Skeletal Radiol DOI 10.1007/s00256-014-1818-5
SCIENTIFIC ARTICLE
Hip imaging of avascular necrosis at 7 Tesla compared with 3 Tesla J. M. Theysohn & O. Kraff & N. Theysohn & S. Orzada & S. Landgraeber & M. E. Ladd & T. C. Lauenstein
Received: 23 October 2013 / Revised: 30 December 2013 / Accepted: 2 January 2014 # ISS 2014
Abstract Objectives To compare ultra-high field, high-resolution bilateral magnetic resonance imaging (MRI) of the hips at 7 Tesla (T) with 3 T MRI in patients with avascular necrosis (AVN) of the femoral head by subjective image evaluations, contrast measurements, and evaluation of the appearance of imaging abnormalities. Materials and Methods Thirteen subjects with avascular necrosis treated using advanced core decompression underwent MRI at both 7 T and 3 T. Sequence parameters as well as resolution were kept identical for both field strengths. All MR images (MEDIC, DESS, PD/T2w TSE, T1w TSE, and STIR) were evaluated by two radiologists with regard to subjective image quality, soft tissue contrasts, B1 homogeneity (fourpoint scale, higher values indicating better image quality) and depiction of imaging abnormalities of the femoral heads (three-point scale, higher values indicating the superiority of 7 T). Contrast ratios of soft tissues were calculated and compared with subjective data.
J. M. Theysohn (*) : O. Kraff : N. Theysohn : S. Orzada : M. E. Ladd : T. C. Lauenstein Erwin L. Hahn Institute for Magnetic Resonance Imaging, University Duisburg-Essen, Essen, Germany e-mail: [email protected] J. M. Theysohn : O. Kraff : N. Theysohn : S. Orzada : T. C. Lauenstein Institute for Diagnostic and Interventional Radiology and Neuroradiology, University Hospital Essen, Essen, Germany S. Landgraeber Department of Orthopedic Surgery, University Hospital Essen, Essen, Germany M. E. Ladd Department of Medical Physics in Radiology, German Cancer Research Center, Heidelberg, Germany
Results 7-T imaging of the femoral joints, as well as 3-T imaging, achieved “good” to “very good” quality in all sequences. 7 T showed significantly higher soft tissue contrasts for T2w and MEDIC compared with 3 T (cartilage/fluid: 2.9 vs 2.2 and 3.6 vs 2.6), better detailed resolution for cartilage defects (PDw, T2w, T1w, MEDIC, DESS>2.5) and better visibility of joint effusions (MEDIC 2.6; PDw/T2w 2.4; DESS 2.2). Image homogeneity compared with 3 T (3.9–4.0 for all sequences) was degraded, especially in TSE sequences at 7 T through signal variations (7 T: 2.1–2.9); to a lesser extent also GRE sequences (7 T: 2.9–3.5). Imaging findings related to untreated or treated AVN were better delineated at 3 T (≤1.8), while joint effusions (2.2–2.6) and cartilage defects (2.5–3.0) were better visualized at 7 T. STIR performed much more poorly at 7 T, generating large contrast variations (1.5). Conclusions 7-T hip MRI showed comparable results in hip joint imaging compared with 3 T with slight advantages in contrast detail (cartilage defects) and fluid detection at 7 T when accepting image degradation medially. Keywords Ultra-high field . UHF . Hip . MRI . Avascular necrosis . AVN
Introduction Magnetic resonance imaging (MRI) is the gold standard or has been suggested to be the imaging test of choice in the diagnostic evaluation of hip joint abnormalities. This is currently realized using MR systems at field strengths of 1.5 or 3 Tesla (T), generating high-resolution images with strong soft-tissue contrasts [1]. While ultra-high field (UHF) MRI systems for human use at 10.5 T (whole-body) and 11.7 T are about to start running early in 2014, approximately 40 UHF systems at 7 T and 9.4 T (ratio ~ 6:1) have been installed worldwide. Active research at 7 T and 9.4 T has been performed for more than 10
Skeletal Radiol
years, with the main focus on the brain and, lagging behind, the musculoskeletal system using smaller fields-of-view (FOV) [2–6]. Compared with lower field strengths, their implementation for fundamental research in volunteers (i.e., functional MRI) and patient imaging has revealed improved diagnostic capabilities in the brain and joints. The roughly twofold or fourfold increased signal-to-noise ratio (SNR) (over 3 T or 1.5 T) was transformed into stronger blood oxygenation level-dependent (BOLD) contrast, higher image resolution, and stronger image contrast. Transferring this new technology to the torso requiring the use of larger FOVs is a rocky road challenged by strong image degradations [7]. Radiofrequency (RF) field distortions in body imaging with large FOVs beginning to appear at 3 T are much more enhanced at 7 T [8, 9]. This has led to the delay in useful thoracic, abdominal, and pelvic imaging at 7 T. Feasibility studies have emerged only over the last few years, after selfbuilt multi-channel transmit body coils had been established [10–16]. Only a small number of comparison studies of 7 T with lower field strengths have been published on imaging these regions, including validation against 3 T in the wrist, ankle, breast, and the heart [17–21]. However, none of these studies included patients, or an analysis of imaging abnormalities. Avascular necrosis (AVN) of the hip is one endpoint of bone destruction due to an altered microcirculation, subchondral oxygen deficiency and necrosis [22–24]. While AVN mostly occurs in middle-aged males it can lead to severe degenerative disease of the hip joint, possibly requiring endoprosthetic hip replacement (10 % of all are due to AVN) [25–27]. Efforts to treat the early stages of AVN that are not manageable using conservative pain strategies include bone decompression [28]. Recently, this procedure has been supplemented with different biological bone graft injections used to fill the decompressed and carved out necrotic lesion [29, 30]. In the event of treated or untreated AVN, MRI can visualize aspects of the necrotic lesion (necrosis, reactive zone/granulation tissue, sclerotic changes, edema), as well as postoperative changes (edema) and the drilling hole or injected bone substitute (bone graft). Technical progress in imaging has enhanced sensitivity for the detection of subchondral fractures in the later stages of the disease. However, sensitivity compared with computed tomography (CT) has been shown only to be fair to moderate [31]. Conspicuity in MRI is dependent on a fine subchondral fluid collection at the location of the bony depression or hypointense line on T1weighted (T1w) images [32, 33]. The reactive border is noticeable as a “double line sign” with an outer sclerotic band (T1w and T2w hypointense) and adjacent T2w hyperintensity representing granulation tissue [34–37]. Comparison studies of 7 T with lower field strengths, preferably 3 T in musculoskeletal imaging, are warranted, including patient data. We hope that the diagnostic advantage
of 7 T, as has been shown for peripheral joints with small cross-sections (i.e., the knee) [6], can also be transferred to the hip joints. The aims of this study were twofold: to compare imaging results of the hips at 7 T with those at 3 T using subjective evaluations and contrast measurements, and to analyze and compare the appearance of imaging abnormalities in patients with AVN of the femoral head between the field strengths.
