reduction of Metal artifacts in Patients with Total hip

Original Research n Musculoskeletal

Imaging

Note: This copy is for your personal non-commercial use only. To order presentation-ready copies for distribution to your colleagues or clients, contact us at www.rsna.org/rsnarights.

Reduction of Metal Artifacts in Patients with Total Hip Arthroplasty with Slice-encoding Metal Artifact Correction and View-Angle Tilting MR Imaging1 Reto Sutter, MD Erika J. Ulbrich, MD Vladimir Jellus, PhD Mathias Nittka, PhD Christian W. A. Pfirrmann, MD, MBA

Purpose:

To compare the new “warp” sequence (slice-encoding metal artifact correction [SEMAC], view-angle tilting [VAT], and increased bandwidth) for the reduction of both through-plane and in-plane magnetic resonance (MR) artifacts with current optimized MR sequences in patients with total hip arthroplasty (THA).

Materials and Methods:

The institutional review board issued a waiver for this study. Forty patients with THA were prospectively included. SEMAC, VAT, and increased bandwidth were applied by using the warp turbo-spin-echo sequence at 1.5 T. Coronal short tau inversion-recovery (STIR)-warp and transverse T1-weighted warp (hereafter, T1-warp) images, as well as standard coronal STIR and transverse T1-weighted sequence images optimized with high bandwidth (STIRhiBW and T1-hiBW), were acquired. Fifteen additional patients were examined to compare the T1-warp and T1hiBW sequence with an identical matrix size. Signal void was quantified. Qualitative criteria (distinction of anatomic structures, blurring, and noise) were assessed on a fivepoint scale (1, no artifacts; 5, not visible due to severe artifacts) by two readers. Abnormal imaging findings were recorded. Quantitative data were analyzed with a t test and qualitative data with a Wilcoxon signed rank test.

Results:

Signal void around the acetabular component was smaller for STIR-warp than STIR-hiBW images (21.6 cm2 vs 42.4 cm2; P = .0001), and for T1-warp than T1-hiBW images (17.6 cm2 vs 20.2 cm2; P = .0001). Anatomic distinction was better on STIR-warp compared with STIR-hiBW images (1.9–2.8 vs 3.6–4.6; P = .0001), and on T1-warp compared with T1-hiBW images (1.3–2.8 vs 1.8–3.2; P , .002). Distortion, blurring, and noise were lower with warp sequences than with the standard sequences (P = .0001). Almost half of the abnormal imaging findings were missed on STIR-hiBW compared with STIR-warp images (55 vs 105 findings; P = .0001), while T1-hiBW was similar to T1-warp imaging (50 vs 55 findings; P = .06).

Conclusion:

STIR-warp and T1-warp sequences were significantly better according to quantitative and qualitative image criteria, but a clinically relevant artifact reduction was only present for STIR images.

1

From the Department of Radiology, Orthopedic University Hospital Balgrist, Forchstrasse 340, 8008 Zurich, Switzerland (R.S., E.J.U., C.W.A.P.); and Healthcare Sector, Imaging & Therapy Division, Magnetic Resonance, Siemens, Erlangen, Germany (V.J., M.N.). Received November 11, 2011; revision requested January 10, 2012; revision received March 13; accepted April 11; final version accepted April 24. Address correspondence to R.S. (e-mail: reto [email protected]). RSNA, 2012

q

204

RSNA, 2012

q

radiology.rsna.org n Radiology: Volume 265: Number 1—October 2012

MUSCULOSKELETAL IMAGING: Reduction of Metal Artifacts in Patients with Total Hip Arthroplasty

M

agnetic resonance (MR) imaging of total hip arthroplasty (THA) has been very limited for many years due to strong local susceptibility artifacts induced by the metal (1,2). More recently, MR imaging has increasingly been used in patients with THA, because unlike radiography, computed tomography (CT), scintigraphy, or positron emission tomography, it also enables visualization of the soft tissues surrounding the hip joint, yet artifacts still hamper MR image quality and image interpretation substantially (3,4). While the soft tissues that are located further from the metallic implant might be depicted well on MR images (5,6), it is particularly challenging to assess the bone and soft tissues directly adjacent to the hip prosthesis (7). In patients without orthopedic implants, fluid-sensitive sequences are helpful in detecting subtle abnormalities of bone and soft tissue (8,9). An MR technique that minimizes metalinduced artifacts and allows the evaluation of THA with anatomic sequences as well as with a fluid-sensitive sequence might be a useful tool for radiologists and orthopedic surgeons. This technique might aid in the visualization of

Advances in Knowledge nn The warp MR sequence, combining slice-encoding metal artifact correction (SEMAC) and view-angle tilting (VAT), achieved smaller artifacts and better image quality in patients with total hip arthroplasty for both short tau inversion-recovery (STIR) and T1-weighted sequences, but only STIR images benefited from a clinically relevant artifact reduction. nn Almost half of the abnormal imaging findings seen on the STIRwarp images were missed on the standard STIR sequence images optimized with high bandwidth (STIR-hiBW), and presumed periprosthetic osteolysis was detected three times more often on the STIR-warp images compared with STIR-hiBW images.

periprosthetic osteolysis, periprosthetic fractures, or joint effusions. Several MR techniques are able to reduce in-plane distortion artifacts such as high-readout bandwidth or viewangle tilting (VAT) (10–13). However, a substantial amount of metal-induced artifacts are due to distortion of the excited plane perpendicular to the readout direction (denoted as “through-plane” distortion), which cannot be addressed by increased readout bandwidth or VAT (14). Slice-encoding metal artifact correction (SEMAC) in combination with VAT is known to reduce both throughplane and in-plane distortion artifacts in MR imaging phantom studies and selected in vivo examples (15,16). Furthermore, a recent article showed that the SEMAC technique reduced artifact size from total knee arthroplasty in comparison with a fast-spin-echo sequence in 14 volunteers and 11 patients (17). In our study, the SEMAC/ VAT technique was applied by means of a “warp” turbo-spin-echo (TSE) sequence that combines these new techniques with an optimized sequence design for high-bandwidth acquisitions. The purpose of our study was to compare the warp sequence for the reduction of both through-plane and inplane MR artifacts to current optimized MR sequences in clinical patients with THA.

