6 0.0 – 3-2 – 2.5 6 0.7

H

C

T

– – – – 3-4 3.5 6 0.7 1-1 3-3 3-2 2.2 6 1.0

2.5 6 1.1

– – – – 5-5 5.0 6 0.0 3-3 4-4 4-4 3.7 6 0.5

4.0 6 0.8

4.4 6 0.9

H C

5-5 5-5 4-5 – – 4.8 6 0.4 – 3-3 – 3.0 6 0.0

C

S STIR cor

1.8 6 0.4

2-2 2-2 2-2 2-2 – 2.0 6 0.0 1-1 – – 1.0 6 0.0

H

T

– – – 3-3 – 3.0 6 0.0 1-1 2-2 2-3 1.8 6 0.8

2.1 6 0.8

– – – 4-5 – 4.5 6 0.7 3-3 2-3 3-4 3.0 6 0.6

3.4 6 0.9

3.4 6 1.3

H C

4-4 4-4 4-4 4-4 – 4.0 6 0.0 1-1 – – 1.0 6 0.0

C

S PD sag

Mean

#1 #2 #3 #4 #5 Mean #6 #7 #8 Mean

T Mode/ patient

Table 3 Distortion In Vivo Graded by Two Radiologists (First-Second) on a Scale From 1 (No Artifacts) to 5 (Severe Artifacts)

S ¼ SEMAC ¼ slice encoding metal artifact correction; T ¼ TSE ¼ turbo spin echo; H ¼ high-bandwidth mode; C ¼ conventional mode; PD ¼ proton density; STIR ¼ short tau inversion recovery; the mean of all patients and subgroups of them (total knee replacements only and titanium inserts only) was calculated for each contrast.

1.5 6 0.6

1575

1.5T

PD sag

S

– – – 2-2 – 2.0 6 0.0 1-1 – – 1.0 6 0.0

Knee Metal Implant Imaging at 3T

among them, plastic grids appear more distorted and blurrier in HiBW mode in vertical direction, which is the direction of frequency encoding. This applies to the polyethylene insert as well (long arrows). Ripple artifacts (short arrows) are prominent in both acquisition modes. The conventional image features more, but less intense pile-ups compared with the HiBW mode. Based on the field map, theoretically 53 SEMAC steps (asymmetric to the central slice, 24 in negative, 28 in positive direction) are required to cover all occurring off-resonances in conventional mode, while only 12 (5 in negative, 6 in positive direction plus the central one) are sufficient in HiBW mode. This virtually reduced SEMAC reconstruction is depicted in Figure 2e and no difference to the fully reconstructed image in Figure 2d is visible. Reducing SES by one would manifest in slightly increased signal void in close proximity to the implant. In contrast, the reconstruction of only 12 SEMAC partitions in conventional mode is dominated by signal voids and aliasing artifacts (Fig. 2b). Signal voids occur close to the implant in areas of strong off-resonances of this specific slice, while folding artifacts are independent of the geometry of the specific slice, but depend on the off-resonances given in other slices. Figures 2c and 2f show the virtual reconstruction of eight SES in conventional and HiBW mode, respectively. In HiBW mode, signal void is slightly increased in areas of maximum positive and negative off-resonances, e.g., the caudal area of the tibial implant, compared with the full coverage in Figure 2e and slight signal pileups between the tibial and femoral condyle reflect signal from other slices. In conventional mode (Fig. 2c), the level of artifacts for 8 SES is significantly increased in contrast to 12 SES. Comparing the acquisitions of both RF modes using eight SES, the shape of the knee implant can be distinguished in HiBW mode, while structures are not identifiable in conventional mode. In Vivo Study The readers’ individual analyses of distortion in vivo are summarized in Table 3. For patients #1 to #3 with TKA, the application of SEMAC with eight sliceencoding steps in the HiBW mode reduces the level of distortion from grade 4.3 6 0.5 in conventional mode to grade 2.0 6 0.0 on average. For patients with titanium fixtures, high-bandwidth TSE sequences improve distortion from grade 3.160.8 in conventional mode to grade 2.0 6 0.8 on average. Please note that on average the distortion is rated less or equal for the group of patients getting a TSE scan. This is due to the fact that SEMAC scans were preferred for those patients having implants with high distortions, according to Table 1. In the following, results of four individual patients with different types of implants are presented in detail and the average rate of both radiologists for distortion is evaluated. Patients With Total Knee Arthroplasty Patient #4 has a total knee arthroplasty of a CoCrMo alloy (Protasul-1). The PD images of the comparison

