|Home | About | Journals | Submit | Contact Us | Français|
To design and compare an eight-channel phased array (PA) coil for carotid imaging to an established four-channel PA design at 3T.
An eight-channel PA (8PA) coil was designed specifically for imaging the carotid bifurcation and compared with the existing four-channel (4PA) design using a phantom and by in vivo black-blood magnetic resonance imaging (MRI). The 8PA and 4PA were compared in terms of coverage, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR).
The 8PA showed up to 1.7-fold improvement in SNR at a depth of 3.5 cm and greater longitudinal coverage at a given SNR on a phantom. The 8PA showed improved vessel wall SNR for high spatial resolution (0.63 mm2) PD, T1, and T2 (1.7, 1.7, 1.6 times, respectively; P ≤ 0.002) and improved CNR (1.7, 1.6, 1.5 times, respectively; P ≤ 0.002). Ultrahigh-resolution (0.27 mm2) T1-weighted images showed better SNR and CNR (1.4 times, P ≤ 0.0001) on 8PA compared to 4PA.
Carotid imaging studies may benefit from the improved SNR and larger coverage provided by use of the 8PA.
Carotid Atherosclerosis is a leading cause of stroke-related morbidity and mortality due to rupture of atherosclerotic plaques formed near the carotid bifurcation. Plaque size and composition are known to be important determinants of plaque vulnerability (1–3). Noninvasive evaluation of plaque morphology and composition by imaging techniques is desirable to predict vulnerable plaques. High-resolution magnetic resonance imaging (MRI) of the carotid vessel wall can accurately measure lesion size and comprehensively assess plaque composition (4,5) and thereby assess plaque vulnerability. Multicontrast carotid MRI can identify vulnerable plaque components such as the presence of a large necrotic core (6–8), thin or ruptured fibrous cap (8,9), intraplaque hemorrhage (10,11), and has been shown to predict patient symptoms (12,13). Carotid MRI is increasingly used to monitor lesion progression and regression in response to antiatherosclerotic therapies (14,15). Higher spatial resolution MRI is required to identify the small changes in plaque morphology and composition typical of short drug trials. Maintaining good signal-to-noise ratio (SNR) at a high spatial resolution within a clinically acceptable scan time is crucial for measuring small plaque components.
Phased-array (PA) surface coils specifically designed for carotid imaging have been shown to provide improved SNR at the carotid bifurcation (16–18). These studies have demonstrated the superiority of PA coils and identified favorable coil element geometries. Availability of clinical 3T scanners and newer black-blood multicontrast pulse sequences optimized for imaging at 3T provide improved SNR over 1.5 T (19). However, limited experience with PA coils for carotid imaging is available at 3 T (20).
Current coil designs have shortcomings for use in large-scale clinical trials and prospective epidemiologic studies of carotid atherosclerosis. The location of the carotid bifurcation varies significantly from patient to patient, and from side to side within the same patient. Thus, larger anatomical coverage along the length of the artery is needed to avoid coil repositioning. Additionally, in large subjects with thicker necks the carotid arteries are deep and located away from the surface. Coil sensitivity at these depths is an important design consideration for studies involving such subjects. With the development of faster scanning techniques and improvements in 3D carotid vessel wall imaging (21–23), longer segments of the artery can be assessed by increasing coil coverage along the length of the artery. Additionally, longitudinal coverage of up to 5 cm of the carotids is desirable for assessment of plaque away from the bifurcation and for maintaining lesion coverage in serial studies. Thus, a dedicated carotid coil optimized to improve both SNR and coverage of the carotid bifurcation, is required to enable high-resolution imaging for use in clinical studies at 3T.
A high SNR four-element PA (4PA) coil for carotid imaging has been previously demonstrated at 1.5T (16). In this study a new coil of similar design with eight elements (8PA) is compared against the 4PA design at 3T in terms of coverage, SNR, and contrast-to-noise ratio (CNR).
4PA and 8PA coils were compared on phantom and in vivo carotid images on volunteers. The imaging parameters of the phantom study were optimized for optimal SNR measurement. The in vivo study followed the current protocols in carotid atherosclerotic imaging.
Two receive-only phased array coils with four and eight elements were designed to provide a high SNR at a limited depth across a moderate field of view (FOV). The 4PA consisted of two transverse pairs of approximately square coils applied bilaterally and has been described previously (16). The 8PA consisted of two sets of four coils occupying points of a square applied bilaterally (Fig. 1). The overlap in the transverse direction minimizes mutual inductance. In order to minimize the valley in signal peaks, the overlap along the longitudinal direction was made greater than the overlap required to cancel the mutual inductance. Excess mutual inductance was compensated by a common capacitor to decouple the longitudinal coils. The transverse overlap minimizes coupling between both laterally adjacent and diagonally adjacent coil pairs. The pattern of conductive loops was fabricated by etching a flexible circuit board. Capacitor values were scaled to provide tuning and matching at the appropriate operating frequencies.