Materials and methods Subjects and hardware After gaining approval from the local institutional ethics committee and written informed consent from all subjects, 13 patients aged 30 to 67 years (mean age 51.4 years, SD 9.4 years; 4 female, 9 male) were included in this study. All patients had a known history of AVN of one or both femoral heads treated with advanced core decompression. During this intervention a surgeon drills a hole through the femoral neck up to the necrotic area in the femoral head, removes part of the necrotic material and refills the cavity and drilling hole with a bone graft substitute [30]. Patients were imaged between 8 weeks and 16 months after surgery. MRI examinations were performed on two different systems not more than 2 weeks apart. A 7-T whole-body scanner (Magnetom 7 T, Siemens Healthcare, Erlangen, Germany) was used in combination with a custom-built eight-channel transmit/receive coil, since neither an integrated transmit body coil as is known from 1.5T or 3-T systems, nor commercially procurable RF coils for ultra-high-field abdominal MRI are available. The RF coil allowed static transmit phase shimming to mitigate signal voids medially out of the region of interest [38]. Therefore, each coil element is connected via a modulator box containing amplitude and phase shifters to individual 1-kW modules of a RF power amplifier (RFPA) situated in the equipment room. A total cable length of more than 20 m resulted in an attenuation of −3 dB or 50 % between RFPA and coil plug, limiting the available peak RF power to approximately 4 kW in total. At 3 T, a state-of-the-art clinical MRI system (Skyra, Siemens Healthcare, Erlangen, Germany) was utilized, paired with a scanner-integrated body transmit coil and receive coils as provided by the vendor (18-channel body matrix coil combined with 16-channel spine array). Here, the available peak RF power amounted to 37.5 kW, with the RFPA situated directly at the magnet for improved performance. The multichannel coils at both field strengths allowed for parallel imaging techniques, allowing shorter acquisition times compared with single-channel coils (generalized autocalibrating partially parallel acquisitions [GRAPPA], acceleration factor of R=2).
Skeletal Radiol
MR imaging An imaging protocol at 7 T had been evaluated in an earlier study [14], adapted from sequences used to image the knee joint with a local volume coil [39] with the intention of investing additional SNR into higher resolution. Table 1 provides an overview of all sequence parameters. Images were acquired in coronal orientation: proton density (PD)/T2w turbo spin echo (TSE), T1w TSE, multi-echo data image combination (MEDIC), double echo steady-state (DESS), and short-tau-inversion-recovery (STIR). For the TSE sequences, prolonged 180° RF pulses (7.68 ms) [40] were implemented to keep the total amount of power within the SAR limits, while playing out the high amplitude pulses needed for pelvic imaging. Owing to conservative SAR limits at 7 T (10 W/kg local SAR, as defined for normal operating mode by IEC guidelines), all TSE sequences were nevertheless limited to 4–5 slices per acquisition, requiring multiple repetitions for full femoral head coverage. The same sequence parameters were used at 3 T, resulting in, for example, identical receiver bandwidths at both field strengths (only minimal adjustments owing to technical limitations were allowed, but no further optimization was performed). With this choice the chemical shift displacement was less at 3 T compared with 7 T and yielded a better delineation of anatomical structures at 3 T. On the other hand a change in receiver bandwidth would have resulted in additional SNR variations, thereby challenging the comparability. Image evaluation All MR images were reviewed using a PACS (picture archive and communication system) workstation (Centricity; General
Electric Healthcare, Barrington, IL, USA) and were assessed by two radiologists in consensus (8 and 6 years of experience in musculoskeletal MRI). Since field-strength was obviously identified through the medial signal voids and change in tissue contrast on 7-T images, blinded reading was not possible. However, radiologists consciously aimed not to overrate 7-T MRI. Qualitative properties (resolution, homogeneity, and soft tissue contrasts among cartilage, fluid, and bone) of each image set of each examination were rated on a four-point scale, looking only at the hip joint (1 = poor, 2 = fair/only partially diagnostic, 3 = good/diagnostic, 4 = excellent image quality) and compared them at the two field strengths. In STIR images the characteristic contrast through fat suppression was graded in a similar fashion. Homogeneity was only evaluated directly around the joint, accepting technically inevitable, medially located signal voids. We performed quantitative signal measurements of soft tissues (cartilage, labrum, fluid, and bone) and calculated contrast ratios: CR=(SIA −SIB)/(SIA +SIB). The signal intensities were measured drawing a free-hand ROI over the tissue of interest after magnifying the selected structure. ROIs for cartilage and bone were placed on the same anatomical site in most patients: lateral to the insertion of the round ligament of the femur. Only if this cartilage was thinned out (n=3) and not measurable was the medial femoral head chosen. Locations at both field strengths were always identical. Again these data was compared at the two field strengths and correlated with subjective results. Depiction of postoperative changes after advanced core decompression of AVN (treated lesion, bone marrow edema) or MR abnormalities of the hip joints not directly related to the surgical procedure (cartilage defect, joint effusion, untreated