1576

Bachschmidt et al.

Figure 3. Patient #4 with total knee arthroplasty, proton density weighted (PD) sagittal: TSE Conv 3T (a), TSE HiBW 3T (b), TSE 1.5T (c), SEMAC Conv 3T (d), SEMAC HiBW 3T (e), SEMAC 1.5T (f).

between TSE and SEMAC for both the conventional and HiBW mode and the corresponding 1.5T reference scans are shown in Figure 3. In conventional mode (Figs. 3a,d), the level of distortion for the TSE image (grade 4.560.7) remains nearly unchanged by the application of SEMAC (grade 4.0 6 0.0), due to insufficient off-resonance coverage and folding artifacts. The HiBW mode improves distortions for both TSE (grade 3.0 6 0.0) and SEMAC (grade 2.0 6 0.0) (Figs. 3b,e). Comparing the HiBW mode at 3T with the scans at 1.5T (Figs. 3c,f), it turns out that the level of distortion is very similar, both for TSE (grade 3.5 6 0.7)

and SEMAC (grade 2.0 6 0.0). Correspondent SEMAC images reveal the shape of the implant clearly. Figure 4 shows data of patient #5. The level of distortion for the STIR contrast acquired with the TSE sequence improves from grade 5.0 6 0.0 to grade 3.5 6 0.7, when the HiBW mode is used instead of the conventional mode (Figs. 4a,d). Using equal methods for the acquisition of a T1-weighted contrast, differences between the HiBW (grade 3.5 6 0.7) and conventional mode (grade 4 6 0.0) are less significant. Further reduction of distortion is achieved by SEMAC in HiBW mode only. In that acquisition mode, the ringing

Figure 4. Patient #5 with total knee arthroplasty: short tau inversion recovery (STIR) coronal TSE Conv (a), STIR coronal TSE HiBW (d), T1 coronal TSE Conv (b), T1 coronal TSE HiBW (e), T1 coronal SEMAC Conv (c), T1 coronal SEMAC HiBW (f).

Knee Metal Implant Imaging at 3T

1577

Figure 5. Patient #6 with intramedullary rod, PD sagittal: TSE Conv 3T (a), TSE HiBW 3T (b), TSE 1.5T (c), SEMAC Conv 3T (d), SEMAC HiBW 3T (e), SEMAC 1.5T (f).

artifact in the caudal area of the tibial implant becomes dominant (Fig. 4f). It is not visible in the conventional mode (Fig. 4e) due to overlaying stronger artifacts. Patients With Other Implants The PD results of the comparison between TSE and SEMAC for both the conventional and HiBW modes and the corresponding 1.5T reference scans are shown in Figure 5 for patient #6. The intramedullary

rod in the tibia is made of a titanium alloy. At 3T, both the TSE HiBW mode and SEMAC in any excitation mode minimize distortion, while in TSE conventional mode, distortion was rated as clearly visible, but without impairment of diagnostic quality. For this patient, the TSE sequence in HiBW mode is superior to all other acquisition modes, both for 3T and 1.5T, with respect to total scan time and distortion. Regarding the STIR contrast in coronal orientation of patient #7 (Figs. 6a–d) with a tibial titanium osteosynthesis, distortion is improved incrementally by both the use

Figure 6. Patient #7 with osteosynthesis of the tibia, STIR coronal: TSE Conv (a), TSE HiBW (b), SEMAC Conv (c), SEMAC HiBW (d); Patient #8 with osteosynthesis of the tibia, T1 coronal: TSE Conv (e), TSE HiBW (f), SEMAC Conv (g), SEMAC HiBW (h).

