Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
J Magn Reson Imaging. Author manuscript; available in PMC 2011 April 1.
Published in final edited form as:
PMCID: PMC2894155

Accelerated Slice Encoding for Metal Artifact Correction



To demonstrate accelerated imaging with artifact reduction near metallic implants and different contrast mechanisms.

Materials and Methods

Slice-encoding for metal artifact correction (SEMAC) is a modified spin echo sequence that uses view-angle tilting and slice-direction phase encoding to correct both in-plane and through-plane artifacts. Standard spin echo trains and short-TI inversion recovery (STIR) allow efficient PD-weighted imaging with optional fat suppression. A completely linear reconstruction allows incorporation of parallel imaging and partial Fourier imaging. The SNR effects of all reconstructions were quantified in one subject. 10 subjects with different metallic implants were scanned using SEMAC protocols, all with scan times below 11 minutes, as well as with standard spin echo methods.


The SNR using standard acceleration techniques is unaffected by the linear SEMAC reconstruction. In all cases with implants, accelerated SEMAC significantly reduced artifacts compared with standard imaging techniques, with no additional artifacts from acceleration techniques. The use of different contrast mechanisms allowed differentiation of fluid from other structures in several subjects.


SEMAC imaging can be combined with standard echo-train imaging, parallel imaging, partial-Fourier imaging and inversion recovery techniques to offer flexible image contrast with a dramatic reduction of metal-induced artifacts in scan times under 11 minutes.

Keywords: SEMAC, Metal Artifact, Distortion, Slice-Encoding, Artifact Reduction


Numerous orthopedic disorders are treated using metallic implants including joint replacements and spinal fixation hardware. Magnetic resonance imaging (MRI) provides excellent soft-tissue contrast, and would thus be an ideal modality for post-surgical examination of these implants. However, metal-induced artifacts, such as distortion, signal voids and “pile-up” artifact, often render MRI non-diagnostic. These artifacts exist along both in-plane and through-slice directions, resulting from severe static field inhomogeneities near metal due to the susceptibility difference between tissues and metal (1).

Many methods have been explored for the reduction of metal-induced artifacts in MRI. The use of spin echoes recovers much of the signal loss and is common in orthopedic imaging. View-angle tilting (VAT) (2) is a simple method to suppress in-plane distortion, but does not correct through-plane distortion. Numerous methods that estimate a distortion map have been proposed, but these generally assign a distortion field to each pixel, which is inadequate for very rapid spatial variations of the static magnetic field (3,4). In most cases, maximizing the slice-selection and readout bandwidths will reduce artifacts, but at a cost of signal-to-noise ratio (5). However, since the range of frequencies around metallic implants typically far exceeds the excitation pulse bandwidth, it is simply not possible to both correct distortion and avoid signal loss without increasing the amount of data acquired.

Two recently proposed spin-echo methods use additional encoding to almost completely eliminate both signal loss and distortion. Multi-acquisition variable-resonance image combination (MAVRIC) acquires non-spatially-selective 3D spin echo acquisitions at multiple center frequencies (6). Because the bandwidth of each acquisition is limited, the in-plane distortion is typically limited to the size of a pixel, and the combination of all acquisitions almost completely corrects signal loss by covering a wide bandwidth. Slice encoding for metal artifact correction (SEMAC) selectively excites 2D slices and uses 3D phase encoding for each slice to resolve through-plane distortion (7). In-plane distortions can be corrected either using a shear correction or by using VAT. Comparing the two methods, MAVRIC is a 3D technique that divides data based on resonance frequency, while SEMAC divides data based on standard slices (selected using a gradient), corresponding to a tilted region in the slice–frequency plane. Although SEMAC uses a series of 2D excitations, the image acquisition is 3D.