Materials and Methods Patients This study was submitted to the institutional review board and a waiver of specific formal review was issued. All patients signed informed consent as part of their research hospital admission. Forty consecutive patients (mean age, 64.5 years; range, 36–92 years) with THA referred for MR imaging of Implication for Patient Care nn The STIR-warp sequence reduces metal artifacts around hip prostheses to an extent that might be helpful for the evaluation of periprosthetic complications.

Radiology: Volume 265: Number 1—October 2012 n radiology.rsna.org

Sutter et al

the abductor tendons were included between March and May 2011. The study population consisted of 21 women (mean age, 66.4 years; range, 36–92 years) and 19 men (mean age, 62.4 years; range, 43–86 years). Twenty-three patients had THA of the left hip, and 17 had THA of the right hip. Exclusion criteria were pregnancy or the lack of informed consent. No patients were excluded. Between January and February 2012, an MR imaging examination was performed in an additional 15 patients (mean age, 59.3 years; range, 19–79 years) to compare the T1-weighted warp sequence and the high-bandwidth-optimized T1-weighted sequence with an identical matrix size. This patient population consisted of six women (mean age, 61.7 years; range, 47–77 years) and nine men (mean age, 56.8 years; range, 19–79 years); 11 patients had THA of the left hip, and four had THA of the right hip.

MR Imaging All patients were examined with a 1.5T imager (Magnetom Avanto; Siemens Healthcare, Erlangen, Germany) with a gradient field strength of 45 mT/m and a slew rate of 200 T/m/sec. For this Published online before print 10.1148/radiol.12112408 Content code: Radiology 2012; 265:204–214 Abbreviations: SEMAC = slice-encoding metal artifact correction STIR = short tau inversion-recovery STIR-hiBW = STIR sequence optimized with high bandwidth T1-hiBW = T1-weighted sequence optimized with high bandwidth THA = total hip arthroplasty TSE = turbo spin echo VAT = view-angle tilting Author contributions: Guarantor of integrity of entire study, R.S.; study concepts/ study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; literature research, all authors; clinical studies, R.S., E.J.U.; statistical analysis, R.S.; and manuscript editing, all authors Potential conflicts of interest are listed at the end of this article.

205


Figure 1

Figure 1: Pulse sequence diagram of the warp sequence used in this study that implements the SEMAC and VAT techniques. The warp sequence is based on a conventional TSE sequence, and additional gradients are applied for section phase encoding and VAT. Besides a high-bandwidth radiofrequency pulse and increased readout bandwidth, the section-selection gradients for excitation, refocusing, and the optional inversion recovery pulse are set to identical amplitudes to minimize off-resonance–related artifacts. G = gradient, IR = optional inversion recovery pulse, RF = radiofrequency excitation.

study we used a work-in-progress software package providing a TSE sequence adapted for implant imaging (warp TSE; Siemens Healthcare) that implements the SEMAC and VAT techniques into a standard TSE sequence combined with high-bandwidth radiofrequency pulses as well as increased readout bandwidth (Fig 1). Two coauthors (V.J. and M.N.) are employees of Siemens and were involved in the technical development of the warp sequence. They did not have access to key portions of the study data. The study data were controlled independently by the other authors (R.S., E.J.U., and C.W.A.P). Patients were positioned supine on the MR examination table and a body matrix phased-array surface coil (six receiver channels) in combination with a spine matrix coil (six receiver channels selected) were used for image acquisition. The following four sequences were acquired over the hip and proximal femur and subsequently analyzed in this study: 1. A coronal short tau inversionrecovery (STIR) sequence (called “STIR-warp” in our study) optimized with SEMAC in combination with VAT 206

(repetition time msec/echo time msec, 5000/29; section thickness, 3.5 mm; refocusing flip angle, 150°; field of view, 22 cm; matrix, 205 3 256; one signal acquired; excitation bandwidth, 1.4 kHz; readout bandwidth, 781 Hz/pixel; echo spacing, 5.84 msec; echo train length, 23; inversion time, 145 msec; number of sections, 15; section-encoding steps, 10; acquisition time, 5:57 min). 2. A coronal standard STIR sequence optimized with high bandwidth (STIRhiBW) (5000/30; section thickness, 3.5 mm; refocusing flip angle, 150°; field of view, 22 cm; matrix, 205 3 256; two signals acquired; excitation bandwidth, 1.2 kHz; readout bandwidth, 781 Hz/ pixel; echo spacing, 7.44 msec; echo train length, 23; inversion time, 145 msec; number of sections, 15; sectionencoding steps, not applicable; acquisition time, 3:07 min). 3. A transverse T1-weighted warp sequence (hereafter, “T1-warp”) optimized with SEMAC in combination with VAT (470/6.4; section thickness, 6 mm; refocusing flip angle, 130°; field of view, 20 cm; matrix, 240 3 320; one signal acquired; excitation bandwidth, 1.4 kHz; readout bandwidth, 781 Hz/