1578

of SEMAC and HiBW mode compared with TSE conventional mode. In HiBW mode, both for TSE and SEMAC, there is substantial shading in the area of convergence of the two top screws. Considering the results of the T1 contrast of patient #8 (Figs. 6e–h) in coronal orientation with a tibial and femoral titanium osteosynthesis and taking the TSE conventional mode as reference, distortion is reduced in all other modes. Residual through-plane artifacts in the femoral area are visible in SEMAC conventional mode (Fig. 6g), while the TSE HiBW mode (Fig. 6f) shows stronger signal pile-ups close to screws. In SEMAC HiBW mode (Fig. 6h), ripple artifacts are intensified.

DISCUSSION Sequence-based methods for artifact compensation comprise ORS, high-bandwidth RF pulses or SEMAC to limit or resolve distortion through-plane. When using SEMAC, the increased acquisition time is a true draw-back in clinical routine due to patient comfort, motion artifacts and economic considerations, despite the use of acceleration techniques like GRAPPA. Although SEMAC can be combined with ORS, which limits the number of required SEMAC encoding steps SESr (32), this is not done in this work due to increased signal voids introduced by this method. Another option to reduce SESr is the use of thinner slices. However, in that case more slices are required to cover the anatomy. Considering the high level of SAR deposition in SEMAC imaging in general, this method will mostly countervail the benefit in scan time gained by reduced SESr. At 3T, off-resonances are increased and RF pulse bandwidth is more limited due to SAR restrictions as compared to 1.5T. Both aspects increase the number of required SESr and hence scan time. Exemplary through-plane analysis of phantom experiments confirms that the proportion of required SEMAC encoding steps for artifact reduction relates inversely to the proportion of accordant RF bandwidths: the coverage of a selected off-resonance featuring high signal intensity requires 12 SES in conventional mode and 3 SES in HiBW mode, approximately reflecting the relation of bandwidths of 4.7. According to theory, high RF bandwidths impair image quality in-plane: if the readout gradient is kept constant, the application of VAT gradients increases blurring in readout direction due to a stronger lowpass filter and the view angle increases from 9.4 in conventional mode to 37.9 in HiBW mode. This enhanced blurring is visible in the phantom evaluation using SEMAC, which comprises VAT, in proximity to plastic structures. Methods reducing blurring introduced by VAT focus on sparse sampling of k-space edges (33,34), but have not been combined with SEMAC yet. Besides, the high-bandwidth SEMAC image is composed of less, but more intense transitions between different SEMAC partitions. These ripple artifacts are caused by rapid spatial variation of B0 and also depend on the extension of the covered off-resonances by the RF pulse and its slice profile.

Bachschmidt et al.