Although both MAVRIC and SEMAC provide excellent correction of metal artifacts, both incur significantly prolonged scan times (around a factor of 15–25). Practical scan times for orthopedic protocols are limited to 5–10 minutes, primarily by patient comfort and motion considerations. In this work, we accelerate the SEMAC acquisition by incorporating many standard fast-imaging approaches such as echo-train imaging (8), parallel imaging (911) and partial Fourier imaging (12). In addition to reducing scan times, these methods also allow flexible contrast mechanisms including proton-density (PD) weighting, T1-weighting, and fat suppression using short-TI inversion recovery (STIR) (13). First we experimentally quantify the SNR reductions in SEMAC parallel imaging and partial Fourier imaging compared with standard imaging. Next, we demonstrate these SEMAC acquisitions in phantoms and human subjects, and provide examples of practical clinical imaging near metallic implants.


We review SEMAC imaging, including incorporation of echo-train imaging, fat suppression, parallel imaging and partial Fourier imaging, before demonstrating accelerated SEMAC.

SEMAC Overview

The goal of SEMAC imaging is to completely correct the signal loss and distortion artifacts resulting from metal-induced static magnetic field inhomogeneities (7). Signal loss due to intravoxel dephasing is addressed by using spin echoes (14). In standard 2D Cartesian MRI, a resonant frequency shift Δf results in spatial shifts in the slice-encode and readout directions of Δf/BWRF slices and Δf/BWpix pixels respectively, where BWRF is the excitation bandwidth and BWpix is the bandwidth per pixel. Near metal, there is a large range of resonant frequencies that results in image distortion, signal loss, and signal pile-up artifacts in both the slice-selection and readout directions (15). In the phase-encode direction, however, there are no distortions, since phase differences induced by phase-encode gradients are independent of field inhomogeneities.

To correct through-plane distortion, SEMAC uses the approach shown in Fig. 1. The resonant frequency shifts, Δf are assumed to be in the range |Δf|≤Δfmax. A finite-width slice of thickness Δs is excited, using an RF pulse of bandwidth BWRF. When distortions are considered, the slice may now excite a range ±ΔfmaxΔs/BWRF. To resolve this profile, we add additional phase encoding in the slice direction spanning a FOVz, which we will refer to as SEMAC phase encoding (Fig. 1). After acquisition and a Fourier transform in the SEMAC phase-encode (slice or z) direction, we have resolved the actual z signal locations, which we will refer to as sections. The spatially resolved sections from all excited slices are then added to form an image with no through-plane distortion. Since these steps are linear, the Fourier transforms in x and y can be applied at any stage of the reconstruction.

Figure 1
SEMAC acquisition and reconstruction. (a) The desired volume and number of sections are prescribed. (b) An arbitrary number of excited slices cover the prescribed volume. For each slice (example highlighted), the phase-encoded FOVz includes enough sections ...

Two key points about SEMAC through-slice correction should be emphasized: (1) Phase encoding, rather than slice selection, completely resolves the position in the slice direction, so the slice width does not affect the final through-slice resolution. (2) The extent of distortions requires a limited amount of additional phase encoding (i.e. a range of ±ΔfmaxΔs/BWRF).

To correct in-plane (readout-direction) distortions, SEMAC uses view-angle tilting (VAT), where the slice-select gradient is replayed during readout, to exactly cancel in-plane shifts with slice-direction shifts (Fig. 2) (16). An alternative view of VAT is that by playing the same slice-select gradient during the readout, the frequency of all spins in the excited slice is kept within the RF excitation bandwidth, which is low enough to avoid in-plane distortion greater than the voxel tilt (5). VAT has been used in numerous studies for reducing metal artifacts around prostheses (5, 1720).

Figure 2
SEMAC pulse sequence. Standard RF waveforms are used, with centered phase-ramps to shift the slice location without altering relative slice-to-slice phase. A VAT gradient is played during the readout to correct in-plane distortions, and SEMAC phase encoding ...

Pulse Sequence

The SEMAC pulse sequence is implemented with a standard Carr-Purcell-Meiboom-Gill (CPMG) echo train as shown in Fig. 2, where the relative phases of refocusing pulses differ by 90° from the phase of the initial 90° excitation pulse, and the refocusing flip angle is often decreased from 180° to reduce RF power. The echo train can provide T1-weighted, T2-weighted or PD-weighted contrasts as with standard spin-echo train (RARE, FSE or TSE) imaging (8).