Sutter et al

pixel; echo spacing, 6.4 msec; echo train length, seven; number of sections, 20; section-encoding steps, 12; acquisition time, 6:32 min). 4. A transverse standard T1-weighted sequence optimized with high bandwidth (T1-hiBW) (484/9.7; section thickness, 6 mm; refocusing flip angle, 180°; field of view, 20 cm; matrix, 512 3 512; three signals acquired; excitation bandwidth, 1.2 kHz, readout bandwidth, 425 Hz/pixel; echo spacing, 9.7 msec; echo train length, three; number of sections, 20; section-encoding steps, not applicable; acquisition time, 4:26 min). The T1-warp sequence can be performed with a maximum matrix of 240 3 320 on our imager without changing further imaging parameters, while the T1-hiBW sequence used in clinical routine has a larger matrix. To compare both T1-weighted sequences with the same matrix size, we performed an additional analysis in 15 patients both with the T1-warp sequence and with a modified T1-hiBW sequence. The latter sequence was the same as the T1warp sequence, but with SEMAC and VAT switched off, and the other imaging parameters (including the matrix) kept constant. The modified T1-hiBW sequence had the following imaging parameters: 470/6.4; section thickness, 6 mm; refocusing flip angle, 130°; field of view, 20 cm; matrix, 240 3 320; one signal acquired; excitation bandwidth, 1.4 kHz; readout bandwidth, 781 Hz/ pixel; echo spacing, 6.4 msec; echo train length, seven; number of sections, 20; section-encoding steps, not applicable; acquisition time, 0:35 minute. Our routine protocol further included a sagittal T1-weighted fast spinecho sequence, a coronal T2-weighted fast spin-echo sequence, and a transverse STIR sequence, all optimized with high bandwidth for metal artifact reduction.

Quantitative Image Analysis A fellowship-trained musculoskeletal radiologist (R.S., with 6 years of experience) measured the cumulative area of signal void on the image, which was defined as the area without discernible



anatomic information, including both low- and high-signal-intensity artifacts induced by the prosthesis. On coronal STIR images, two signal void measurements were performed: The first measurement area included the acetabular component and the head of the prosthesis (at the level of the center of the head of the prosthesis), and the second measurement area included the neck and shaft of the prosthesis (at the level at which the neck and shaft of the prosthesis were visualized to the largest extent). On transverse T1-weighted images, two other signal void measurements were performed: The first measurement area included the acetabular component and the head of the prosthesis at the level of the center of the head (Fig 2) as well as the part of the neck and shoulder of the prosthesis visible on the same image. The second measurement area included the shaft of the prosthesis (at the level of the lesser trochanter). All measurements were performed at the same level for the corresponding sequences. The region of insufficient inversion (ie, incomplete fat suppression and also reduced water suppression) was measured for both STIR sequences at the level of the greatest extent of the artifact: Insufficient inversion was defined as the area adjacent to the orthopedic hardware where markedly increased signal intensity was present in a nonanatomic distribution (18) (Fig 3).

Qualitative Image Analysis Two fellowship-trained musculoskeletal radiologists (R.S. and E.J.U., with 6 and 7 years of experience, respectively), blinded to the clinical diagnosis, independently compared the STIR-hiBW sequence with the STIRwarp sequence, as well as the T1-hiBW sequence with the T1-warp sequences. These observers evaluated the distinction of several defined anatomic structures and image quality (geometric distortion, spatial blurring, noise). In a separate consensus reading, the number of abnormal imaging findings discernible was noted. Distinction of anatomic structures.—On both the coronal and

Sutter et al

Figure 2

Figure 2: Images in 84-year-old female patient with THA of the right hip. (a) Image obtained with standard transverse T1-hiBW technique. (b) Transverse T1-warp image at the same level. Top images show signal void area (outlined area), and identical bottom images demonstrate normal anatomic structures. Signal void area is larger on the T1-hiBW image than on the T1-warp image. Several anatomic structures, such as the sartorius muscle (∗) and the musculotendinous junction of both the gluteus minimus (arrowhead) and gluteus medius muscle (arrow), are well depicted on b but are not visible on a. Image noise is more pronounced on the T1-hiBW compared with the T1-warp image.

transverse images the gluteus minimus and the gluteus medius tendon attachments were assessed. On the coronal STIR images, the indirect head of the rectus femoris tendon and the hip joint capsule were assessed as anatomic structures that are located superior to the THA, and the iliopsoas tendon and the hip joint capsule were assessed directly inferior to the THA. On the transverse T1-weighted images, the iliopsoas tendon and the hip joint capsule were assessed anterior to the THA, and the internal obturator tendon was assessed posterior to the THA. Anatomic structures were classified according to a fivepoint scale: score of 1, good depiction of structure; 2, structure fully visible, slight blurring of borders; 3, structure


fully visible, substantial blurring of borders; 4, structure only partially visible; and 5, structure not visible. Image quality.—Geometric image distortion, spatial blurring, and image noise were graded on a five-point scale: score of 1, no artifacts; 2, barely visible artifacts; 3, visible artifacts without impairment of diagnostic quality; 4, moderate artifacts with moderate impairment of diagnostic quality; and 5, severe artifacts and nondiagnostic image. If abnormal imaging findings were present, such as trochanteric bursitis, gluteal tendon tears, periprosthetic osteolysis, or hip joint effusion (1,19,20), this was recorded and the number of such findings per sequence was noted 207


Sutter et al

(Fig 4). Qualitative and quantitative analysis were performed at least 2 weeks apart (range, 2–8 weeks) to minimize a possible recollection bias.

Figure 3

Figure 3: Images in 63-year-old female patient with THA of the left hip. Distortion artifacts, signal pile-up, and insufficient inversion (arrows) are substantially larger on the (a) coronal STIR-hiBW image compared with the (b) coronal STIR-warp image. The insertion of the gluteus medius tendon at the lateral facet of the greater trochanter (arrowhead) is better depicted with STIR-warp (b) compared with STIR-hiBW (a) imaging.

