A 128-Channel Receive-Only Cardiac Coil for Highly Accelerated Cardiac MRI at 3 Tesla

doi:10.1002/mrm.21598

Magn Reson Med. Author manuscript; available in PMC 2008 December 1.

Published in final edited form as:

Magn Reson Med. 2008 June; 59(6): 1431–1439.

doi: 10.1002/mrm.21598

PMCID: PMC2548273

NIHMSID: NIHMS66973

A 128-Channel Receive-Only Cardiac Coil for Highly Accelerated Cardiac MRI at 3 Tesla

Melanie Schmitt,¹ Andreas Potthast,² David E. Sosnovik,^1,³ Jonathan R. Polimeni,¹ Graham C. Wiggins,¹ Christina Triantafyllou,^1,⁴ and Lawrence L. Wald^1,⁵^*

¹ Department of Radiology, A.A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, Massachusetts, USA

² MR Division, Siemens Medical Solutions, Charlestown, Massachusetts, USA

³ Department of Cardiology, Massachusetts General Hospital, Boston, Massachusetts, USA

⁴ Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA

⁵ Harvard-MIT Division of Health Sciences Technology, Cambridge, Massachusetts, USA

^* Correspondence to: Lawrence L. Wald, A.A. Martinos Center for Biomedical Imaging, Bldg 149 13th St, Charlestown MA 02129. E-mail: wald/at/nmr.mgh.harvard.edu

The publisher's final edited version of this article is available at Magn Reson Med

See other articles in PMC that cite the published article.

Abstract

A 128-channel receive-only array coil is described and tested for cardiac imaging at 3T. The coil is closely contoured to the body with a “clam-shell” geometry with 68 posterior and 60 anterior elements, each 75 mm in diameter, and arranged in a continuous overlapped array of hexagonal symmetry to minimize nearest neighbor coupling. Signal-to-noise ratio (SNR) and noise amplification for parallel imaging (G-factor) were evaluated in phantom and volunteer experiments. These results were compared to those of commercially available 24-channel and 32-channel coils in routine use for cardiac imaging. The in vivo measurements with the 128-channel coil resulted in SNR gains compared to the 24-channel coil (up to 2.2-fold in the apex). The 128- and 32-channel coils showed similar SNR in the heart, likely dominated by the similar element diameters of these coils. The maximum G-factor values were up to seven times better for a seven-fold acceleration factor (R = 7) compared to the 24-channel coil and up to two-fold improved compared to the 32-channel coil. The ability of the 128-channel coil to facilitate highly accelerated cardiac imaging was demonstrated in four volunteers using acceleration factors up to seven-fold (R = 7) in a single spatial dimension.

Keywords: MR array coil, cardiac MRI, parallel imaging

MRI is playing an increasingly important role in the evaluation of cardiovascular disease. Cardiac MRI is now routinely used to assess ventricular function, perfusion, and viability (1). However, cardiac MR is still hampered by the need for precise double-oblique slice prescriptions and insufficient spatial resolution, particularly for coronary artery imaging. These issues could potentially be addressed through the acquisition of highly accelerated 3D images, which, by providing high-resolution isotropic data, could also be resampled offline for visualization in any plane. The performance of isotropic 3D whole-heart imaging in a single breathhold, however, is not possible with current radio frequency (RF) coil technology.

Phased array surface coils have been shown to improve the sensitivity of MRI (2), and facilitate complex encoding schemes in MRI (3–5). The sensitivity of array coils matches the sensitivity of larger element coils in the center of the body, but provides substantial gains in sensitivity closer to the surface (6–10). Highly accelerated cardiac imaging has been explored with 32-channel systems, with promising results (12–14). Recent studies with a 32-channel cardiac coil at 1.5T, for instance, have shown that sevenfold acceleration factors (R = 7) (14) can be accomplished in a single phase-encode direction using the spatial-temporal time-adaptive sensitivity encoding (TSENSE) method (3) and that acceleration factors of R = 12–16 (11) can be accomplished with parallel imaging in two spatial dimensions. Although they are important and significant advances, present 32-channel systems still do not allow whole-heart, isotropic three-dimensional (3D) imaging to be performed with the resolution and ease approaching that obtained with cardiac CT imaging.