The pulse sequence uses view-angle tilting gradients (Fig. 2) that replay the slice-selection gradient during all readouts (16), and the acquired echo must therefore be demodulated at the slice-selection frequency. Note that the duration of the readout is matched to that of the main lobe of the excitation to avoid modulation-induced blurring in VAT (21).

To resolve actual slice location, phase encoding in the slice direction is included. Phase encoding often introduces small amounts of linear phase, depending on which lines are sampled. In our implementation it is important to not introduce any phase variation across the resolved FOVz, so that signals from different slices add coherently. A simple approach is to sample an odd number (N) of symmetric phase encode lines that include kz=0 and setting an extra line to 0. Subsequently, an N+1-point discrete Fourier transform is applied along the kz direction such that there is no phase variation between resolved spins. In this work we use N=15.

It is also important to avoid phase variation between excited slices. Although other options exist, we use phase-ramp waveforms to shift the slice position. Identical slopes are used on the 90° pulses, refocusing pulses, and during readout, with zero-phase at the pulse center (Fig. 2). Typically the relative delay of these ramps compared to gradients and data acquisition must be within ±4μsec. This was easily achieved in a tuning stage, where we used a simple 2D multislice acquisition (i.e. no z phase encoding) in a phantom with a quadrature birdcage coil and tuned the delays to minimize phase variation between reconstructed slices.

We used a straightforward approach to stabilize the phase in CPMG echo trains (22). The phase of the first refocusing pulse is modified by a value [var phi], which varies with slice location. To determine [var phi] for each slice, we use a pre-scan, whereby the echo train is repeated with different values of [var phi] to acquire an FID. The central points of the FID are added over the echo train, and the[var phi] that produces the greatest magnitude is used during acquisition. In addition, the amplitude of the first refocusing pulse is adjusted to increase the overall echo train amplitude (23).

An optional inversion pulse is used for short-TI inversion recovery (STIR) fat suppression (13). We use an adiabatic hyperbolic secant inversion pulse (24) with 2 kHz bandwidth and 8.6 ms duration. The inversion pulse is played with a slice-selection gradient that has identical amplitude to that of the excitation and refocusing pulses, to ensure that any displacement or distortion of the inverted slice due to frequency shifts is the same as that of the excited and refocused slice. STIR inversions are interleaved such that at least one slice is imaged during the inversion time (TI) of another slice, to improve sequence efficiency.

Standard k-space acquisitions and ordering are adopted for fast imaging techniques. Partial Fourier and parallel imaging are performed along the ky direction, which typically has many more phase encoding steps than the number of SEMAC (kz) phase encodes. Standard interleaving of excited slices is applied, with the SEMAC phase encoding as the outer k-space encoding loop, and ky phase encoding as the inner loop, so that any echo-train artifacts do not affect metal artifact reduction.


The combination of acquisition and reconstruction, shown in Fig. 1, is achieved with a simple linear “pre-processing” stage that results in a distortion-corrected 3D volume, which is then passed to completely standard reconstruction algorithms.

The pre-processing reconstruction (1) Fourier transforms data along kz to resolve actual z position, and (2) adds the resolved signal to a distortion-corrected volume in the correct place. To ensure coherent addition of the signals, we carefully avoid phase variations between excited slices during acquisition, and between phase-encoded sections during both acquisition and reconstruction, as described in the previous section. As the received SEMAC phase-encoded volumes are not phase-modulated to shift the field of view, some wrap along the kz direction in the volumes occurs (Fig. 1b–c). To address this, the recon simply uses modulo-arithmetic to select the appropriate phase-encoded sections and add them to the output volume (Fig. 1d). After pre-processing, the output volume becomes an undistorted 3D acquisition in kxkyz space.

After pre-processing, we apply autocalibrating reconstruction for Cartesian imaging (ARC) to fill in missing data along the ky direction. ARC is a data-driven parallel imaging approach that generates reconstruction weights to efficiently fill in missing data using correlation between coils (25,26). Other parallel imaging methods, such as GRAPPA [11] or mSENSE (27), could be used at this stage.

Partial Fourier reconstruction is applied on a coil-by-coil basis using a homodyne reconstruction, which includes the in-plane Fourier transformation [12]. Despite severe static magnetic field variations, partial Fourier reconstruction is possible because the combination of spin echoes, sequence implementation, and reconstruction steps avoids both in-plane and through-plane phase variations. Data from multiple coils are combined using a standard square-root of sum-of-squares approach.