Both 4PA and 8PA have a flexible design to allow adjustment to the neck shape and close positioning to the carotid bifurcation region on the neck. The coil padding on the 8PA is slightly larger to accommodate the coil elements but did not hamper placement of the coil on the neck. Both coils are shown positioned on a volunteer in Fig. 2.
A cylindrical bottle (6.5 cm diameter) filled with mineral oil was scanned using the standard resolution T1w turbo spin-echo sequence: TR(msec)/TE(msec)/echo train length/Slice Thickness/Matrix/Number of Excitations = 800/9/11/2/256×192/1 using both 4PA and 8PA on a Philips (Best, Netherlands) Achieva 3T scanner. Position of the coils was marked on the phantom so as to ensure similar positioning of coils. Both axial and coronal images were obtained with similar imaging parameters. An FOV of 250 × 250 mm was used for axial and 400 × 400 mm for coronal phantom images in order to obtain noise estimates unaffected by sum-of-squares recombination of individual coil elements. Both axial and coronal SNR profiles were measured. The axial intensity profile was measured along a 20 pixel wide line drawn in the region of maximal coil sensitivity on an axial image. A coronal profile was measured along a 20 pixel wide line at a depth of 3.5 cm from the surface of the phantom. Axial and coronal line intensity profiles were measured on corresponding locations on both coil images. The standard deviation of noise was measured from a region of interest (ROI) placed in the air surrounding the phantom. SNR was calculated using the line intensity profiles with appropriate correction factors for the number of channels (24).
A total of five healthy subjects (four males and one female, 28–41 years old) with informed consent participated in the study according to Institutional Review Board guidelines. Bilateral carotid arteries were scanned. Altogether, 10 arteries and 150 slices were available for review. An ultrahigh-resolution protocol (0.27 mm2 in-plane) was acquired on four of the subjects (four arteries and 24 slices).
All subjects were scanned with the 4PA and 8PA using the same protocol. The time interval between both scans did not exceed 7 days. The body coil was used for signal transmission and PA coils were used for signal reception. A multicontrast carotid imaging protocol with T1-, T2-, and PD-weighted black-blood images was used to evaluate the performance of the coils. The imaging parameters used (Table 1) were similar to currently established carotid imaging protocols. A 2-mm slice thickness, 160 × 160 mm rectangular FOV, 256 × 256 matrix, and 1 excitation were used providing an in-plane resolution of 0.63 mm. A quadruple inversion recovery (25) (QIR) preparation was used for blood suppression in the T1 weighted sequence while multislice double inversion recovery (26) (MDIR) was used for T2- and PD-weighted sequences. To assess possible improvements in resolution, an ultrahigh-resolution T1- weighted sequence with an FOV of 140 × 140 mm and larger matrix size of 512 × 512 (in-plane resolution = 0.27 mm) was used with both PA coils.
Images were outlined by an independent reader blinded to coil configuration. Images were presented in random order with respect to coil configuration and were not reviewed simultaneously. Lumen and outer wall boundaries were identified by the reviewer using a semiautomated snake algorithm implemented in custom-designed image analysis software. Images obtained with 4PA and 8PA coils were matched using the carotid bifurcation as an internal reference. Quantitative SNR and CNR were assessed automatically from the lumen and outer wall contours. The signal intensity was measured for all pixels contained between lumen and outer wall boundaries. Noise was measured from an ROI placed in the background free from signal and artifacts. The lumen SNR (SNRl), wall SNR (SNRw), and black blood CNR were calculated. SNR was calculated as SNR = S/σ, where S is the true signal intensity corrected for the noise contribution and σ is the true standard deviation (SD) of the noise. S was obtained from the measured magnitude signal (Sm) and the measured magnitude of the background noise (Sn): S = (Sm2 − Sn2)1/2(27). σ was calculated from the measured SD of noise (SDn) and the number of receivers N. The signal in magnitude images reconstructed from multiple receivers follows a noncentral chi-square distribution (24). In magnitude images with SNR 2.0, SDn and σ are related by (28):
Where N is the number of receivers and β is given by:
Thus, SDn = 0.695σ for a 4-element coil and 0.702σ for an 8-element coil. CNR was defined as SNRw – SNRl.
Artery level SNR and CNR obtained by averaging across locations were compared using a two-tailed paired Student’s t-test. All statistical tests were done using SPSS 12.0 (Chicago, IL).
SNR along both axial and longitudinal directions was greater using 8PA on the phantom. The SNR along the axial direction was on average 1.7 times greater using the 8PA with greater gains at increasing depth from the surface (Fig. 3a, Table 2). From Table 2 it can be seen that at a given depth the 8PA provides up to twice the improvement in SNR.
