This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.
To show the feasibility of metal artifact reduction in MRI at 3 T in comparison to 1.5 T using innovative MRI sequence strategies.
Materials and Methods
State-of-the-art techniques for metal artifact reduction including view angle tilting (VAT), slice encoding for metal artifact correction (SEMAC), and a combination thereof (SEMAC-VAT) were evaluated. Multi-sequence images were acquired on a 1.5 T system as well as on a 3 T system. Metallic implants within agarose and tissue phantoms as well as 3 healthy volunteers were imaged. Artifact reduction in the agarose phantoms was assessed by the artifact volume measurements utilizing all three techniques. A blinded read was conducted to similarly compare artifacts in tissue phantoms. In patients, SEMAC-VAT was compared to a conventional sequence.
Compared to conventional acquisitions with a high bandwidth, metal artifacts due to the stainless steel screw as judged by artifact volume were reduced, respectively, by 79%, 21%, 89% on T1 VAT, SEMAC without VAT, and SEMAC with VAT at 1.5 T and, respectively, by 62%, 63% and 76% on T1 VAT, SEMAC without VAT and SEMAC-VAT at 3 T. T2, PD and STIR demonstrated similar artifact reduction. For overall image quality in tissue phantoms, images were ranked as follows: SEMAC-VAT > VAT > SEMAC without VAT > conventional acquisition (kappa = 0.8). SEMAC-VAT was superior to the conventional acquisition in all cases.
SEMAC in combination with VAT yields superior reduction of artifacts from metallic orthopedic implants compared to conventional imaging or either technique alone.
Key Words: MRI - Metal Artifact - SEMAC - VAT
Clinical magnetic resonance imaging systems operating at a field strength of 3 T have become more widely adopted in recent years . The main motivation for the transition from 1.5 T to 3 T MRI systems is the improvement in the signal-to-noise ratio (SNR) which increases proportionally to field strength . Today, the availability of 3 T MRI systems has substantially increased and these systems now are used for imaging of all anatomical areas. MR imaging is safely performed in patients with metallic orthopedic implants, the vast majority of which are not ferromagnetic . As a result, postoperative MR imaging of these patients has become more common. Nevertheless, all commonly used metal implants induce MR artifacts, which often interfere with evaluation of the area of interest (i.e. soft tissue surrounding the implant or the bone marrow itself), thus diminishing the diagnostic value of the MR exam. A common application for post-operative MR imaging is the clinical scenario of suspected bone or soft tissue infections.
A brief theoretical discussion of the origins of metallic artifacts in MR images is warranted to better understand ways to reduce such artifacts. Susceptibility artifacts result from local inhomogeneity of the main magnetic field. These local areas of inhomogeneity cause dephasing of spins and frequency shifts of the surrounding tissues. After application of radiofrequency pulses and magnetic field gradients in an MR sequence, the radiofrequency signal received from the tissue may be corrupted by the presence of metal, altering its amplitude, frequency and phase. Intra-voxel dephasing decreases the signal received and results in hypointense areas in the MR image. Frequency shifts result in spatial misregistration seen as bright and dark areas in the image and as spatial distortion of surrounding anatomy. Spatial distortion may occur both in the image plane or as a 'bending' of the plane in the 3rd dimension. This includes severe thinning, or thickening of the slice profile as well as splitting it into multiple excited regions. Thus, at the location and in the immediate vicinity of metallic hardware, artifactually bright and dark areas may be seen . Spin Echo (SE) or Turbo Spin Echo (TSE) sequences are used due to their ability to re-phase off-resonant signal. A high radiofrequency bandwidth, high readout bandwidth, small slice thickness, and avoiding spectral fat saturation (by using STIR) are all measures that can be employed to reduce artifacts. However, a major disadvantage of all these approaches is a decrease in the signal to noise ratio (SNR) of the image.
Several sequence specific approaches have been reported recently in the literature for the reduction of metal induced artifact . One of those is view angle tilting (VAT). This technique uses an extra gradient in the slice select direction during the read-out gradient, such that the image pixels appear as if they were viewed from an angle. The sum of artifactual frequency shifts in the slice-select and the read-out direction results in a frequency shift with oblique direction. By viewing from this oblique angle during read out, the received signals can be projected into the correct pixel position of the image matrix. SEMAC (Slice Encoding for Metal Artifact Correction) adds phase encoding in the slice direction in order to resolve distortions of the slice profile . During image reconstruction this information is used to sort distorted pixels into their correct slice position. The principle drawback of this technique is the requisite increase in scan time. As MR imaging is an exquisite modality to evaluate osseous and soft tissue structures, the aim of our study is to evaluate the feasibility of metal artifact reduction, in the presence of orthopaedic implants, on imaging at 1.5 T as compared to 3 T, using an approach combining SEMAC and VAT.