Figure 4

Statistical Analysis All analyses were performed with statistical software (SPSS for Windows, release 17.0; SPSS, Chicago, Ill). Descriptive statistics were used to compare signal void size and areas with insufficient inversion, and mean values and standard deviations were calculated. Differences in signal void size and areas with insufficient inversion were assessed by using the paired t test, with P , .05 indicating a significant difference. Differences in qualitative data (distinction of anatomic details, image quality, and number of abnormal imaging findings) were assessed with the Wilcoxon signed-rank test. After Bonferroni correction to adjust for multiple comparisons, a P value of less than .0125 was considered to be indicative of a statistically significant difference for the distinction of anatomic details, and a P value of less than .0167 was considered to be indicative of a statistically significant difference for image quality. The number of discordant cases of abnormal imaging findings detected by means of the different MR sequences was analyzed with a McNemar test. Agreement between the two readers was determined by calculating k values. A k value of 0 indicated poor agreement; 0.01–0.20, slight agreement; 0.21–0.40, fair agreement; 0.41–0.60, moderate agreement; 0.61– 0.80, good agreement; and 0.81–1.00, excellent agreement (21). Results

Figure 4: Images in 69-year-old male patient with THA of the left hip. Artifacts are larger on the (a) coronal STIR-hiBW image compared with the (b) coronal STIR-warp image. Note differences in artifact size both for the acetabular component (open arrows) as well as for the femoral component and fixation wires (solid arrows). The periprosthetic osteolysis (arrowhead) at the lateral aspect of the shaft is better demarcated at STIR-warp (b) than at STIR-hiBW (a) imaging, due to the reduced level of artifacts on the STIR-warp image. 208

Quantitative Image Analysis On coronal STIR images the area of signal void for the acetabular component (mean 6 standard deviation) was 21.6 cm2 6 9.5 and 42.4 cm2 6 27.9 (for STIR-warp [combining SEMAC, VAT, and increased bandwidth] and STIRhiBW imaging, respectively) and for the neck and shaft of the prosthesis it was 27.6 cm2 6 9.2 and 43.5 cm2 6 18.3,



respectively. For both comparisons, the area of signal void was statistically significantly smaller for STIR-warp compared with STIR-hiBW images (P = .0001) (Fig 5a). The area of insufficient inversion was also statistically significantly smaller for STIR-warp compared with STIR-hiBW images, with a mean area of 3.1 cm2 6 7.0 for STIR-warp images and 27.9 cm2 6 27.0 for STIRhiBW images (P = .0001) (Fig 5a). On transverse T1-weighted images, the area of signal void at the proximal level was significantly smaller for T1warp than for T1-hiBW imaging (17.6 cm2 6 4.9 for T1-warp and 20.2 cm2 6 5.9 for T1-hiBW images; P = .0001) (Fig 5b). There was no statistically significant difference between the area of signal void between the two T1weigthed sequences at the distal level that consisted of the shaft of the prosthesis (3.5 cm2 6 1.7 for T1-warp imaging and 3.6 cm2 6 1.5 for T1-hiBW imaging; P = .4). In the 15 additional patients, when the T1-warp sequence was compared with the T1-hiBW sequence with an identical matrix, the area of signal void at the proximal level was significantly smaller for T1-warp than for T1-hiBW imaging (18.5 cm2 6 4.0 for T1-warp images, 20.3 cm2 6 8.7 for T1-hiBW images; P = .028). There was no statistically significant difference however between the T1-weighted sequences at the distal level (4.1 cm2 6 2.1 for T1-warp images, 4.4 cm2 6 2.4 for T1hiBW images; P = .5). The extent of artifact reduction for the T1-weighted images in the additional 15 patients was similar to the main part of the study.

Qualitative Image Analysis Distinction of anatomic structures.— Distinction of all structures near the orthopedic hardware was significantly better for STIR-warp compared with STIR-hiBW imaging (P = .0001 for all structures) and for T1-warp compared with T1-hiBW imaging (P = .0001–.002 for different structures) (Table 1). For example, the gluteus minimus tendon had a mean rating of 1.90 and 1.93 (for reader 1 and 2, respectively) at STIRwarp imaging compared with 3.83 and

3.73 at STIR-hiBW imaging. Differences between the T1-warp and T1hiBW sequences were less pronounced than those between the two STIR sequences (Table 1). For all techniques, distinction of the hip capsule and the structures directly adjacent to the neck of the prosthesis was worse compared with the distinction of the abductor tendon insertions. For example, the anterior hip capsule and iliopsoas tendon had a mean rating of 2.78 and 2.95 (for readers 1 and 2, respectively) at T1warp imaging compared with 1.90 and 1.93 for the gluteus minimus tendon insertion. Interobserver agreement for distinction of anatomic structures was also good for both STIR-warp (k = 0.73) and STIR-hiBW (k = 0.71) imaging, and it was excellent for T1-warp (k = 0.89) and T1-hiBW (k = 0.84) imaging. Image quality.—Image distortion had a mean rating of 2.53 and 2.70 (for readers 1 and 2, respectively) at STIR-warp imaging, compared with a larger distortion of 4.38 and 4.35 at STIR-hiBW imaging (Table 2). Blurring was rated at a mean of 2.23 and 2.35 (for readers 1 and 2, respectively) with STIR-warp imaging compared with more blurring with 3.80 and 3.90 with STIR-hiBW imaging (Fig 6). Image noise was lower by using the STIR-warp sequence, with a mean of 2.10 and 2.08, compared with 3.15 and 3.10 by using the STIRhiBW sequence. Distortion, blurring, and image noise were all significantly better for STIR-warp compared with STIR-hiBW imaging (P = .0001 for all comparisons). Differences in image quality between the T1-warp and T1-hiBW techniques were similar to the differences encountered between the two STIR techniques (Table 2): Distortion, blurring, and image noise were all significantly better for T1-warp compared with T1-hiBW imaging (P = .0001 for all comparisons). Interobserver agreement for image quality was good for STIR-warp imaging (k = 0.63) and excellent for STIRhiBW imaging (k = 0.84). Interobserver agreement for image quality was good