Table 1 Sequence protocol for 7 T and identicala 3 T
Scan timeb (min:s) Acquisition matrix Resolution, interpolated (mm2) Resolution, acquired (mm2) Slice thickness (mm) Flip angle (°) Echo timeb (ms) Repetition timeb (ms) Turbo factor
MEDIC
DESS
PD/T2
STIR
T1
05:33 512×512 0.38 0.76 1.5 30° 15 1,000
08:20 512×512 0.38 0.76 0.76 15° 4.1 11
03:26c 512×512 0.38 0.76 3 180° 29/88 4,350 7
03:32c 384×384 0.51 1.02 3 180° 32 5,000 5
01:35c 512×512 0.38 0.76 2 180° 8.3 900 5
All imaging was performed using a field of view (FOV) of 390 mm×390 mm and, except for T1w, with an acceleration factor of GRAPPA=2 a
After transferring sequences from 7 T to 3 T only minimal adjustments were allowed owing to technical limitations, but no further optimization was performed
b
Scan times at 3 T varied ± 15 s owing to slightly different permitted echo and repetition times
c
Sequences were repeated at 7 T to cover the femoral heads. Limitations in the allowed number of slices owing to SAR restrictions made this necessary
Skeletal Radiol
lesion) were evaluated with regard to which field strength performs better (1: 7 T inferior, 2: 7 T equal to 3 T, and 3: 7 T superior). Sequences were only evaluated for depiction of MR abnormalities, if visible at at least one field-strength (i.e., if no edema was present, it was not rated). All quantitative and qualitative measurements are given as mean and standard deviation. Statistical significance testing was performed using a two-sided Wilcoxon signed-rank test (with p3.0; Figs. 1, 2), with only 7-T T2w lagging behind in the field comparison (3.2), compared with 3 T (3.6). Throughout all sequences 3 T provided greater homogeneity of the hip joints (3.9-4.0) compared with 7 T (Figs. 3, 4), where, despite shifting signal voids medially, residual B1 heterogeneities remained visible overlying the joints (2.1–3.5). TSE sequences at 7 T (2.1–2.9) were much more degraded than gradient echo imaging (2.9–3.5). Looking at cartilage/fluid contrast 7 T was graded superiorly in MEDIC (3.6 vs 2.6; Fig. 5a, b), on T2w (2.9 vs 2.2), and STIR (2.0 vs 1.0), while 3 T performed slightly better in PDw (2.9 vs 2.5) and DESS (2.8 vs 2.6). For cartilage/bone contrast only T2w generated noteworthy advantages of 7 T (3.0 vs 2.2). Finally, the STIR contrast, or fat suppression respectively, at 7 T (2.1) could not compete with 3 T (4.0; Figs. 2, 4). The comparison of measured contrast ratios proved equal for both field strengths in the majority of cases (Table 3). While 3 T performed better for the cartilage–fluid contrast for T2w imaging (0.66 vs 0.57), MEDIC showed a minor advantage at 7 T (0.32 vs 0.20). While labrum–fluid contrast had lower values at 7 T for T2w (0.49 vs 0.75), a dramatic difference in T1w images was apparent with only 3 T showing visible
identical resolution, slight subjective advantages in detail resolution at 7 T in a, e MEDIC, b, f DESS, c, g PDw, and d, h T2w; c, d only on the 7-T TSE sequences can a faint signal loss toward the edge of the image (star) be seen
Skeletal Radiol
Fig. 2 Coronal views of the left hip joint in a 44-year-old male patient suffering from AVN 3 months post-ACD of the left femoral head (see also Figs. 4 , 5c–e). While using identical resolutions at 7 T (top row), images appear slightly crisper and with somewhat stronger areas of contrast compared with 3 T (bottom row) in a, e MEDIC, b, f PDw, c, g T2w,
and d, h STIR. On the 7-T TSE sequences (b–d) a residual signal loss toward the edge of the image (star) can be seen and signal variations cause slight contrast variations, making 3 T better in the depiction of the treated lesion (f, g) and surrounding postoperative changes (h). Arrow in c: drilling hole refilled with a bone graft material
contrast (0.46 vs 0.04); PDw, DESS, and MEDIC were of equal value at the two field strengths. Finally, cartilage–bone contrasts had mainly superior 7 T values (PDw 0.67 vs 0.41, T2w 0.59 vs 0.14, DESS 0.81 vs 0.49), with only 3-T MEDIC proving slightly advantageous (0.81 vs 0.69). When subjective and measured contrast areas are being held side by side many parallelisms can be acknowledged. Measured cartilage–fluid contrast for MEDIC was superior at 7 T (0.32 vs 0.20) with a good correlation in the subjective data (3.6 vs 2.6). While DESS and PDw agree with comparable results, a slight contrast advantage for cartilage–fluid in T2w exists at 3 T (0.66 vs 0.57); 7 T dominates in the subjective data (2.9 vs 2.2), but showing a fairly high standard deviation (1.0). A much stronger measurement of cartilage–bone contrast for T2w at 7 T (0.59 vs 0.14) agrees with the subjective impression (3.0 vs 2.2). Smaller measured 7 T advantages for PDw (0.67 vs 0.41) and DESS (0.81 vs 0.49), or a smaller disadvantage for MEDIC (0.69 vs 0.81) are not mirrored in the subjective evaluations, where field strengths perform to a comparably high standard (3.0 vs 3.1; 3.8 vs 3.6; 3.8 vs 3.6). Imaging findings (necrosis, reactive zone/granulation tissue, sclerotic changes, edema, bone graft) related to untreated or treated AVN are better delineated at 3 T with values ≤1.8 (Table 4) (Figs. 2–4); this includes the lack of reasonable depiction of bone marrow edema at 7 T on STIR images (1.3; Figs. 2d, 4d, h) and changing contrasts in TSE sequences close to the medial signal void (Fig. 3b–d). Joint effusions were
better visualized at 7 T on MEDIC (2.6), PDw (2.4), T2w (2.4), and DESS (2.2; Fig. 2); but 7-T STIR performed much more poorly compared with 3 T, generating large variations in image contrast (1.5). Finally, for cartilage defects 7 T was superior to 3 T on MEDIC (3.0), DESS (2.8), T2w (2.6), T1w (2.6), and PDw (2.5; Fig. 5).