Parameter Selection

In SEMAC, the scan parameters must be carefully chosen to avoid artifacts. Furthermore, the choices of scan parameters are more interrelated than in normal pulse sequences (7). To minimize the spatial slice distortion, we first maximize the bandwidth of the 90° and 180° RF pulses, which is limited by maximum RF amplitude. (BWRF = 2 kHz in this implementation.) The SEMAC (z) phase encode resolution is set to the desired resolution in that direction, and currently equals the slice thickness (Δs). Assuming a range of frequencies of ±Δfmax=±12 kHz, a reasonable number (typically N =15) of z phase encodes is chosen to provide an adequate FOVz. Finally the readout bandwidth is typically maximized such that SNR is acceptable (7).

Experimental Methods

This section describes experiments in phantoms and volunteers to illustrate the combination of SEMAC with different acceleration and contrast techniques. Unless stated otherwise, our SEMAC scans used 15 SEMAC (z) phase-encodes to resolve slice profile, a slice-selection bandwidth of 2.0 kHz, a readout bandwidth of ±125 kHz, and VAT gradients as described above. We have scanned 15 human subjects following informed consent according to the investigational review board at our institution. Other scan parameters vary with the subject and the body part, and we present specific examples below. All scans were done using General Electric 1.5 T HDx scanners (GE Healthcare, Waukesha WI) with a maximum gradient amplitude of 40 mT/m and maximum slew rates of 150 mT/m/ms. All coils used were standard GE product coils.

To illustrate the combination of PD-weighted SEMAC with partial Fourier and parallel imaging, we scanned a gel phantom containing a shoulder implant (titanium shaft with cobalt-chromium head) using an 8-channel head coil at 1.5T with a PD-weighted SEMAC sequence with TR/TE = 3400/12 ms, ETL=8, 24 slices (2.5 mm thick), 256×144 matrix over a 20×15 cm2 FOV for a 16:26 scan time. 15 SEMAC phase encodes were used to resolve distortions. SEMAC was repeated with 2× acceleration using ARC, and the combination of 2× ARC with 60% partial Fourier (ky) imaging, resulting in scan times of 10:05 and 7:22. Finally, a spin echo sequence with the same parameters, but without VAT, was included for comparison.

To demonstrate the SNR impact of SEMAC and different acceleration methods, we performed quantitative SNR measurements beginning with a fully-sampled T1-weighted SEMAC acquisition in a voluneer with small screws following anterior cruciate ligament reconstruction. The acquisition used an 8-channel knee coil at 1.5T with TR/TE = 500/9 ms, 16 slices (5mm thick), 256×128 matrix over 14×14 cm2 FOV, with 7 SEMAC phase encodes, in 7:36. 112 lines of data were also acquired with no excitation, and used to compute the noise covariance matrix (28). The data were then sub-sampled in three ways: (1) retaining only the central kz plane, enabling a standard 2D multislice reconstruction, (2) by subsampling outer k-space by 2× and 3× to emulate a parallel imaging acquisition, and (3) by discarding half of k-space to emulate a partial Fourier acquisition. In all parallel imaging and partial Fourier cases, 24 central ky lines were retained for calibration. Using all combinations of these 3 categories of subsampling produced 12 datasets. SNR of each combination was measured using the “pseudo multiple replica” method (28). Using the noise covariance matrix, synthesized noise was added to each of the 12 datasets, before reconstructing images with the appropriate combination of SEMAC, parallel imaging (ARC) and partial Fourier reconstructions. The noise addition and reconstruction was repeated 30 times for each dataset, and then mean and standard deviation were measured at each pixel over the 30 repetitions for each of the 12 datasets. SNR at each pixel was then measured as the ratio of mean to standard deviation.

To compare metal artifact correction at different acceleration rates, we scanned a volunteer with stainless steel screws in his knee, and compared the reconstruction with the full acquisition to those where data were discarded followed by use of ARC with 2× and 3× acceleration. We used an 8-channel transmit/receive knee coil, 32 slices (3mm thick), TR/TE = 540/11ms, ETL=1, and 256×128 matrix over a 16cm FOV. The effective scan times were 18:48, 11:06 and 8:29.