The 8PA showed better SNR over a larger longitudinal extent than the 4PA (Fig. 3b). The improved SNR results in a better spatial coverage. For example, at the depths of 3.5 cm from the surface, full-width half-maximum of SNR profiles were 6.77 cm for 4PA and 9.45 cm for 8PA. For SNR levels of 6, 5, 4, and 3, the head–foot coverage of the 4PA was 3.02 cm, 4.35 cm, 5.83 cm, and 6.98 cm, respectively. For the same SNR levels the head–foot coverage of the 8PA was 10.18 cm, 11.09 cm, 11.98 cm, and 13.26 cm, respectively. Thus, the 8PA can provide an increased coverage of around 6 cm along the head–foot direction at a given SNR level.
The depth of the carotid bifurcations in these volunteers was 2.8 ± 0.4 cm. Artery wall SNR and CNR were significantly greater in standard resolution in vivo images acquired using 8PA (Table 3), with 1.6–1.8-fold improvement in wall SNR (P < 0.0001 for PD, T1, and T2 images) and 1.5–1.7-fold improvement in CNR (P < 0.0001 for PD and T1, P = 0.002 for T2 images). Although a moderate increase in the residual blood signal did occur, the overall quality of black-blood imaging was significantly higher in the 8PA images, as indicated by the lumen-wall CNR. The 8PA images provided better delineation of vessel boundaries than 4PA due to higher contrast and lesser noise (Fig. 4). Ultrahigh-resolution T1-weighted images obtained with 8PA showed a 1.37-fold improvement in wall SNR and 1.49-fold improvement in CNR. The mean improvement in SNR and CNR on ultrahigh-resolution images is shown in Fig. 5.
At a given SNR deeper arteries can be imaged with the 8PA compared to 4PA. Patients prone to atherosclerotic disease are often obese and have thick necks with the carotids located deep from the surface. In studies involving such patients, use of the 8PA coil will provide images of diagnostic quality in a greater proportion of patients compared to the 4PA. The larger longitudinal coverage of the 8PA is also advantageous considering the variability in the location of the carotid bifurcation among patients. Even in the same patient, if the carotids bifurcate at different levels the larger longitudinal coverage will provide better SNR at both bifurcations compared to 4PA.
Although carotid plaque is most frequent at the bifurcation, it may occur anywhere along the length of the artery. Currently, imaging is restricted to a few slices around the bifurcation. In longitudinal studies the coverage may be further reduced due to mismatch between the two timepoint scans. Increased coverage along the artery would benefit longitudinal studies by providing more slices for matching between multiple timepoints. Use of 8PA will support acquisition of a larger number of slices and thus is better suited than 4PA for such studies.
The correction factor for measured signal depends both on the number of coils and the SNR (24). The measured signal overestimates true signal at low SNR areas such as the lumen and has to be corrected for by the appropriate correction factor depending on SNR and number of coil elements. Instead of numerical calculation of the correction factor in vivo, we used a simple correction procedure to calculate the true signal S = (Sm2 − Sn2)1/2. This has been previously shown to closely approximate the analytical value of the correction factor (27).
The correction factor for noise assumes that the noise measurement is uncorrelated between different coil elements. We measured noise from an ROI in the periphery of the image where there was no contribution from signal due to motion or wrap-around. The phantom experiments used a large FOV so that noise ROIs free from artifacts could be measured. In clinical protocols, depending on the size of the subject’s neck and motion artifacts, a clean measure of noise may not be possible. This may explain the slight discrepancies between the SNR ratios measured on phantom and ultrahigh-resolution images. In this regard, our measurements represent the improvements that can be expected with the use of the 8PA in a clinical protocol over the 4PA.
The CNR is a measure of lumen-wall discrimination at the blood–tissue interface and is therefore affected by unsuppressed flow signal. The slight difference between SNR and CNR ratios in our study is likely due to residual flow signal in the arterial lumen.
In conclusion, our initial comparison of two PA carotid coils on volunteers shows the potential benefits of using 8PA over 4PA. With appropriate design, increasing the number of coil elements of a receive-only coil can increase the SNR at a predetermined depth from the skin surface. The current 8PA design provides improved SNR at depths of 2–5 cm where the carotids usually occur, compared to the 4PA coil. This new eight-element PA coil provides a significant improvement in vessel wall visualization in multicontrast weighted black-blood imaging of the carotid artery wall. This improvement permits an increase in spatial resolution that may lead to visualization of small plaque components. Additionally, the coil provides increased coverage along the artery and is thus better suited for longitudinal studies of plaque progression/regression.
Mark Mathis and Tim Wilbur helped with the construction and testing of the coils.
Contract grant sponsor: National Institutes of Health (NIH); Contract grant number: NIH HL056784.