Material and Methods
A Work In Progress software package ('WARP' WIP#648, Siemens Healthcare, Germany) was utilized for imaging. The contained 2D TSE sequence features increased RF pulse bandwidths, as well as the VAT and the SEMAC techniques. VAT and SEMAC can be switched on and off independently.
In all cases described below, the metallic hardware was oriented parallel to the direction of the main magnetic field (B0). The experimental design is detailed in Table 1.
A stainless steel screw (Smith & Nephew 3.5 mm standard cortical 38 mm locking screws) and a stainless steel plate (Smith & Nephew Periloc 6-hole 4.5 mm locking plates) were placed in two separate plastic containers filled with agar. Each implant was placed in the central aspect of each phantom. Each phantom was positioned at the center of a flexible 4-element receiver coil and scanned on both a 1.5 T (Magnetom Espree) and a 3 T (Magnetom Verio) MR-system. Images with T1-, T2-, PD- and STIR contrast were obtained with four different techniques including the standard default protocol (called product), VAT, SEMAC without VAT and SEMAC-VAT (resulting in sixteen scans). T1 and PD scans were obtained in axial and sagittal planes. The details of the MR parameters are listed in Table 2 with scan times provided in Table 3. Common imaging parameters for these four sequences were a 20 cm field of view, a 256x256 matrix, a 2.4 mm slice thickness and an anterior-posterior frequency encoding direction. Representative images are provided in Figure 1.
Artifact volume was measured offline using a dedicated workstation (GE Advantage Workstation AW 4.2). The semiquantitative approach for artifact volume quantification originally described in Lee et al was adopted for analysis of each pulse sequence . This technique was favourable due to its ease of implementation and ability to account for both high and low signal intensity artifact. To summarize the approach, an agarose gel phantom without any metallic implants was first imaged. The signal intensity distribution of a standardized central area was calculated utilizing the histogram tool of the workstation. From this, the normal range of signal intensities was defined as being voxels with signal intensity within three standard deviations of the mean signal intensity. Next, the agarose gel containing the metallic implant was imaged with the same pulse sequence. A histogram analysis of signal intensity was performed similarly. The number of pixels falling outside the normal range of signal intensities was computed, and these were considered artifact. Analogous calculations were performed throughout the entire imaging volume to compute a measure of total artifact volume. This is illustrated in Figure 2.
Pig leg phantom scans
A stainless steel plate was surgically attached to the femur of a pig leg, with three stainless steel screws in one phantom and six stainless steel screws in another. The pig legs were imaged first on a 1.5 T MR-system (Siemens Espree) and subsequently on a 3 T MR-system (Siemens Verio). The standard 4 channel flex coil was used for signal reception at both field strengths. Ethical approval is not required at our institution for animal cadaver studies. Imaging parameters were held as constant as possible for the different field strengths. An 18 cm field of view, a 256x256 matrix and a slice thickness of 2.5 mm were used. A field of view encompassing the pig leg, given its size and orientation, was selected. Imaging was performed in the coronal plane. T1, T2, PD and STIR sequences were acquired with all four techniques including specifically product, VAT, SEMAC without VAT and SEMAC-VAT.
The images from this in vitro tissue phantom evaluation were assessed by a blinded read. The images were ranked from 1 to 4. Two readers with over 5 years of experience evaluated the images as did two radiology residents. The readers were not aware of the magnetic field strength or sequence type at the time of interpretation. Image quality with the four different sequences was compared. The images from each technique with the same contrast were compared side by side.
Images from each sequence were assigned a score ranging from 1-4. Images ranked as 1 demonstrated the best image quality and 4 the worst image quality within a group. If two techniques exhibited similar image quality, the images were given the same ranking. (Figure 3)
Three patients with metallic implants were included in this Institutional Review Board approved study to provide a limited in vivo assessment of the sequences in question, with the intent of confirming the results obtained in phantoms. 1 was a man, and 2 were women, with ages ranging from 48 to 58 years. All patients had undergone surgical fixation with metallic implants from a few months to several years previous. Implants evaluated included stainless steel, titianium alloy and cobalt-chromium. All examinations were performed on both the 1.5 T and 3 T Siemens scanners. An extremity coil was used for all studies. STIR coronal product, T1 axial product, PD sagittal product and STIR coronal SEMAC-VAT, T1 axial SEMAC-VAT and PD sagittal SEMAC-VAT were acquired.