Recently, this was studied in detail (35), but no solutions have been proposed up to now. For plates, marrow nails and screws of titanium, subjective evaluation of in vivo results indicates a similar level of artifact reduction for TSE in HiBW mode and SEMAC in conventional mode, while latter one involves a significantly higher acquisition time. For maximum signal void reduction, SEMAC with total off-resonance coverage is indispensable. However, the margin in signal void reduction compared with TSE HiBW is limited for titanium plates, nails, and screws. A larger range of off-resonances is corrected in SEMAC conventional mode compared with the covered range of off-resonances in TSE HiBW mode. Still, through-plane artifacts can derange the image in SEMAC conventional mode: aliased signal not only from the central partition, but from many slices, contributes to the final reconstruction. Furthermore, SEMAC introduces ripple artifacts, i.e., pileups, in any RF excitation mode. Based on these disadvantages of SEMAC and our results, in patients without total joint replacements, imaging of titanium implants with TSE in HiBW mode is preferred over SEMAC at 3T. The observation of locally increased shading hints to certain adverse effects of the HiBW mode that need further investigation. Potential explanations could be, e.g., electric field induced currents related to the stronger transmit B1 or eddy currents related to the stronger slice select gradients. For patients with TKA, the use of HiBW RF pulses in combination with TSE is not sufficient to tackle distortion at 3T. The use of SEMAC is suggested. This technique combined with HiBW RF pulses allows the reduction of the number of required SEMAC encoding steps and is hence very efficient in achieving clinically more acceptable scan times. Apart from economic considerations, shorter scan times help to avoid motion artifacts and provide increased patient comfort. In clinical experiments, eight SES per slice appeared to be a good trade-off between off-resonance coverage and an acceptable scan time of 7 to 9 min per acquisition. However, because the extension of distortions may vary significantly based on size and material of the implant, the optimal number of SES would have to be determined on an individual basis. A potential solution would be the implementation of a scout scan to calculate the number of required SEMAC partitions in advance of the actual imaging scan. On the basis of this information, asymmetric off-resonance coverage can be realized and saves additional scan time (36). Minimizing the number of SES is beneficial regarding acquisition time, but Fourier transform of only few phase-encoding steps can result in a loss of image quality due to through-plane ghosting and blurring. Using SEMAC, blurring can be affected additionally by both overlapping spectral profiles (37) and effects induced by VAT, which are linked to the RF modes. The exclusive effect of high RF bandwidths on blurring and in-plane distortion in SEMAC, VAT, and TSE may be subject to future studies. A limitation of this work is the small sample size. Data of eight patients does not allow statistical implications, but substantiates the clinical applicability of

Knee Metal Implant Imaging at 3T

the results of phantom measurements. For quantitative evaluation of distortion, which is an important parameter of image quality, knowledge about the exact dimensions of the implants would be required. In the literature, only few authors target the effects of increased susceptibility artifacts at 3T induced by metal implants. MRI close to metal implants at 3T has been compared with 1.5T with respect to susceptibility artifacts in-plane (21), heating (38), and B1 effects (39). Publications focusing on 3T imaging are limited to metallic dental materials (30) and scan time reduction for SEMAC (40). Earlier, MRI at 3T was disadvised in this context (10). To our knowledge, so far throughplane artifacts have not been studied with regard to different RF pulse bandwidths, neither at 1.5T nor at 3T. At 1.5T, high RF bandwidths are less advantageous due to the increased in-plane distortion and blurring, and finally because the level of total distortion is lower. In this study, we used a local transmit knee coil to achieve increased bandwidth at 3T, circumventing the more limiting RF-properties of a whole-body transmit coil. The same could be applied to other local transmit coils, such as for hand, elbow, ankle, etc. In conclusion, our results show that the application of high-bandwidth RF pulses by local transmit coils can mitigate the two main limiting factors that hamper implant imaging at 3T: SAR limits can be used to full capacity to reduce susceptibility artifacts. Although high-bandwidth RF pulses may increase blurring, total acquisition time can be reduced significantly: throughplane distortion is decreased such that either SEMAC can be applied with fewer encoding steps, or even SEMAC encoding is not required any more, e.g., for small pieces of metal or material with low susceptibility like titanium. Adaption of the bandwidth also includes inversion pulses for fat-suppressed STIR imaging. Thus, all relevant contrasts can be obtained at a level of artifacts comparable to 1.5T and benefit from improved image quality at 3T without a penalty in scan time. REFERENCES 1. Hasenboehler E, Rikli D, Babst R. Locking compression plate with minimally invasive plate osteosynthesis in diaphyseal and distal tibial fracture: a retrospective study of 32 patients. Injury 2007;38:365–370. 2. Kurtz S, Mowat F, Ong K, Chan N, Lau E, Halpern M. Prevalence of primary and revision total hip and knee arthroplasty in the united states from 1990 through 2002. J Bone Joint Surg Am 2005;87:1487–1497. 3. Kurtz S, Ong K, Lau E, Mowat F, Halpern M. Projections of primary and revision hip and knee arthroplasty in the United States from 2005 to 2030. J Bone Joint Surg Am 2007;89: 780–785. 4. Chen CA, Chen W, Goodman SB, et al. New MR imaging methods for metallic implants in the knee: artifact correction and clinical impact. J Magn Reson Imaging 2011;33:1121–1127. 5. Wong S, Steinbach L, Zhao J, Stehling C, Ma BC, Link TM. Comparative study of imaging at 3.0 T versus 1.5 T of the knee. Skeletal Radiol 2009;38:761–769. 6. Shapiro L, Staroswiecki E, Gold G. MRI of the knee: optimizing 3T imaging. Semin Roentgenol 2010;45:238–249. 7. Shapiro L, Harish M, Hargreaves B, Staroswiecki E, Gold G. Advances in musculoskeletal MRI: Technical considerations. J Magn Reson Imaging 2012;36:775–787. 8. Kuo R, Panchal M, Tanenbaum L, Crues JV III. 3.0 Tesla imaging of the musculoskeletal system. J Magn Reson Imaging 2007;25: 245–261.