Discussion With this comparative study it was demonstrated that hip imaging at 7 T essentially meets standards set by 3 T, looking at the joints themselves. This is particularly interesting as, from a technical point of view, there are major differences in the RF chain between the two MR systems. Not only was the number of receive elements underrepresented at 7 T compared with 3 T by nearly a factor of 3, the available peak RF power also differs significantly (37.5 kW “DirectRF” at 3 T vs 4 kW at 7 T). Nordmeyer-Massner et al. have used an identical GRE sequence and identical receive coils, proving 0- (bone) to 2fold (cartilage, nerves) SNR gains at 7 T compared with 3 T depending on the tissue [18]. While larger signal differences in tissues can translate into stronger contrast, as has been shown by our data, they could not improve the visibility of anatomical structures. In patients, it was found that 7 T may display cartilage defects more accurately, using identical sequences at both field strengths, while a comparable depiction
Skeletal Radiol
Fig. 3 Coronal views of the pelvis in a 41-year-old male patient suffering from AVN of both femoral heads 4 weeks post-ACD of the left femoral head. Comparison of T1w images at a 3 T and b 7 T visualize relatively strong residual signal voids overlying the right hip joint (star in b)— compare Fig. 4 for the good shimming results mostly seen in our population. Signal alteration can not only reduce signal medially (star in b–d), but can cause contrast to vary across the image: in c 7-T PDw and d 7-T T2w the border zone of the right-sided, untreated necrotic lesion changes between hyper- and hypointense signal intensity (black arrows). The reactive border zone between bone graft material and surrounding bone can be appreciated as a hyperintense line (arrowhead) in c (PDw) and d (T2w)
of intraosseous imaging abnormalities due to AVN or postsurgical changes is within reach. Lack of decent bone marrow edema contrast at 7 T is still one of the biggest challenges because of the need for T2-weighted spin echo sequences. Since images at both field strengths were acquired with identical (high) resolutions, it is not surprising that subjective resolution evaluation issued high values for all sequences at both field strengths. The disadvantage of medial image degradation at 7 T has expectedly lower grading in all 7-T sequences (spin echo >>gradient echo) because 7 T suffers from B1 heterogeneities and a limited penetration depth, which are not truly visible at 3 T. Such interferences arise when an object is excited by waves whose length is shorter
than the object’s diameter: wave length in tissue is about 15 cm at 7 T and 30 cm at 3 T [41]. This explains why the problem has been identified and described at 3 T, but because of its low degree of severity has not truly required its extinction [42, 43]. STIR imaging yielded lowest image quality at 7 T as this sequence depends on two consecutive high flip angle pulses for the inversion and refocusing pulses. Coping with challenges at 7 T will further improve 3-T imaging results. Mastering these obstacles includes the use of the relatively new multichannel transmit/receive technology at 7 T with possible modulation of B1 heterogeneities, which was used here and has been illustrated in publications on hip imaging [11, 13, 14]. Juras et al. used optimized sequences in ten healthy volunteers at 3 T and 7 T to compare imaging performance in the ankle [17]. While they observed a CNR increase for PDw TSE and most 3D T1w GRE images (but not T1w TSE), our results in patients showed more heterogeneity. However, taking into account differences in cross-sections and thus the aforementioned challenges in B1 homogeneity and penetration depth, this difference may be less surprising. Furthermore, Juras et al. used an different RF coil setting for their comparison, namely a commercially available 28-channel coil at 7 T versus an eight-channel coil at 3 T. In our study, the superiority of 7-T MEDIC for cartilage–fluid contrast in qualitative and quantitative data (T2w only in the qualitative data) can be explained by higher intrinsic SNR and the fact that MEDIC suffers less from image degradation at 7 T than TSE sequences, owing to its nature as a low flip angle gradient echo sequence. This contrast enhancement may be translated to clinically used spin echo sequences as engineering process advances at 7 T. Other sequences (PDw, T2w, DESS) do not show the domination of one field strength, putting 7 T on a par with 3 T. Identical reasoning can be alleged for stronger cartilage–bone contrast at 7 T, mainly for T2w and to a lesser extend PDw and DESS, with intrinsically higher SNR producing stronger contrasts between fluid-rich tissue (cartilage) and low-signal tissue (bone). Lastly, while measured labrum–fluid contrast at 7 T showed equal values for most sequences, a slight contrast reduction at 7 T might be due to our population mainly having degenerated joints with calcifications of the labrum; stronger susceptibility effects at 7 T could account for this signal reduction. The fact that, in our study, 7 T mostly performs with even quality compared with 3 T might be attributable to the early stage of 7-T prototype coil technology, while 3-T coils are commercially available and well optimized. This lagging behind in coil development may be a reason for losing some of the potential of the theoretically more than twofold increase in SNR. Deligianni et al. have shown that a dedicated high-resolution 3D GRE sequence could be well-integrated into 3 T as well as 7 T for musculoskeletal imaging [20]. Inventing new sequences as well as further optimizing 7-T sequences in our study, aiming to exploit the full potential,
Skeletal Radiol
Fig. 4 Coronal views of the pelvis in a 44-year-old male patient known from Fig. 2 (see also Fig. 5c–e). High-resolution a, b T1w and c, d STIR images of both hip joints at 7 T (left) and 3 T (right). 7-T images represent the typical localization of the medially shifted signal void (star) after B1 shimming with good depiction of both femoral heads. On T1w images one can appreciate the better delineation of the joint effusion (white
circles) in a at 7 T while in b 3 T generates almost no contrast to the surrounding muscle tissue. Quality of fat suppression and contrast homogeneity on STIR images in c at 7 T cannot compete with d 3 T so far. Black arrow: drilling hole refilled with a bone graft material, surrounded by postoperative changes
could further facilitate finding clinically useful advantages at 7 T. Clinically relevant superiority is the key to each new modality earning its way into routine and specialized imaging settings. For UHF MRI higher resolution and changes in tissue
contrasts could be the basis for better and earlier depiction of imaging abnormalities. However, in the end, added value for the patient has to be proven. This study does not answer the last point, but shows clear advantages in cartilage defect visualization. This thin, high proton density tissue seems to
Table 2 Subjective image evaluation using a four-point scale (4 = very good; 3 = good; 2 = fair; 1 = unacceptable) displaying average value, SD, and significance Resolution
Homogeneity in ROI
Contrast cartilage/7bone
Contrast cartilage/fluid
MEDIC
7T 3T p
3.9 4.0 0.35
0.4 0.0
2.9 4.0
SCIENTIFIC ARTICLE
Hip imaging of avascular necrosis at 7 Tesla compared with 3 Tesla J. M. Theysohn & O. Kraff & N. Theysohn & S. Orzada & S. Landgraeber & M. E. Ladd & T. C. Lauenstein
Received: 23 October 2013 / Revised: 30 December 2013 / Accepted: 2 January 2014 # ISS 2014
Abstract Objectives To compare ultra-high field, high-resolution bilateral magnetic resonance imaging (MRI) of the hips at 7 Tesla (T) with 3 T MRI in patients with avascular necrosis (AVN) of the femoral head by subjective image evaluations, contrast measurements, and evaluation of the appearance of imaging abnormalities. Materials and Methods Thirteen subjects with avascular necrosis treated using advanced core decompression underwent MRI at both 7 T and 3 T. Sequence parameters as well as resolution were kept identical for both field strengths. All MR images (MEDIC, DESS, PD/T2w TSE, T1w TSE, and STIR) were evaluated by two radiologists with regard to subjective image quality, soft tissue contrasts, B1 homogeneity (fourpoint scale, higher values indicating better image quality) and depiction of imaging abnormalities of the femoral heads (three-point scale, higher values indicating the superiority of 7 T). Contrast ratios of soft tissues were calculated and compared with subjective data.