Sutter et al

Figure 5

Figure 5: (a) Bar chart shows differences in metal artifact size. Coronal STIR-warp images have a significantly smaller area of signal void for the acetabular component and for the prosthesis neck and shaft and a significantly smaller area of insufficient inversion when compared with coronal STIR-hiBW images (P = .0001 for all three comparisons). Whiskers = standard deviation. (b) Bar chart shows differences in metal artifact size. Transverse T1-warp images have a significantly smaller area of signal void proximally (at the level of the acetabular component) compared with transverse T1-hiBW images (P = .0001). Regarding the area of signal void distally (at the level of the shaft), the two sequences are not different (P = .4). Whiskers = standard deviation.

for both T1-warp (k = 0.74), and T1hiBW (k = 0.75) imaging. Abnormal image findings.—The number of abnormal findings noted on STIR-warp images (105 findings) was significantly greater than the number of findings detected on STIR-hiBW images (55 findings; P = .0001), with more than half of the findings missed on STIR-hiBW images (Table 3). All abnormal imaging findings detected on 209


Sutter et al

Table 1 Distinction of Anatomic Structures Reader 1 Imaging Technique and Anatomic Structure STIR imaging Gluteus minimus attachment Gluteus medius attachment Inferior hip capsule Superior hip capsule T1-weighted imaging Gluteus minimus attachment Gluteus medius attachment Anterior hip capsule Posterior hip capsule

Reader 2

Warp Sequence

hiBW Sequence

P Value

Warp Sequence

hiBW Sequence

P Value

1.9 6 1.1 1.7 6 1.0 2.8 6 1.0 2.6 6 0.9

3.8 6 1.3 3.7 6 1.3 4.4 6 0.8 4.6 6 0.7

.0001 .0001 .0001 .0001

1.9 6 1.1 1.7 6 1.0 3.1 6 1.2 2.9 6 1.0

3.7 6 1.3 3.4 6 1.3 4.5 6 0.7 4.5 6 0.8

.0001 .0001 .0001 .0001

1.5 6 0.8 1.3 6 0.6 2.8 6 0.8 2.7 6 0.8

1.9 6 1.1 1.8 6 0.9 3.2 6 0.9 3.0 6 0.9

.002 .0001 .0001 .001

1.5 6 0.8 1.3 6 0.7 3.0 6 0.9 2.9 6 0.9

1.9 6 1.0 1.8 6 0.9 3.2 6 1.0 3.2 6 0.9

.002 .0001 .002 .002

Note.—Anatomic structures were assessed by two readers on a five-point scale from 1 (good depiction) to 5 (not visible). Data are mean 6 standard deviation. After Bonferroni correction, P , .0125 denotes statistical significance. hiBW = sequence optimized with high bandwidth (STIR-hiBW or T1-hiBW).

Table 2 Effect of Metal Artifacts on Image Quality Reader 1 Imaging Technique and Image Quality STIR imaging Distortion Blurring Image noise T1-weighted imaging Distortion Blurring Image noise

Reader 2

Warp Sequence

hiBW Sequence

P Value

Warp Sequence

hiBW Sequence

P Value

2.5 6 0.8 2.2 6 0.5 2.1 6 0.3

4.4 6 0.8 3.8 6 0.8 3.2 6 0.6

.0001 .0001 .0001

2.7 6 0.8 2.4 6 0.6 2.1 6 0.3

4.4 6 0.8 3.9 6 0.8 3.1 6 0.5

.0001 .0001 .0001

2.5 6 0.8 2.1 6 0.7 1.3 6 0.5

3.4 6 0.9 2.9 6 0.8 2.9 6 0.8

.0001 .0001 .0001

2.7 6 0.9 2.2 6 0.7 1.4 6 0.6

3.5 6 0.9 2.9 6 0.7 2.7 6 0.8

.0001 .0001 .0001

Note.—Distortion, blurring, and image noise were assessed by two readers on a five-point scale from 1 (no artifacts) to 5 (severe artifacts and nondiagnostic image). Data are mean 6 standard deviation. After Bonferroni correction, P , .0167 denotes statistical significance. hiBW = sequence optimized with high bandwidth (STIR-hiBW or T1-hiBW).

STIR-hiBW images were also noted on STIR-warp images. An area of presumed periprosthetic osteolysis was found in 16 patients on STIR-warp images but in only five patients on STIR-hiBW images (P = .001). Additionally, the number of patients in whom a joint effusion was detected was substantially greater on STIR-warp (23 patients) compared with STIR-hiBW (nine patients) images (P = .0001) (Fig 6). Tendinopathic changes of the gluteus minimus tendon were better seen at STIR-warp imaging (14 patients) compared with STIR-hiBW imaging (six patients) (P = .008). The difference in the number of abnormal imaging findings noted with the two T1-weighted sequences was small, 210

with 55 findings detected at T1-warp imaging and 50 findings detected at T1-hiBW imaging–this difference was not statistically significant (P = .06). All abnormal imaging findings detected on T1-hiBW images were also noted on T1-warp images. It is worth noting that small ossicles were seen only on the T1-warp and T1-hiBW images (four patients each) and were not detected on the STIR-warp and STIR-hiBW images. However, signs of a possible prosthesis loosening (periprosthetic osteolysis) (1,19) were only detected on the STIR sequence images (16 and five patients for STIR-warp and STIR-hiBW, respectively), while these findings were not seen on T1-warp and T1-hiBW images.