The frequency-encoding axis was set to an anterior-posterior direction for all patients. Image assessment was performed in a blinded read. The readers were provided all slices for the acquired images and were asked to make a judgment of the overall degree of artifact. The readers were not aware of the magnetic field strength or sequence type at the time of interpretation. The images were ranked from 1 to 4. Two readers evaluated the images. Both readers had at least 5 years of radiology experience. The readers assessed four specific factors: size of the metal artifact, distortion of normal structures adjacent to the metal, ability to visualize the bone marrow, bone cortex and soft tissue and image blur. The images from each technique with the same contrast were compared side by side, with results from the two different field strengths also directly compared. The images were ranked by the readers in a similar fashion to the comparison described above. (see Figure 4)
Statistical analysis was performed using commercial statistical software
(R, v 2.9.0). As multiple readers were evaluated, a special test created by Fleiss as a modification of CohenÂ´s Kappa was used to calculate the value of the Kappa coefficient with K < 0.40 indicating low association, K between 0.40 and 0.70 indicating medium association; and values greater than 0.70 indicating high association between the different readers.
The measured metallic artifact volume sizes for the stainless steel screw was reduced by 79%, 21%, and 89% on T1 VAT, SEMAC without VAT and SEMAC-VAT, respectively, at 1.5 T and by 62%, 63% and 76% on T1 VAT, SEMAC without VAT and SEMAC-VAT, respectively, at 3 T when comparing to the equivalent product sequences. Representative images are provided in Figure 2. The measured metallic artifact volume sizes for the stainless steel plate was reduced by 67%, 87% and 91% on T1 VAT, SEMAC without VAT and SEMAC-VAT, respectively, at 1.5 T and by 36%, 51% and 59% on T1 VAT, SEMAC without VAT and SEMAC-VAT, respectively, at 3 T. T2, PD and STIR demonstrated similar artifact reduction (Table 4).
The measured artifact volume for 3 T when compared to 1.5 T scans was increased by 48% for T1, 56% for T2, 79% for PD and 56% for STIR. The overall measured artifact volume for 3 T scans in comparison to 1.5 T scans increased over 60%.
We also compared the measured artifact volume at 3 T for SEMAC-VAT sequence to conventional acquisitions optimized to reduce metallic artifact on a 1.5 T scanner. Metal artifacts due to the stainless screw were reduced by 50% for T1, 45% for T2, 28% for PD and 42% for the STIR sequence. Artifact volumes are illustrated in Figure 5.
In the qualitative analysis, the images were ranked on the basis of the following parameters: artifact size, distortion of normal structures, and the ability to visualize the bone marrow, bone cortex and soft tissues. The sequences were ranked SEMAC-VAT at 1.5 T > SEMAC-VAT at 3 T > conventional sequence 1.5 T > conventional sequence 3 T for artifact size in the blinded read. For the evaluation of the bone marrow itself, images were ranked SEMAC-VAT at 1.5 T > SEMAC-VAT at 3 T > conventional sequence 1.5 T > conventional sequence 3 T. For the evaluation of distortion, images were ranked SEMAC-VAT at 1.5 T > SEMAC-VAT at 3 T > conventional sequence 1.5 T > conventional sequence 3 T. Representative images are provided in Figure 3. For the evaluation of blur, the images were ranked as follows, conventional sequence 3 T > conventional sequence 1.5 T > SEMAC-VAT 3 T > SEMAC-VAT 1.5 T. The Kappa for the blinded read was 0.788, which indicates a strong level of agreement between the two readers.
In terms of overall image quality, images were ranked as follows, SEMAC-VAT 1.5 T > SEMAC-VAT 3 T > conventional sequence 1.5 T > conventional sequence 3 T.
For volunteer scans, SEMAC-VAT 1.5 T was ranked superior to SEMAC-VAT 3 T > conventional sequence 1.5 T > conventional sequence 3 T for T1,T2 and PD. For STIR, the images were ranked SEMAC-VAT 3 T > SEMAC-VAT 1.5 T > conventional sequence 1.5 T > conventional sequence 3 T. Representative images are provided in Figure 4.