J. M. Theysohn (*) : O. Kraff : N. Theysohn : S. Orzada : M. E. Ladd : T. C. Lauenstein Erwin L. Hahn Institute for Magnetic Resonance Imaging, University Duisburg-Essen, Essen, Germany e-mail: [email protected] J. M. Theysohn : O. Kraff : N. Theysohn : S. Orzada : T. C. Lauenstein Institute for Diagnostic and Interventional Radiology and Neuroradiology, University Hospital Essen, Essen, Germany S. Landgraeber Department of Orthopedic Surgery, University Hospital Essen, Essen, Germany M. E. Ladd Department of Medical Physics in Radiology, German Cancer Research Center, Heidelberg, Germany
Results 7-T imaging of the femoral joints, as well as 3-T imaging, achieved “good” to “very good” quality in all sequences. 7 T showed significantly higher soft tissue contrasts for T2w and MEDIC compared with 3 T (cartilage/fluid: 2.9 vs 2.2 and 3.6 vs 2.6), better detailed resolution for cartilage defects (PDw, T2w, T1w, MEDIC, DESS>2.5) and better visibility of joint effusions (MEDIC 2.6; PDw/T2w 2.4; DESS 2.2). Image homogeneity compared with 3 T (3.9–4.0 for all sequences) was degraded, especially in TSE sequences at 7 T through signal variations (7 T: 2.1–2.9); to a lesser extent also GRE sequences (7 T: 2.9–3.5). Imaging findings related to untreated or treated AVN were better delineated at 3 T (≤1.8), while joint effusions (2.2–2.6) and cartilage defects (2.5–3.0) were better visualized at 7 T. STIR performed much more poorly at 7 T, generating large contrast variations (1.5). Conclusions 7-T hip MRI showed comparable results in hip joint imaging compared with 3 T with slight advantages in contrast detail (cartilage defects) and fluid detection at 7 T when accepting image degradation medially. Keywords Ultra-high field . UHF . Hip . MRI . Avascular necrosis . AVN
Introduction Magnetic resonance imaging (MRI) is the gold standard or has been suggested to be the imaging test of choice in the diagnostic evaluation of hip joint abnormalities. This is currently realized using MR systems at field strengths of 1.5 or 3 Tesla (T), generating high-resolution images with strong soft-tissue contrasts [1]. While ultra-high field (UHF) MRI systems for human use at 10.5 T (whole-body) and 11.7 T are about to start running early in 2014, approximately 40 UHF systems at 7 T and 9.4 T (ratio ~ 6:1) have been installed worldwide. Active research at 7 T and 9.4 T has been performed for more than 10
Skeletal Radiol
years, with the main focus on the brain and, lagging behind, the musculoskeletal system using smaller fields-of-view (FOV) [2–6]. Compared with lower field strengths, their implementation for fundamental research in volunteers (i.e., functional MRI) and patient imaging has revealed improved diagnostic capabilities in the brain and joints. The roughly twofold or fourfold increased signal-to-noise ratio (SNR) (over 3 T or 1.5 T) was transformed into stronger blood oxygenation level-dependent (BOLD) contrast, higher image resolution, and stronger image contrast. Transferring this new technology to the torso requiring the use of larger FOVs is a rocky road challenged by strong image degradations [7]. Radiofrequency (RF) field distortions in body imaging with large FOVs beginning to appear at 3 T are much more enhanced at 7 T [8, 9]. This has led to the delay in useful thoracic, abdominal, and pelvic imaging at 7 T. Feasibility studies have emerged only over the last few years, after selfbuilt multi-channel transmit body coils had been established [10–16]. Only a small number of comparison studies of 7 T with lower field strengths have been published on imaging these regions, including validation against 3 T in the wrist, ankle, breast, and the heart [17–21]. However, none of these studies included patients, or an analysis of imaging abnormalities. Avascular necrosis (AVN) of the hip is one endpoint of bone destruction due to an altered microcirculation, subchondral oxygen deficiency and necrosis [22–24]. While AVN mostly occurs in middle-aged males it can lead to severe degenerative disease of the hip joint, possibly requiring endoprosthetic hip replacement (10 % of all are due to AVN) [25–27]. Efforts to treat the early stages of AVN that are not manageable using conservative pain strategies include bone decompression [28]. Recently, this procedure has been supplemented with different biological bone graft injections used to fill the decompressed and carved out necrotic lesion [29, 30]. In the event of treated or untreated AVN, MRI can visualize aspects of the necrotic lesion (necrosis, reactive zone/granulation tissue, sclerotic changes, edema), as well as postoperative changes (edema) and the drilling hole or injected bone substitute (bone graft). Technical progress in imaging has enhanced sensitivity for the detection of subchondral fractures in the later stages of the disease. However, sensitivity compared with computed tomography (CT) has been shown only to be fair to moderate [31]. Conspicuity in MRI is dependent on a fine subchondral fluid collection at the location of the bony depression or hypointense line on T1weighted (T1w) images [32, 33]. The reactive border is noticeable as a “double line sign” with an outer sclerotic band (T1w and T2w hypointense) and adjacent T2w hyperintensity representing granulation tissue [34–37]. Comparison studies of 7 T with lower field strengths, preferably 3 T in musculoskeletal imaging, are warranted, including patient data. We hope that the diagnostic advantage
of 7 T, as has been shown for peripheral joints with small cross-sections (i.e., the knee) [6], can also be transferred to the hip joints. The aims of this study were twofold: to compare imaging results of the hips at 7 T with those at 3 T using subjective evaluations and contrast measurements, and to analyze and compare the appearance of imaging abnormalities in patients with AVN of the femoral head between the field strengths.