Additional T1-weighted imaging comparison.—In the 15 additional patients with THA, in whom the T1-warp sequence was compared with the T1hiBW sequence with an identical matrix, the qualitative analysis showed very similar results to those of the main part of the study, with a distinction of all structures near the orthopedic hardware that was significantly better for T1-warp compared with T1-hiBW imaging (P = .0001–.001 for different structures) (Table 4). Distortion, blurring, and image noise were significantly better for T1-warp compared with T1-hiBW imaging (P = .0001 for all comparisons) (Table 4). For both T1-weighted sequences, 18 abnormal imaging findings were detected (P . .99).



Figure 6

Figure 6: Images in 71-year old female patient with revision THA of the left hip. (a) Coronal STIR-hiBW image is nondiagnostic due to large areas of signal void, distortion, and blurring. (b) Coronal STIR-warp image shows signs of prosthesis loosening around the shaft (arrows) and joint effusion (∗). Note fixation wires (arrowheads) in the proximal femur on b that are not discernible on a.

Discussion The number of patients undergoing THA has increased by 50% in the past decade: In the United States, more than 400 000 THA procedures were performed in 2009, and the hip was the second most commonly replaced joint, after the knee (22). Different types of prosthesis-associated complications may occur in patients with THA, such as mechanical loosening of prosthesis components, periprosthetic osteolysis (also known as microparticle disease), infection, hardware failure, or periprosthetic fractures (1,19). While radiographs are usually able to show hardware failure and periprosthetic fractures, CT and MR imaging are more sensitive in depicting periprosthetic osteolysis than radiographs, with MR imaging being the most accurate modality for detection of lesions that are smaller than 3 cm in diameter (3). Although MR imaging is helpful in assessing the tendons around the hip joint and is also the most accurate imaging modality for

the detection of periprosthetic osteolysis and wear-induced synovitis (1,23), there are still distinct and detrimental artifacts in the bone and soft tissue adjacent to the metallic hardware that limit the use of MR imaging in patients with THA (4,23). Orthopedic hardware induces artifacts that occur both in the plane that is imaged (in-plane artifacts—ie, signal displaced within the plane) and in adjacent planes (through-plane artifacts— ie, signal displaced to other planes) (15). The predominant types of artifacts are signal loss or signal pile-up, displacement artifacts, and insufficient inversion (13). Signal loss is caused by spin dephasing due to large resonance frequency variations of the magnetic field caused by the metal implants but can also occur as a result of displacement artifacts (13). In the case of signal displacement, the signal is shifted away from its actual position, resulting either in areas of signal voids or in areas of signal pile-up (13). This signal displacement is caused by local changes of the


Sutter et al

magnetic field around metallic objects: Imaging gradients for frequency encoding and section selection are distorted by these local changes, and consequently image pixels are dislocated both in the frequency encoding direction within the image plane depicted and also into adjacent image planes (15,24). Fat-suppression techniques are much more susceptible to variations in the resonance frequency, which often results in an insufficient suppression of fat and water not only adjacent to the orthopedic implant, but also over large parts of the field of view (13). There are several strategies used for reduction of in-plane artifacts: Maximizing the bandwidth during signal readout reduces the displacement of image pixels in the readout direction (24). This approach is limited by imager hardware (maximum readout gradient amplitude) and reduces the signal-to-noise ratio efficiency (24). VAT applies a replay of the sectionselection gradient during signal readout, which results in a cancelation of the signal displacement in readout direction for off-resonance spins (10). However, VAT is not able to remove the distortion of the section profile perpendicular to the image plane. Moreover, VAT can induce some blurring of the image caused by two distinct effects: The first effect is the geometrical tilting of the image pixels (10), and the second effect is a low pass filter superimposing on the signal readout (11). The blurring due to the low pass filtering effect results in a general loss of image detail even if the nominal resolution is unchanged, whereas the tilting effect is most visible at the edges of anatomic structures and is more pronounced when the section thickness is increased (11,15,25). In our study, blurring effects due to the low pass filtering effect of the readout were minimized by using short, high-bandwidth readouts (11), whereas blurring of edges remained as an inherent effect of the tilted section profile, even though this effect was not very prominent for the section thickness we used. Fat suppression is a particular challenge in patients with metal implants. Conventional spectral fat suppression 211


Sutter et al

Table 3 Number of Abnormal Imaging Findings Abnormal Finding Joint effusion Gluteus minimus tear Gluteus medius tear Gluteus minimus tendinopathy Gluteus medius tendinopathy Trochanteric bursitis Periprosthetic osteolysis Soft tissue edema Bone marrow edema Synovitis Small ossicles Displaced bone avulsion Other All abnormal imaging findings

STIR-warp

STIR-hiBW

P Value

T1-warp

T1-hiBW

P Value

23 7 8 14 8 11 16 4 4 5 0 2 3 105

9 4 7 6 6 9 5 2 3 3 0 1 0 55

.0001* .25 ..99 .008* .5 .5 .001* .5 ..99 .5 NA ..99 NA .0001*

3 7 7 13 9 4 0 0 0 0 4 2 6 55

3 4 6 13 9 4 0 0 0 0 4 2 5 50

..99 .25 ..99 ..99 ..99 ..99 NA NA NA NA ..99 ..99 ..99 .06

Note.—NA = not applicable. * Indicates a statistically significant difference (P , .05).