1579 9. Eustace S, Goldberg R, Williamson D, et al. MR imaging of soft tissues adjacent to orthopaedic hardware: techniques to minimize susceptibility artefact. Clin Radiol 1997;52:589–594. 10. Lee M-J, Kim S, Lee S-A, et al. Overcoming artifacts from metallic orthopedic implants at high-field-strength MR imaging and multidetector CT. Radiographics 2007;27:791–803. 11. Sofka CM, Potter HG, Figgie M, Laskin R. Magnetic resonance imaging of total knee arthroplasty. Clin Orthop Relat Res 2003: 129–135. 12. Sutter R, Hodek R, Fucentese SF, Nittka M, Pfirrmann, CW. Total knee arthroplasty MRI featuring slice-encoding for metal artifact correction: reduction of artifacts for STIR and proton density– weighted sequences. AJR Am J Roentgenol 2013;201:1315–1324. 13. Naraghi AM, White LM. Magnetic resonance imaging of joint replacements. Semin Musculoskelet Radiol 2006;10:98–106. 14. Vandevenne JE, Vanhoenacker FM, Parizel PM, Butts Pauly K, Lang RK. Reduction of metal artefacts in musculoskeletal MR imaging. JBR-BTR 2007;90:345–349. 15. Guermazi A, Miaux Y, Zaim S, Peterfy CG, White D, Genant HK. Metallic artefacts in MR imaging: effects of main field orientation and strength. Clin Radiol 2003;58:322–328. 16. Butts K, Pauly JM, Gold GE. Reduction of blurring in view angle tilting MRI. Magn Reson Med 2005;53:418–424. 17. Jung KJ, Moon CH. Effect of slice angle on inhomogeneity artifact and its correction in slice selective MR imaging. Concepts Magn Reson Part A 2009;34A:238–248. 18. T€ o rm€ anen 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;65:805–811. 19. Viano AM, Gronemeyer SA, Haliloglu M, Hoffer FA. Improved MR imaging for patients with metallic implants. Magn Reson Imaging 2000;18:287–295. 20. White LM, Kim JK, Mehta M, et al. Complications of total hip arthroplasty: MR imaging–initial experience. Radiology 2000;215: 254–262. 21. Farrelly C, Davarpanah A, Brennan S, Sampson M, Eustace SJ. Imaging of soft tissues adjacent to orthopedic hardware: comparison of 3-T and 1.5-T MRI. AJR Am J Roentgenol 2010;194:W60– W64. 22. Lu W, Pauly KB, Gold GE, Pauly J, Hargreaves B. SEMAC: slice encoding for metal artifact correction in MRI. Magn Reson Med 2009;62:66–76. 23. Koch KM, Lorbiecki JE, Hinks RS, King KF. A multispectral three-dimensional acquisition technique for imaging near metal implants. Magn Reson Med 2009;61:381–390. 24. Bos C, den Harder CJ, van Yperen G. MR Imaging near orthopedic implants with artifact reduction using view-angle tilting and off-resonance suppression. In: Proceedings of the 18th Annual Meeting of ISMRM, Stockholm, 2010. (abstract 129). 25. Hargreaves BA, Worters PW, Pauly KB, Pauly JM, Koch KM, Gold GE. Metal-induced artifacts in MRI. AJR Am J Roentgenol 2011; 197:547–555. 26. Olsen RV, Munk PL, Lee MJ, et al. Metal artifact reduction sequence: early clinical applications. Radiographics 2000;20: 699–712. 27. Ulbrich EJ, Sutter R, Aguiar RF, Nittka M, Pfirrmann CW. STIR sequence with increased receiver bandwidth of the inversion pulse for reduction of metallic artifacts. AJR Am J Roentgenol 2012;199:W735–W742. 28. Hargreaves BA, Chen W, Lu W, et al. Accelerated slice-encoding for metal artifact correction. J Magn Reson Imaging 2010;31: 987–996. 29. Sutter R, Ulbrich EJ, Jellus V, Nittka M, Pfirrmann CW. Reduction of metal artifacts in patients with total hip arthroplasty with slice-encoding metal artifact correction and view-angle tilting MR imaging. Radiology 2012;265:204–214. 30. Zho S, Kim M, Lee K, Kim D. Artifact reduction from metallic dental materials in T1-weighted spin-echo imaging at 3.0 Tesla. J Magn Reson Imaging 2013;37:471–478. 31. Griswold MA, Jakob PM, Heidemann RM, et al. Generalized autocalibrating partially parallel acquisitions (GRAPPA). Magn Reson Med 2002;47:1202–1210. 32. den Harder CJ, Blume UA, Bos C. MR Imaging near orthopedic implants using slice-encoding for metal artifact correction and off-resonance suppression. In: Proceedings of the 19th Annual Meeting of ISMRM, Montreal, 2011. (abstract 3170).