Materials and methods Subjects and hardware After gaining approval from the local institutional ethics committee and written informed consent from all subjects, 13 patients aged 30 to 67 years (mean age 51.4 years, SD 9.4 years; 4 female, 9 male) were included in this study. All patients had a known history of AVN of one or both femoral heads treated with advanced core decompression. During this intervention a surgeon drills a hole through the femoral neck up to the necrotic area in the femoral head, removes part of the necrotic material and refills the cavity and drilling hole with a bone graft substitute [30]. Patients were imaged between 8 weeks and 16 months after surgery. MRI examinations were performed on two different systems not more than 2 weeks apart. A 7-T whole-body scanner (Magnetom 7 T, Siemens Healthcare, Erlangen, Germany) was used in combination with a custom-built eight-channel transmit/receive coil, since neither an integrated transmit body coil as is known from 1.5T or 3-T systems, nor commercially procurable RF coils for ultra-high-field abdominal MRI are available. The RF coil allowed static transmit phase shimming to mitigate signal voids medially out of the region of interest [38]. Therefore, each coil element is connected via a modulator box containing amplitude and phase shifters to individual 1-kW modules of a RF power amplifier (RFPA) situated in the equipment room. A total cable length of more than 20 m resulted in an attenuation of −3 dB or 50 % between RFPA and coil plug, limiting the available peak RF power to approximately 4 kW in total. At 3 T, a state-of-the-art clinical MRI system (Skyra, Siemens Healthcare, Erlangen, Germany) was utilized, paired with a scanner-integrated body transmit coil and receive coils as provided by the vendor (18-channel body matrix coil combined with 16-channel spine array). Here, the available peak RF power amounted to 37.5 kW, with the RFPA situated directly at the magnet for improved performance. The multichannel coils at both field strengths allowed for parallel imaging techniques, allowing shorter acquisition times compared with single-channel coils (generalized autocalibrating partially parallel acquisitions [GRAPPA], acceleration factor of R=2).
Skeletal Radiol
MR imaging An imaging protocol at 7 T had been evaluated in an earlier study [14], adapted from sequences used to image the knee joint with a local volume coil [39] with the intention of investing additional SNR into higher resolution. Table 1 provides an overview of all sequence parameters. Images were acquired in coronal orientation: proton density (PD)/T2w turbo spin echo (TSE), T1w TSE, multi-echo data image combination (MEDIC), double echo steady-state (DESS), and short-tau-inversion-recovery (STIR). For the TSE sequences, prolonged 180° RF pulses (7.68 ms) [40] were implemented to keep the total amount of power within the SAR limits, while playing out the high amplitude pulses needed for pelvic imaging. Owing to conservative SAR limits at 7 T (10 W/kg local SAR, as defined for normal operating mode by IEC guidelines), all TSE sequences were nevertheless limited to 4–5 slices per acquisition, requiring multiple repetitions for full femoral head coverage. The same sequence parameters were used at 3 T, resulting in, for example, identical receiver bandwidths at both field strengths (only minimal adjustments owing to technical limitations were allowed, but no further optimization was performed). With this choice the chemical shift displacement was less at 3 T compared with 7 T and yielded a better delineation of anatomical structures at 3 T. On the other hand a change in receiver bandwidth would have resulted in additional SNR variations, thereby challenging the comparability. Image evaluation All MR images were reviewed using a PACS (picture archive and communication system) workstation (Centricity; General
Electric Healthcare, Barrington, IL, USA) and were assessed by two radiologists in consensus (8 and 6 years of experience in musculoskeletal MRI). Since field-strength was obviously identified through the medial signal voids and change in tissue contrast on 7-T images, blinded reading was not possible. However, radiologists consciously aimed not to overrate 7-T MRI. Qualitative properties (resolution, homogeneity, and soft tissue contrasts among cartilage, fluid, and bone) of each image set of each examination were rated on a four-point scale, looking only at the hip joint (1 = poor, 2 = fair/only partially diagnostic, 3 = good/diagnostic, 4 = excellent image quality) and compared them at the two field strengths. In STIR images the characteristic contrast through fat suppression was graded in a similar fashion. Homogeneity was only evaluated directly around the joint, accepting technically inevitable, medially located signal voids. We performed quantitative signal measurements of soft tissues (cartilage, labrum, fluid, and bone) and calculated contrast ratios: CR=(SIA −SIB)/(SIA +SIB). The signal intensities were measured drawing a free-hand ROI over the tissue of interest after magnifying the selected structure. ROIs for cartilage and bone were placed on the same anatomical site in most patients: lateral to the insertion of the round ligament of the femur. Only if this cartilage was thinned out (n=3) and not measurable was the medial femoral head chosen. Locations at both field strengths were always identical. Again these data was compared at the two field strengths and correlated with subjective results. Depiction of postoperative changes after advanced core decompression of AVN (treated lesion, bone marrow edema) or MR abnormalities of the hip joints not directly related to the surgical procedure (cartilage defect, joint effusion, untreated