In the same volunteer, we also performed PD-weighted spin echo and SEMAC scans with the same matrix size (256×192), FOV and slice thickness. The standard spin echo scan used TR/TE=2000/20 ms and ETL=20 for a 2:01 scan time, while SEMAC used TR/TE=4000/11 ms, ETL=8 and 2× ARC acceleration, with a full ky acquisition for a 10:48 scan time. Both sequences used a ±125 kHz receive bandwidth.

To demonstrate different contrast with SEMAC, we scanned a volunteer with a total knee arthroplasty with the same 8-channel knee coil. We used standard spin echo and both PD- and T1-weighted SEMAC, with 32 slices, each 4 mm thick over a 16×16 cm2 FOV. For PD weighting we used TR/TE = 3450/11 ms with an ETL of 8. For T1 weighting we used TR/TE = 580/11 ms. Both scans used a receive bandwidth of ±125 kHz, 15 SEMAC phase encodes, 60% partial ky acquisition and a 2× ARC acceleration.

As a further example of varied contrast with SEMAC, we have scanned several patients with hip replacements using standard T1-weighted fast spin echo (FSE), standard STIR, and PD-weighted and STIR-PD SEMAC. We show one example using an 8-channel cardiac phased array coil and coronal acquisition with 24 slices, 5mm thick. Both SEMAC scans used a 40×40 cm2 FOV and 256×192 matrix with 60% ky acquisition and 2× ARC acceleration, while standard scans used a 34×34 cm2 FOV and 512×192 matrix. Standard T1-Weighted FSE used TR/TE = 500/7 ms and ETL=4 for a 3:16 scan time. Standard STIR scans used TR/TE/TI = 4700/54/150 ms and ETL=8 for a 5:40 scan time. PD-weighted SEMAC used TR/TE=2100/10 ms and ETL=8 for a scan time of 5:06. STIR SEMAC used 24 slices, TR/TE/TI=4700/54/150 ms and ETL=8 for a scan time of 6:18.


In all scans using accelerated SEMAC, the distortion and signal loss near metallic implants was markedly reduced compared with spin echo and FSE scans. Figure 3 shows phantom image comparisons between spin-echo and fully sampled SEMAC, SEMAC with 2× ARC acceleration, and SEMAC with 2× ARC acceleration plus 60% partial Fourier (ky) acquisition. The correction of distortions is the same for all SEMAC methods, allowing visualization of the area around the neck of the implant device. As is standard, the SNR decreases both with ARC acceleration and with partial Fourier acquisition due to the reduced total acquisition window.

Figure 3
Fully sampled SEMAC image of the neck of a shoulder prosthesis (a), and images comparing the highlighted region with different techniques (b–e) and reformatted images in an orthogonal plane (f–i). Fully sampled SEMAC, 16:26 (b,f), 2× ...

Figure 4 shows the results of SNR measurement comparing SEMAC reconstruction to a standard multislice scan for 1×, 2× and 3× parallel imaging, both with and without partial Fourier imaging. The SNR values shown are the average over a large central ROI in the image including approximately 250 cm3, (83,000 voxels) over 6 slices to give an overall SNR comparision. Most importantly, the SNR in SEMAC acquisitions is identical to standard 2D multislice imaging with the same parameters in all combinations, as predicted in ref. (7). Full k-space acquistions show SNR loss due to both scan time reduction and “g-factor” noise amplification, with the latter becoming much more significant with 3× acceleration. Finally, the cost of 2× and 3× acceleration increases when using partial Fourier imaging. This is mostly due to the fact that higher-frequency noise is amplified by parallel imaging, then further amplified by the partial Fourier reconstruction.

Figure 4
Measured SNR using SEMAC (red) and standard 2D multi-slice (blue) acquisitions and reconstructions, for 1×, 2× and 3× outer k-space sub-sampling, and with full and 60% ky acquisitions. Hollow boxes show the SNRs normalized so that ...