The clinical use of 3 T MRI continues to increase, principally due to the improved SNR relative to 1.5 T, which can be employed for either improved spatial resolution and/or faster scan times. However, the increased prominence of susceptibility artifacts at 3 T have prevented its mainstream use for imaging of patients with prosthetic devices , whereas 1.5 T MRI is considered clinically useful for evaluating peri-prosthetic tissue . Depending on the strength of the main magnetic field and magnetic susceptibility of the implanted metallic hardware, substantial changes of the local field around the implant may occur. As a consequence, MR imaging of the tissue adjacent to metallic implants suffers from severe signal off-resonance and de-phasing effects, decreasing diagnostic utility.
Our study demonstrate that VAT and SEMAC sequences previously utilized at 1.5 T , can be adapted to 3 T with similar success. More importantly both methods may be combined to correct both through-plane and in-plane distortion. Previous work on improving MR imaging near metallic implants was based on the view angle tilting technique, originally described by Cho et al . Kolind et al previously described a VAT-SE sequence utilizing high RF and readout bandwidths. This sequence was referred to as a "metal artifact reduction sequence" (MARS). However, these studies only addressed in-plane distortion correction, and did not deal with the topic of through-plane distortions. In contrast to VAT, SEMAC reduces through-plane distortion. This technique corrects metal artifacts utilizing additional phase encoding steps for each excited slice. Implementation of the SEMAC technique only requires installation of appropriate software updates on existing whole-body MRI systems and does not depend upon any additional hardware installation . However, as it requires additional phase-enconding steps, SEMAC does increase scan time, a major drawback limiting its use in clinical routine.
The evaluation of VAT or SEMAC-based sequences alone demonstrated a similar amount of artifact reduction both at 1.5 T and 3 T, when compared to conventional sequences. However, for successful correction of through-plane distortion, SEMAC requires a certain number of z-phase encoding steps, depending upon the actual degree of distortion. Since scan time linearly increases with the applied number of phase encoding steps (`z`), values beyond 15 may easily exceed a scan duration that is acceptable for routine exams. In our study we used 15 z-phase encoding steps, which shows effective artifact correction with an acquisition time of 8:41 min:sec for 1.5 T and 12:32 min:sec at 3 T for T1 scans (Table 3). When comparing VAT and SEMAC without VAT, VAT demonstrated better artifact reduction, as proven by both volumetric measurements as well as the blinded read. However, since metallic artifact occurs in three dimensions, with both in-plane and through-plane artifacts, it is difficult to compare SEMAC without VAT to VAT directly.
Our study demonstrated that the combined SEMAC technique with VAT reliably corrected for metal artifacts at both field strengths with utility for T1, T2, PD and STIR sequences. Of all image contrasts, SEMAC with VAT provided the least artifact reduction on the STIR sequence. The mechanism for disparities in artifact reduction based on image contrast remains unclear. It is possible that the semiquantitative methodology employed for artifact volume measurements could be susceptible to thresholding effects. Thus, this method of volumetric artifact calculation may only be useful in comparing scans with identical image contrasts at the same field strength. However, in that context the method is a useful way to determine the efficacy of artifact reduction among the SEMAC and VAT sequences.
Scan time can be considerably reduced by utilizing dedicated acquisition techniques such as parallel imaging, partial Fourier sampling, and long echo trains. A high RF pulse bandwidth is desired to limit the extent of slice distortion. However, an increase in RF pulse bandwidth increases the Specific Absorption Rate (SAR), a critical limitation at 3 T. In this study a 1.8 kHZ RF bandwidth was chosen for SEMAC-VAT. Combined with parallel imaging (GRAPPA, acceleration factor of 2) the scan time was shortened to 8-12 minutes per acquisition compared to acquisitions without GRAPPA.
Blurring may be observed on VAT sequences due to the shear effect of the slice and a low-pass filter effect in the frequency domain during signal readout . Butts et al reduced VAT associated blurring by shortening the readout time. Generally, readout time should not exceed the main lobe of the RF pulse's duration . This approach was implemented in the work herein. Of note, the shear effect blurs voxels in structures perpendicular to the slice, and the blur may be less if the structure is obliquely oriented relative to the slice. Higher readout bandwidth or a reduction in the number of readout samples can be used to shorten the readout at the expense of lowered resolution respectively. The results from the present work demonstrate that high readout bandwidth (931 Hz per pixel) reduced the blurring in the imaging with VAT and SEMAC-VAT in comparison to VAT and only slight blurring could be recognized on tissue images.