Table 1 Sequence protocol for 7 T and identicala 3 T
Scan timeb (min:s) Acquisition matrix Resolution, interpolated (mm2) Resolution, acquired (mm2) Slice thickness (mm) Flip angle (°) Echo timeb (ms) Repetition timeb (ms) Turbo factor
MEDIC
DESS
PD/T2
STIR
T1
05:33 512×512 0.38 0.76 1.5 30° 15 1,000
08:20 512×512 0.38 0.76 0.76 15° 4.1 11
03:26c 512×512 0.38 0.76 3 180° 29/88 4,350 7
03:32c 384×384 0.51 1.02 3 180° 32 5,000 5
01:35c 512×512 0.38 0.76 2 180° 8.3 900 5
All imaging was performed using a field of view (FOV) of 390 mm×390 mm and, except for T1w, with an acceleration factor of GRAPPA=2 a
After transferring sequences from 7 T to 3 T only minimal adjustments were allowed owing to technical limitations, but no further optimization was performed
b
Scan times at 3 T varied ± 15 s owing to slightly different permitted echo and repetition times
c
Sequences were repeated at 7 T to cover the femoral heads. Limitations in the allowed number of slices owing to SAR restrictions made this necessary
Skeletal Radiol
lesion) were evaluated with regard to which field strength performs better (1: 7 T inferior, 2: 7 T equal to 3 T, and 3: 7 T superior). Sequences were only evaluated for depiction of MR abnormalities, if visible at at least one field-strength (i.e., if no edema was present, it was not rated). All quantitative and qualitative measurements are given as mean and standard deviation. Statistical significance testing was performed using a two-sided Wilcoxon signed-rank test (with p3.0; Figs. 1, 2), with only 7-T T2w lagging behind in the field comparison (3.2), compared with 3 T (3.6). Throughout all sequences 3 T provided greater homogeneity of the hip joints (3.9-4.0) compared with 7 T (Figs. 3, 4), where, despite shifting signal voids medially, residual B1 heterogeneities remained visible overlying the joints (2.1–3.5). TSE sequences at 7 T (2.1–2.9) were much more degraded than gradient echo imaging (2.9–3.5). Looking at cartilage/fluid contrast 7 T was graded superiorly in MEDIC (3.6 vs 2.6; Fig. 5a, b), on T2w (2.9 vs 2.2), and STIR (2.0 vs 1.0), while 3 T performed slightly better in PDw (2.9 vs 2.5) and DESS (2.8 vs 2.6). For cartilage/bone contrast only T2w generated noteworthy advantages of 7 T (3.0 vs 2.2). Finally, the STIR contrast, or fat suppression respectively, at 7 T (2.1) could not compete with 3 T (4.0; Figs. 2, 4). The comparison of measured contrast ratios proved equal for both field strengths in the majority of cases (Table 3). While 3 T performed better for the cartilage–fluid contrast for T2w imaging (0.66 vs 0.57), MEDIC showed a minor advantage at 7 T (0.32 vs 0.20). While labrum–fluid contrast had lower values at 7 T for T2w (0.49 vs 0.75), a dramatic difference in T1w images was apparent with only 3 T showing visible
identical resolution, slight subjective advantages in detail resolution at 7 T in a, e MEDIC, b, f DESS, c, g PDw, and d, h T2w; c, d only on the 7-T TSE sequences can a faint signal loss toward the edge of the image (star) be seen
Skeletal Radiol
Fig. 2 Coronal views of the left hip joint in a 44-year-old male patient suffering from AVN 3 months post-ACD of the left femoral head (see also Figs. 4 , 5c–e). While using identical resolutions at 7 T (top row), images appear slightly crisper and with somewhat stronger areas of contrast compared with 3 T (bottom row) in a, e MEDIC, b, f PDw, c, g T2w,
and d, h STIR. On the 7-T TSE sequences (b–d) a residual signal loss toward the edge of the image (star) can be seen and signal variations cause slight contrast variations, making 3 T better in the depiction of the treated lesion (f, g) and surrounding postoperative changes (h). Arrow in c: drilling hole refilled with a bone graft material
contrast (0.46 vs 0.04); PDw, DESS, and MEDIC were of equal value at the two field strengths. Finally, cartilage–bone contrasts had mainly superior 7 T values (PDw 0.67 vs 0.41, T2w 0.59 vs 0.14, DESS 0.81 vs 0.49), with only 3-T MEDIC proving slightly advantageous (0.81 vs 0.69). When subjective and measured contrast areas are being held side by side many parallelisms can be acknowledged. Measured cartilage–fluid contrast for MEDIC was superior at 7 T (0.32 vs 0.20) with a good correlation in the subjective data (3.6 vs 2.6). While DESS and PDw agree with comparable results, a slight contrast advantage for cartilage–fluid in T2w exists at 3 T (0.66 vs 0.57); 7 T dominates in the subjective data (2.9 vs 2.2), but showing a fairly high standard deviation (1.0). A much stronger measurement of cartilage–bone contrast for T2w at 7 T (0.59 vs 0.14) agrees with the subjective impression (3.0 vs 2.2). Smaller measured 7 T advantages for PDw (0.67 vs 0.41) and DESS (0.81 vs 0.49), or a smaller disadvantage for MEDIC (0.69 vs 0.81) are not mirrored in the subjective evaluations, where field strengths perform to a comparably high standard (3.0 vs 3.1; 3.8 vs 3.6; 3.8 vs 3.6). Imaging findings (necrosis, reactive zone/granulation tissue, sclerotic changes, edema, bone graft) related to untreated or treated AVN are better delineated at 3 T with values ≤1.8 (Table 4) (Figs. 2–4); this includes the lack of reasonable depiction of bone marrow edema at 7 T on STIR images (1.3; Figs. 2d, 4d, h) and changing contrasts in TSE sequences close to the medial signal void (Fig. 3b–d). Joint effusions were
better visualized at 7 T on MEDIC (2.6), PDw (2.4), T2w (2.4), and DESS (2.2; Fig. 2); but 7-T STIR performed much more poorly compared with 3 T, generating large variations in image contrast (1.5). Finally, for cartilage defects 7 T was superior to 3 T on MEDIC (3.0), DESS (2.8), T2w (2.6), T1w (2.6), and PDw (2.5; Fig. 5).