A similar comparison of different ARC acceleration factors in the knee is shown in Fig. 5. Again, all SEMAC images show very good depiction of the tissue near metal. Some residual artifact is present, but neither the residual artifact nor the extent of correction differs with acceleration factor, suggesting that parallel imaging with SEMAC is limited ultimately by SNR.

Figure 5
Spin echo and SEMAC images in a volunteer with stainless steel screws below the knee. (a) Standard SE acquisition (2:20) (b) Full SEMAC acquisition (18:48). (c) 2× ARC accelerated SEMAC (11:06). (d) 3× ARC accelerated SEMAC (8:29). Accelerated ...

Figure 6 shows a SEMAC scan with 2× ARC and 60% ky acquisition in the same subject with stainless steel screws. The 3-axis views show the extent of distortion correction compared with a typical spin echo acquisition, including a view along one of the screws. Comparing the sagittal image with Fig. 5 illustrates the improved spatial resolution enabled by use of partial Fourier acquisition and reconstruction.

Figure 6
PD-weighted Knee Images with ARC and a higher matrix: 256 × 128, 32 slices, full k-space 9-minute SEMAC scan with 2× ARC acceleration (a–c) PD-weighted Spin Echo, (d–f) PD-weighted SEMAC. Images are acquired in the sagittal ...

An important potential use of SEMAC is the assessment of tissues surrounding an orthopedic implant such as a total knee arthroplasty. Figure 7 shows a spin echo image compared with T1-weighted and PD-weighted SEMAC, both with 2× ARC and 60% ky acquisition. The interface between the femur and the implant is well visualized in both SEMAC images. Furthermore, the area posterior to the implant can be identified as a probable fluid region, since it is hypointense on T1-weighted SEMAC compared with PD-weighted SEMAC.

Figure 7
Sagittal images in a volunteer with a total knee arthroplasty comparing PD-weighted FSE to SEMAC with 2×ARC acceleration and 60% partial Fourier acquisition. (a) PD-weighted FSE, 2.5 min, (b) T1-weighted SEMAC, 12 min and (c) PD-weighted SEMAC, ...

Figure 8 shows a comparison of standard T1-weighted FSE with T1-weighted, PD-weighted and STIR PD-weighted SEMAC images in a patient with bilateral hip replacements. Signal loss and distortion seen on the standard images is again removed on SEMAC images. In this clinical example, the visualization of fluid posterior to the joint replacement, seen on both SEMAC PD-weighted and SEMAC STIR images, is correlated with reported pain. This scan avoided the need for fine needle aspiration. The reformatted T1-weighted images show greatly improved depiction of the hip implants with SEMAC compared to FSE.

Figure 8
Plain film X-ray (a), conventional T1 FSE (b), conventional STIR (c), proton-density SEMAC (d) and STIR SEMAC (e) images in a subject with a total hip replacment. Susceptibility artifacts cause “pile-up” artifacts (dashed arrows) and inaccurate ...


In our studies of imaging near metallic implants, accelerated SEMAC consistently resulted in dramatic reduction of distortion compared with standard spin-echo and fast spin-echo methods. Comparing fully sampled SEMAC to accelerated SEMAC resulted in the expected SNR loss, but no change in ability to correct distortions.

SEMAC corrects slice-direction distortions by resolving the excited slice profiles using phase encoding. Since separate slices are excited, VAT can be used to correct in-plane distortions. The combination of data from different excited slices, shown in Fig. 1, is a simple complex-sum. By performing the combining stage first, a distortion-free dataset is obtained, which can then be reconstructed using standard parallel imaging acceleration and partial Fourier methods. The acceleration methods offer a way to perform the entire SEMAC acquisition in a reasonable scan time - typically under 10 minutes for many clinical imaging examples.

Reconstruction Order

The linearity of the reconstruction allows interchanging of the order of combination of data from excited SEMAC slices, Fourier transforms and parallel imaging reconstruction. Although we have presented a simple and fast reconstruction process, other options exist. For example, once parallel imaging weights and/or the partial Fourier phase reference have been determined, data synthesis or homodyne demodulation could be applied earlier in the reconstruction. This may be important when image-domain denoising and data-combining steps are to be included (6,29).