Table 4 Comparison of T1-weighted Sequences in 15 Additional Patients Reader 1 Finding Anatomic structure Gluteus minimus attachment Gluteus medius attachment Inferior hip capsule Superior hip capsule Image quality Distortion Blurring Image noise

Reader 2

T1-warp

T1-hiBW

P Value

T1-warp

T1-hiBW

P Value

1.7 6 0.6 1.6 6 0.6 2.9 6 0.5 2.9 6 0.6

2.8 6 1.0 2.5 6 0.9 3.9 6 0.5 3.9 6 0.5

.001 .001 .0001 .0001

1.8 6 0.7 1.6 6 0.6 3.0 6 0.7 3.1 6 0.6

2.7 6 1.0 2.5 6 0.8 3.9 6 0.6 4.1 6 0.5

.001 .001 .0001 .0001

2.9 6 0.8 2.3 6 0.5 1.7 6 0.5

3.8 6 0.8 3.5 6 0.5 3.7 6 0.6

.0001 .0001 .0001

2.9 6 0.8 2.3 6 0.6 1.7 6 0.5

3.8 6 0.8 3.5 6 0.7 3.5 6 0.7

.0001 .0001 .0001

Note.—Anatomic structures were assessed by two readers on a five-point scale from 1 (good depiction) to 5 (not visible). Distortion, blurring, and image noise were assessed by two readers on a five-point scale from 1 (no artifacts) to 5 (severe artifacts and nondiagnostic image). Data are mean 6 standard deviation. After Bonferroni correction, P , .0125 (anatomic structures) and P , .0167 (image quality) denote statistical significance.

fails in those areas of the image for which the local field distortion exceeds the spectral separation between fat and water signal of approximately 3.5 ppm (13). In this situation, the STIR technique is the preferred fat-suppression method in patients with metal implants (26,27) since it is much less sensitive to magnetic field variations and thus results in an improved and more homogeneous suppression of fat and water near the metal structure (13). However, since water signal is attenuated by the inversion pulse, STIR TSE images 212

provide a low signal-to-noise ratio compared with TSE images with spectral fat suppression. Furthermore, the STIR contrast may also be impaired by local field distortions if there is a mismatch between the inverted region and the region being imaged—that is, if the section profile of the inversion pulse is distorted in a different manner than the profile of the acquired section. This is avoided by using equal section-selection gradient amplitudes both for the inversion and the excitation radiofrequency pulses within the sequence (13).

The SEMAC technique is based on a spin-echo sequence with an additional phase encoding along the section-select axis (15). After Fourier transformation of the phase-encoded sections, the signal distortion in the direction perpendicular to the imaging plane is resolved. In a subsequent postprocessing step this information is used to shift dislocated signal back to its original section position, resulting in a “through-plane” distortion correction (15,17). Together with the VAT technique and the associated reduction of in-plane distortion,



a technique results that features both in-plane and through-plane distortion correction (15). Regions with strong local distortions may have additional local blurring caused by SEMAC postprocessing. There is some increase of image noise with the SEMAC technique that is caused by the use of fast imaging techniques, such as partial Fourier, parallel imaging, and high-readout bandwidth, as well as by image postprocessing when combining multiple phase-encoded sections into a single section position (15). Another technique that reduces through-plane artifacts is multiple-acquisition with variable resonance image combination, or MAVRIC. Here, non–spatially-selective three-dimensional spin-echo acquisitions are acquired at multiple overlapping frequencies (28). MAVRIC is purely frequency-selective, whereas SEMAC is spatially-selective but with spatial shifts due to frequency. A major drawback of MAVRIC is caused by the nonselective excitation leading to long acquisition times (28). Recently, a combination of MAVRIC with SEMAC has been shown as a valuable technique in phantoms and single patients (16). While we were able to show that the warp sequence (combining SEMAC, VAT, and increased bandwidth) substantially improves the distinction of anatomic details, the improvement was much more pronounced for the STIR images than for the T1-weighted images. This is primarily because the STIR-warp sequence also benefits from the matching of the inversion and excitation radiofrequency pulse properties, as described above. Our study had limitations: When comparing the T1-warp sequence and the T1-hiBW sequence in the main part of the study, a different matrix was used due to the technical limitations described in the Materials and Methods section. Therefore, we performed MR examinations in 15 additional patients in whom T1-warp and T1-hiBW sequences were compared with an identical matrix, and all results of this additional analysis were very similar to those of the main analysis. Consequently, we think that the different matrix for the

T1-weighted sequences in the main part of the study did not lead to a relevant bias. Another limitation is the lack of a reference standard for the abnormal imaging findings. As only a small number of patients underwent surgery or additional CT after the MR examination, we cannot correlate the abnormal findings to a reference standard to exclude possible false-positive findings. Further, in five cases the time period between the quantitative measurements and the qualitative analysis was only 2 weeks, which might have resulted in a possible recall bias. However, as the time period was longer in most patients, we believe that there was fairly minimal recall bias in our study. In conclusion, the SEMAC and VAT techniques, as provided by means of the warp sequence for this study, are feasible for the assessment of the hip and the adjacent soft tissues in patients with THA and achieve a substantial reduction of metal artifacts when compared with the currently optimized MR sequences. The reduction in signal void size with the warp sequence was most pronounced for the acetabular component. STIR-warp and T1-warp imaging were statistically significantly better for quantitative and qualitative image criteria, but a clinically relevant artifact reduction was only present for STIR images and not for T1-weighted images. Disclosures of Potential Conflicts of Interest: R.S. No potential conflicts of interest to disclose. E.J.U. No potential conflicts of interest to disclose. V.J. Financial activities related to the present article: author is an employee of Siemens and is involved in development of software used in the study. Financial activities not related to the present article: none to disclose. Other relationships: none to disclose. M.N. Financial activities related to the present article: author is an employee of Siemens and is involved in development of software used in the study. Financial activities not related to the present article: none to disclose. Other relationships: none to disclose. C.W.A.P. No potential conflicts of interest to disclose.