Discussion With this comparative study it was demonstrated that hip imaging at 7 T essentially meets standards set by 3 T, looking at the joints themselves. This is particularly interesting as, from a technical point of view, there are major differences in the RF chain between the two MR systems. Not only was the number of receive elements underrepresented at 7 T compared with 3 T by nearly a factor of 3, the available peak RF power also differs significantly (37.5 kW “DirectRF” at 3 T vs 4 kW at 7 T). Nordmeyer-Massner et al. have used an identical GRE sequence and identical receive coils, proving 0- (bone) to 2fold (cartilage, nerves) SNR gains at 7 T compared with 3 T depending on the tissue [18]. While larger signal differences in tissues can translate into stronger contrast, as has been shown by our data, they could not improve the visibility of anatomical structures. In patients, it was found that 7 T may display cartilage defects more accurately, using identical sequences at both field strengths, while a comparable depiction
Skeletal Radiol
Fig. 3 Coronal views of the pelvis in a 41-year-old male patient suffering from AVN of both femoral heads 4 weeks post-ACD of the left femoral head. Comparison of T1w images at a 3 T and b 7 T visualize relatively strong residual signal voids overlying the right hip joint (star in b)— compare Fig. 4 for the good shimming results mostly seen in our population. Signal alteration can not only reduce signal medially (star in b–d), but can cause contrast to vary across the image: in c 7-T PDw and d 7-T T2w the border zone of the right-sided, untreated necrotic lesion changes between hyper- and hypointense signal intensity (black arrows). The reactive border zone between bone graft material and surrounding bone can be appreciated as a hyperintense line (arrowhead) in c (PDw) and d (T2w)
of intraosseous imaging abnormalities due to AVN or postsurgical changes is within reach. Lack of decent bone marrow edema contrast at 7 T is still one of the biggest challenges because of the need for T2-weighted spin echo sequences. Since images at both field strengths were acquired with identical (high) resolutions, it is not surprising that subjective resolution evaluation issued high values for all sequences at both field strengths. The disadvantage of medial image degradation at 7 T has expectedly lower grading in all 7-T sequences (spin echo >>gradient echo) because 7 T suffers from B1 heterogeneities and a limited penetration depth, which are not truly visible at 3 T. Such interferences arise when an object is excited by waves whose length is shorter
than the object’s diameter: wave length in tissue is about 15 cm at 7 T and 30 cm at 3 T [41]. This explains why the problem has been identified and described at 3 T, but because of its low degree of severity has not truly required its extinction [42, 43]. STIR imaging yielded lowest image quality at 7 T as this sequence depends on two consecutive high flip angle pulses for the inversion and refocusing pulses. Coping with challenges at 7 T will further improve 3-T imaging results. Mastering these obstacles includes the use of the relatively new multichannel transmit/receive technology at 7 T with possible modulation of B1 heterogeneities, which was used here and has been illustrated in publications on hip imaging [11, 13, 14]. Juras et al. used optimized sequences in ten healthy volunteers at 3 T and 7 T to compare imaging performance in the ankle [17]. While they observed a CNR increase for PDw TSE and most 3D T1w GRE images (but not T1w TSE), our results in patients showed more heterogeneity. However, taking into account differences in cross-sections and thus the aforementioned challenges in B1 homogeneity and penetration depth, this difference may be less surprising. Furthermore, Juras et al. used an different RF coil setting for their comparison, namely a commercially available 28-channel coil at 7 T versus an eight-channel coil at 3 T. In our study, the superiority of 7-T MEDIC for cartilage–fluid contrast in qualitative and quantitative data (T2w only in the qualitative data) can be explained by higher intrinsic SNR and the fact that MEDIC suffers less from image degradation at 7 T than TSE sequences, owing to its nature as a low flip angle gradient echo sequence. This contrast enhancement may be translated to clinically used spin echo sequences as engineering process advances at 7 T. Other sequences (PDw, T2w, DESS) do not show the domination of one field strength, putting 7 T on a par with 3 T. Identical reasoning can be alleged for stronger cartilage–bone contrast at 7 T, mainly for T2w and to a lesser extend PDw and DESS, with intrinsically higher SNR producing stronger contrasts between fluid-rich tissue (cartilage) and low-signal tissue (bone). Lastly, while measured labrum–fluid contrast at 7 T showed equal values for most sequences, a slight contrast reduction at 7 T might be due to our population mainly having degenerated joints with calcifications of the labrum; stronger susceptibility effects at 7 T could account for this signal reduction. The fact that, in our study, 7 T mostly performs with even quality compared with 3 T might be attributable to the early stage of 7-T prototype coil technology, while 3-T coils are commercially available and well optimized. This lagging behind in coil development may be a reason for losing some of the potential of the theoretically more than twofold increase in SNR. Deligianni et al. have shown that a dedicated high-resolution 3D GRE sequence could be well-integrated into 3 T as well as 7 T for musculoskeletal imaging [20]. Inventing new sequences as well as further optimizing 7-T sequences in our study, aiming to exploit the full potential,
Skeletal Radiol
Fig. 4 Coronal views of the pelvis in a 44-year-old male patient known from Fig. 2 (see also Fig. 5c–e). High-resolution a, b T1w and c, d STIR images of both hip joints at 7 T (left) and 3 T (right). 7-T images represent the typical localization of the medially shifted signal void (star) after B1 shimming with good depiction of both femoral heads. On T1w images one can appreciate the better delineation of the joint effusion (white
circles) in a at 7 T while in b 3 T generates almost no contrast to the surrounding muscle tissue. Quality of fat suppression and contrast homogeneity on STIR images in c at 7 T cannot compete with d 3 T so far. Black arrow: drilling hole refilled with a bone graft material, surrounded by postoperative changes
could further facilitate finding clinically useful advantages at 7 T. Clinically relevant superiority is the key to each new modality earning its way into routine and specialized imaging settings. For UHF MRI higher resolution and changes in tissue
contrasts could be the basis for better and earlier depiction of imaging abnormalities. However, in the end, added value for the patient has to be proven. This study does not answer the last point, but shows clear advantages in cartilage defect visualization. This thin, high proton density tissue seems to
Table 2 Subjective image evaluation using a four-point scale (4 = very good; 3 = good; 2 = fair; 1 = unacceptable) displaying average value, SD, and significance Resolution
Homogeneity in ROI
Contrast cartilage/7bone
Contrast cartilage/fluid
MEDIC
7T 3T p
3.9 4.0 0.35
0.4 0.0
2.9 4.0