As demonstrated, SEMAC images have the same SNR as a single spin echo image without SEMAC encoding (7). Although SEMAC phase encoding with 15 steps leads to about a factor of 4 increase in SNR in the source images, there is a subsequent loss in SNR when the data from different excited slices are combined, due to the fact that many excited slices contain no signal at many locations. While different approaches of combining data from different SEMAC slices are being explored (29), we have used a simple complex-addition here to enable easy addition of acceleration methods. As we have shown, the effects of acceleration techniques are the same for SEMAC as for standard imaging.

2D Parallel Imaging

Further acceleration of SEMAC could be obtained by using parallel imaging both in-plane and through-plane, although the latter may be limited by the small FOV in the slice direction. The method presented does not offer through-plane parallel imaging, since the slice information is resolved prior to parallel imaging reconstruction. However, an alternative approach could be to combine only the central calibration data from SEMAC slices to perform parallel imaging calibration. Then the weights could be applied to the full data prior to combination. This approach may also be useful if other methods are used to combine data from different excited slices to improve SNR.

Use of View Angle Tilting

The use of VAT to correct in-plane distortions is also chosen for simplicity. The in-plane correction is accurate to within BWRF/BWpix pixels, similar to that expected in MAVRIC. Although it is sometimes assumed that the voxel tilt in VAT causes blurring, this assumes that structures are somehow aligned with voxels, which is unlikely. Blurring in VAT is more commonly due to modulation effects when the readout duration is long (21). In SEMAC, short, high bandwidth readouts are used to avoid such blurring.

Residual Artifacts

In SEMAC scans, there is a minor residual artifact in areas near metal as discussed previously (7). This artifact is clearly visible along the screw seen in Fig. 5e. Although this is still being studied, we believe this artifact to be due to signal changes and slight (sub-pixel) shifts at boundaries between excited slices. Using the maximum excitation bandwidth possible reduces the artifact. In addition, changing the readout direction changes the appearance of the artifact.

In conclusion, we have demonstrated in phantoms and human subjects that SEMAC can be combined with standard spin echo contrast mechanisms, STIR fat suppression, and standard parallel imaging and partial Fourier techniques. The correction of distortion and pile-up artifacts with accelerated SEMAC is substantial better than with standard spin echo methods in all cases. Overall, these SEMAC variations permit flexible options for practical imaging of patients with metallic implants with reasonable scan times.