References 1. Potter HG, Nestor BJ, Sofka CM, Ho ST, Peters LE, Salvati EA. Magnetic resonance imaging after total hip arthroplasty: evaluation of periprosthetic soft tissue. J Bone Joint Surg Am 2004;86-A(9):1947–1954.


Sutter et al

2. White LM, Kim JK, Mehta M, et al. Complications of total hip arthroplasty: MR imaging-initial experience. Radiology 2000;215(1):254–262. 3. Walde TA, Weiland DE, Leung SB, et al. Comparison of CT, MRI, and radiographs in assessing pelvic osteolysis: a cadaveric study. Clin Orthop Relat Res 2005;(437):138–144. 4. Sugimoto H, Hirose I, Miyaoka E, et al. Low-field-strength MR imaging of failed hip arthroplasty: association of femoral periprosthetic signal intensity with radiographic, surgical, and pathologic findings. Radiology 2003;229(3):718–723. 5. Pfirrmann CW, Notzli HP, Dora C, Hodler J, Zanetti M. Abductor tendons and muscles assessed at MR imaging after total hip arthroplasty in asymptomatic and symptomatic patients. Radiology 2005;235(3):969– 976. 6. Springer I, Müller M, Hamm B, Dewey M. Intra- and interobserver variability of magnetic resonance imaging for quantitative assessment of abductor and external rotator muscle changes after total hip arthroplasty. Eur J Radiol 2011;81(5):928–933. 7. Schenck JF. The role of magnetic susceptibility in magnetic resonance imaging: MRI magnetic compatibility of the first and second kinds. Med Phys 1996;23(6):815–850. 8. Blankenbaker DG, De Smet AA, Vanderby R, McCabe RP, Koplin SA. MRI of acute bone bruises: timing of the appearance of findings in a swine model. AJR Am J Roentgenol 2008;190(1):W1–W7. 9. Ohishi T, Ito T, Suzuki D, Banno T, Honda Y. Occult hip and pelvic fractures and accompanying muscle injuries around the hip. Arch Orthop Trauma Surg 2011;132(1):105– 112. 10. Cho ZH, Kim DJ, Kim YK. Total inhomogeneity correction including chemical shifts and susceptibility by view angle tilting. Med Phys 1988;15(1):7–11. 11. Butts K, Pauly JM, Gold GE. Reduction of blurring in view angle tilting MRI. Magn Reson Med 2005;53(2):418–424. 12. Kolind SH, MacKay AL, Munk PL, Xiang QS. Quantitative evaluation of metal artifact reduction techniques. J Magn Reson Imaging 2004;20(3):487–495. 13. Hargreaves BA, Worters PW, Pauly KB, Pauly JM, Koch KM, Gold GE. Metal-induced artifacts in MRI. AJR Am J Roentgenol 2011;197(3):547–555. 14. Lu W, Pauly KB, Gold GE, Pauly JM, Hargreaves BA. Slice encoding for metal artifact correction with noise reduction. Magn Reson Med 2011;65(5):1352–1357.

213


15. Lu W, Pauly KB, Gold GE, Pauly JM, Hargreaves BA. SEMAC: Slice Encoding for Metal Artifact Correction in MRI. Magn Reson Med 2009;62(1):66–76. 16. Koch KM, Brau AC, Chen W, et al. Imaging near metal with a MAVRIC-SEMAC hybrid. Magn Reson Med 2011;65(1):71–82. 17. 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(5):1121–1127. 18. Roemer FW, Guermazi A, Lynch JA, et al. Short tau inversion recovery and proton density-weighted fat suppressed sequences for the evaluation of osteoarthritis of the knee with a 1.0 T dedicated extremity MRI: development of a time-efficient sequence protocol. Eur Radiol 2005;15(5): 978–987. 19. Toms AP, Marshall TJ, Cahir J, et al. MRI of early symptomatic metal-on-metal total hip arthroplasty: a retrospective review of

214

radiological findings in 20 hips. Clin Radiol 2008;63(1):49–58. 20. Anderson H, Toms AP, Cahir JG, Goodwin RW, Wimhurst J, Nolan JF. Grading the severity of soft tissue changes associated with metal-on-metal hip replacements: reliability of an MR grading system. Skeletal Radiol 2011;40(3):303–307. 21. Landis JR, Koch GG. An application of hierarchical kappa-type statistics in the assessment of majority agreement among multiple observers. Biometrics 1977;33(2):363–374. 22. Agency for Healthcare Research and Quality. HCUPnet HCaUP. National statistics: Hip replacement 2009 and 1999. Rockville, MD: Agency for Healthcare Research and Quality. http://hcupnet.ahrq.gov. Accessed September 16, 2011. 23. Weiland DE, Walde TA, Leung SB, et al. Magnetic resonance imaging in the evaluation of periprosthetic acetabular osteolysis: a cadaveric study. J Orthop Res 2005;23(4):713–719.

Sutter et al

24. Lee MJ, Kim S, Lee SA, et al. Overcoming artifacts from metallic orthopedic implants at highfield-strength MR imaging and multi-detector CT. RadioGraphics 2007;27(3):791–803. 25. Chang SD, Lee MJ, Munk PL, Janzen DL, MacKay A, Xiang QS. MRI of spinal hardware: comparison of conventional T1weighted sequence with a new metal artifact reduction sequence. Skeletal Radiol 2001;30(4):213–218. 26. 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(3):251–254. 27. Czerny C, Krestan C, Imhof H, Trattnig S. Magnetic resonance imaging of the postoperative hip. Top Magn Reson Imaging 1999;10(4):214–220. 28. Koch KM, Lorbiecki JE, Hinks RS, King KF. A multispectral three-dimensional acquisition technique for imaging near metal implants. Magn Reson Med 2009;61(2):381–390.