1580 33. Li G, Zhang B, Guo H, Zhang Q. BLADE-VAT for geometric distortion correction. In: Proceedings of 16th Annual Meeting of ISMRM, Toronto, 2008. (abstract 1355). 34. Zho S, Kim D. Spiral imaging with view angle tilting for application to metal artifact correction. In: Proceedings of the 19th Annual Meeting of ISMRM, Montreal, 2011. (abstract 4618). 35. Koch KM, King KF, Carl M, Hargreaves BA. Imaging near metal: the impact of extreme static local field gradients on frequency encoding processes. Magn Reson Med 2014;71:2024–2034. 36. Hargreaves BA, Gold GE, Pauly JM, Pauly KB. Adaptive slice encoding for metal artifact correction. In: Proceedings of 18th Annual Meeting of ISMRM, Stockholm, 2010. (abstract 3083).

Bachschmidt et al. 37. Koch KM, Brau AC, Chen W, et al. Imaging near metal with a MAVRIC-SEMAC hybrid. Magn Reson Med 2011;65:71–82. 38. Liu Y, Chen J, Shellock FG, Kainz W. Computational and experimental studies of an orthopedic implant: MRI-related heating at 1.5-T/64-MHz and 3-T/128-MHz. J Magn Reson Imaging 2013; 37:491–497. 39. Graf H, Lauer UA, Berger A, Schick F. RF artifacts caused by metallic implants or instruments which get more prominent at 3 T: an in vitro study. Magn Reson Imaging 2005;23:493–499. 40. Worters PW, Sung K, Stevens KJ, Koch KM, Hargreaves BA. Compressed-sensing multispectral imaging of the postoperative spine. J Magn Reson Imaging 2013;37:243–248.