This work was supported by NIH-EB008190, NIH-RR009784 and GE Healthcare


1. Schenck JF. The role of magnetic susceptibility in magnetic resonance imaging: MRI magnetic compatibility of the first and second kinds. Med Phys. 1996;23:815–850. [PubMed]
2. Cho ZH, Ahn CB, Kim JH, Lee YE, Mun CW. Proc., SMRM, 6th Annual Meeting. New York: 1987. Phase error corrected interlaced echo planar imaging; p. 912.
3. Chang H, Fitzpatrick J. A technique for accurate magnetic resonance imaging in the presence of field inhomogeneities. IEEE Transactions on medical imaging. 1992;11:319–329. [PubMed]
4. Hoff M, Xiang Q. Proceedings of the 17th Annual Meeting of ISMRM. Honolulu: 2009. Eliminating metal artifact distortion using 3d-place; p. 570.
5. Olsen RV, Munk PL, Lee MJ, et al. Metal artifact reduction sequence: early clinical applications. Radiographics. 2000;20:699–712. [PubMed]
6. 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. [PubMed]
7. Lu W, Pauly KB, Gold GE, Pauly JM, Hargreaves BA. SEMAC: Slice encoding for metal artifact correction in MRI. Magn Reson Med. 2009;62:66–76. [PMC free article] [PubMed]
8. Hennig J, Nauerth A, Friedburg H. RARE imaging: A fast imaging method for clinical MR. Magn Reson Med. 1986;3:823–833. [PubMed]
9. Sodickson DK, Manning WJ. Simultaneous acquisition of spatial harmonics (SMASH): Fast imaging with radiofrequency coil arrays. Magn Reson Med. 1997;38:591–603. [PubMed]
10. Pruessmann KP, Weiger M, Scheidegger MB, Boesiger P. SENSE: Sensitivity encoding for fast MRI. Magn Reson Med. 1999;42:952–962. [PubMed]
11. Griswold MA, Jakob PM, Heidemann RM, et al. Generalized autocalibrating partially parallel acquisitions (GRAPPA) Magn Reson Med. 2002;47:1202–1210. [PubMed]
12. Noll DC, Nishimura DG, Macovski A. Homodyne detection in magnetic resonance imaging. IEEE Trans Med Imaging. 1991;10:154–163. [PubMed]
13. Bydder GM, Pennock JM, Steiner RE, Khenia S, Payne JA, Young IR. The short TI inversion recovery sequence – an approach to MR imaging of the abdomen. Magn Reson Imaging. 1985;3:251–254. [PubMed]
14. Hahn EL. Spin echoes. Phys Rev. 1950;80:580–594.
15. White LM, Kim JK, Mehta M, Gross AE, et al. Complications of total hip arthroplasty: MR imaging-initial experience. Radiology. 2000:215. [PubMed]
16. Cho Z, Kim D, Kim Y. Total inhomogeneity correction including chemical shifts and susceptibility by view angle tilting. Medical Physics. 1988;15:7–11. [PubMed]
17. McGowen A, MacKay A, Xiang QS. Reduction of image distortion in the presence of metal. International Society for Magnetic Resonance in Medicine; Vancouver: 1997.
18. Vandevenne J, Heiss S, Butts K, Bergman G, an SG, Lang P. MR imaging of orthopedic prostheses: Metal artifact reduction using view angle titlting. RSNA; Chicago: 1998. p. 465.
19. Lee MJ, Janzen DL, Munk PL, MacKay A, Xiang QS, McGowen A. Quantitative assessment of an MR technique for reducing metal artifact: application to spin-echo imaging in a phantom. Skeletal Radiol. 2001;30:398–401. [PubMed]
20. Kolind SH, MacKay AL, Munk PL, Xiang QS. Quantitative evaluation of metal artifact reduction techniques. J Magn Reson Imaging. 2004;20:487–495. [PubMed]
21. Butts K, Pauly JM, Gold GE. Reduction of blurring in view angle tilting MRI. Magn Reson Med. 2005;53:418–424. [PubMed]
22. Hinks R. Fast spin echo prescan for MRI system. US Patent. 5,378,985 Jan 3, 1995.
23. Hennig J, Scheffler K. Easy improvement of signal-to-noise in RARE-sequences with low refocusing flip angles. rapid acquisition with relaxation enhancement. Magn Reson Med. 2000;44:983–985. [PubMed]
24. Silver M, Joseph R, Hoult D. Selective spin inversion in nuclear magnetic resonance and coherent optics through an exact solution of the Bloch-Riccati equation. Physical Review A. 1985;31:2753–2755. [PubMed]
25. Beatty P, Brau A, Chang S, et al. Proceedings of the 15th Annual Meeting of ISMRM. Berlin: 2007. A method for autocalibrating 2D-accelerated volumetric parallel imaging with clinically practical reconstruction times; p. 1749.
26. Brau ACS, Beatty PJ, Skare S, Bammer R. Comparison of reconstruction accuracy and efficiency among autocalibrating data-driven parallel imaging methods. Magn Reson Med. 2008;59:382–395. [PubMed]
27. Wang J, Kluge T, Nittka M, Jellus V, Kuehn B, Kiefer B. Proceedings of the First Wurzburg Workshop on Parallel Imaging Basics and Clinical Applications. Wurzburg: 2001. Parallel acquisition techniques with modified SENSE reconstruction mSENSE.
28. Robson PM, Grant AK, Madhuranthakam AJ, Lattanzi R, Sodickson DK, McKenzie CA. Comprehensive quantification of signal-to-noise ratio and g-factor for image-based and k-space-based parallel imaging reconstructions [PMC free article] [PubMed]
29. Lu W, Pauly K, Gold G, Pauly J, Hargreaves B. Proceedings of the 17th Annual Meeting of ISMRM. Honolulu: 2009. Noise considerations in slice encoding for metal artifact correction; p. 2781.