This study had several limitations. Comparing images acquired at 1.5 T and 3 T entails comparing images obtained on different machines. The MRI units in this study differed in field strength but were made by the same manufacturer. Differences in image quality may nevertheless relate to disparities in hardware design unrelated to field strength. Second, the method of volumetric measurements is limited, not only due to the potential for thresholding effects described above, but also due to the fact that it represents a semiquantitative measure requiring a user to perform histogram analyses. Nevertheless, the ease of use and of obtaining three-dimensional artifact measurements is advantageous. Third, the readers were blinded to the field strength at which the images were obtained. However, qualitative differences in the acquired images may have allowed the observers to discern which images were obtained at 1.5 T and at 3 T. This is a potential source of bias. The volumetric measurements of artifact reduction also differed considerably between field strengths with the SEMAC (without VAT) technique. The reduction in relative artifact volume at 3 T (63%) was considerably greater than at 1.5 T (21%). As the SEMAC technique corrects through-plane distortions, and as these are generally much greater at 3 T, the artifact volume reduction may as a result be much more pronounced when SEMAC is applied. However, careful further study of these techniques at both field strengths would be useful to confirm this effect. Fourth, we only included three volunteer scans. In the future, further in vivo studies of patients with metal implants will be required to obtain more conclusive information in regards to the clinical utility of these techniques. Finally, the current long acquisition times (9-12 min) at 1.5 T and 3 T for SEMAC-VAT sequences make it difficult to implement these sequences in routine clinical imaging. However it should be emphasized that few other effective alternatives exist today other than this approach, for visualization of soft tissue adjacent to metal implants.
In summary, SEMAC in combination with VAT provides for substantial reduction of metal induced artifacts, a historical limitation to the clinical MRI, in particular at 3 T. The use of SEMAC-VAT reduces susceptibility artifacts to a level less than that observed in sequences performed without such corrections at 1.5 T, improving markedly soft tissue visualization adjacent to orthopedic hardware at 3 T. This valuable imaging method may help the radiologist in diagnosing recurrent tumor, fracture, infection in the region of implant or loosening of the implant. Whether applied at 1.5 or 3 T, SEMAC-VAT demonstrated a consistent reduction in the extent of metallic artifacts.
Table 1. Summary of Sequences Evaluated
Coronal T1, T2, PD, STIR +
axial, sagittal T1 & PD
Coronal T1, T2, PD, STIR
Coronal T1, T2, PD, STIR +
axial, sagittal T1 & PD
Coronal T1, T2, PD, STIR
Coronal T1, T2, PD, STIR +
Coronal T1, T2, PD, STIR
axial, sagittal T1 & PD
Coronal T1, T2, PD, STIR +
Coronal T1, T2, PD, STIR
axial, sagittal T1 & PD
Table 2. Scan Parameters of Sequences Performed at 1.5 and 3 T
Table 3. Acquisition Times of Evaluated Sequences
Table 4. Percentage Artifact Reduction for Each Evaluated Sequence Relative to Conventional Acquisition
Figure and Captions
Fig. 1 - Agarose phantom containing one stainless steel screw evaluated at 1.5 T (A,B) and 3 T (C,D) using conventional T1 weighted scans (A,C) and SEMAC-VAT (B,D). At both field strengths, a substantial reduction in metal artifact is noted, although overall the artifact is greater at 3 T.
Fig. 2 - Illustration of the semiquantitative approach utilized for volume quantification as further described in the text. Histogram analysis of the agarose gel was used to obtain normal signal intensity values. These were defined as being within 3 standard deviations of the mean signal intensity. As shown above, a similar analysis was applied to the phantom containing the metallic implant. The normal value cutoffs were set on the resulting histogram and areas of low and high signal intensity artifact computed. This was performed analogously throughout the imaging volume to obtain a measure of total artifact volume.
Fig. 3 - Pig leg phantom containing one plate and six stainless steel screws, evaluated at 1.5 T (A,B) and 3 T (C,D), using conventional PD scans (A,C) and SEMAC-VAT (B,D). At both field strengths a substantial reduction in metal artifact is noted, and the bone marrow adjacent to the metal is better visualized, although overall the artifact is greater at 3 T.
Fig. 4 - Volunteer scan with stainless steel plate and six stainless steel screws placed for a Weber C fracture, evaluated at 1.5 T (A,B) and 3 T (C,D) using conventional STIR scans (A,C) and SEMAC-VAT (B,D). At both field strengths, a substantial reduction in metal artifact is noted with SEMAC-VAT.
Fig. 5 - Metal artifact volume (average of all sequence contrasts) for 1.5 T versus 3 T