Simulations of systems with even larger numbers (32 channels and above) of independent coil elements and channels have suggested that signal-to-noise ratio (SNR) improvements, particularly close to the surface of the body are possible, in addition to significant improvements in the SNR of accelerated imaging (8,9). An 128-channel body array has also recently been demonstrated with significant advances in sensitivity and capability to accelerate image encoding (11,15). Experimental studies with brain arrays have also examined the potential benefit of extending the number of receivers beyond 32 channels, confirming the potential benefits of ultrahigh element count arrays for accelerated imaging (16). In this work, we tested the utility of expanding the parallel detection approach in cardiac imaging to a regime in which the coil designer was essentially unconstrained by the number of RF channels available on the instrument. An 128-channel cardiac coil, previously presented in abstract form (15), was thus developed for use on a prototype 3T scanner with 128 independent receive channels. We describe the technical issues related to the design of the coil including interele-ment coupling, noise correlation, loading, and common-mode rejection, which all become more challenging with the higher the number of elements. The array’s performance was characterized with G-factor and SNR maps calculated from images of both phantoms and human volunteers and compared to commercially available 32- and 24-channel coils. The ability of the 128-channel coil to facilitate the acquisition of high-quality highly-accelerated images was then demonstrated in normal volunteers.

MATERIALS AND METHODS

Coil Design

The coil consists of a fiberglass mold adapted to a male (weight ≈ 85 kg) thorax, mounted on the scanner table and designed to go into the scanner with the subject in the standard head-first supine position. The coil housing was split into an anterior and posterior section hinged to open in a “clam-shell” geometry. A total of 60 circular loop coils were mounted on the anterior section and 68 on the posterior section. All the coils have an inner diameter of 71 mm with a trace width of 4 mm milled from 0.031 in thick FR4 circuit-board material, and were arranged in a continuous array of hexagonal symmetry and overlapped to minimize nearest neighbor coupling (2). The overall geometry of the coil, and a schematic of a representative coil element with its matching/tuning circuit, as well as an overall cabling diagram, is shown in Fig. 1.

FIG. 1

a: A 128-channel cardiac coil consisting of a fiberglass cradle with 60 anterior and 68 posterior surface coils. The coils are arranged in overlapping hexagonal symmetry to minimize next-neighbor coupling. The preamplifiers are positioned 3 cm above each (more ...)

Each element in the coil was tuned to resonance with six series capacitors (56 pF each), one of which was linked in parallel with an adjustable capacitor (1–30 pF) for fine tuning of the coil’s resonance frequency. The coil was matched to present the load impedance at the preamplifier required to minimize the noise figure of the preamplifier. The coil matching was adjusted by connecting across the capacitor C_m, as shown in Fig. 1b. A C_m value of 56 pF was found to transform the coil load to the correct impedance in a test coil configuration and this matching capacitance was replicated on the other elements without individual adjustment. An active detuning trap was formed around capacitor C_m using a small inductor and PIN diode (MA4P7470F-1072T; Macom, Lowell, MA, USA) to produce a resonant parallel circuit to detune the loop during RF transmission with the body coil.

The preamplifier for each element was mounted approximately 3 cm above the center of the element to improve the compactness of the coil, and then connected to the coil via a 4.5-cm semi-rigid coaxial cable (coax). The length of this coaxial cable was optimized for preamplifier decoupling. The preamplifier also contained a common mode rejection transformer between the first and second stage. The detuning trap was also used as part of the preamplifier decoupling circuit. To achieve preamplifier decoupling, the input impedance of the preamplifier was transformed by the 4.5-cm coax to a low impedance (~2Ω) across the PIN diode on the coil. Thus the preamplifier input impedance is used to insure the presence of a high impedance series element in the coil, which reduces inductive coupling between neighboring elements (2). Finally, unloaded/loaded Q for 10 elements on the top and 10 elements on the bottom were evaluated by measuring the S12 coupling between two broadband loosely coupled probes with and without the body load present. The Q was measured as the ratio of the response width at the 3 dB points to the resonance frequency. The active PIN diode detuning between the tuned and detuned state was determined by measuring the S12 coupling between two coupled probes with the PIN diode powered and the coil tuned to resonance (i.e., with the PIN diode not powered and the preamplifier removed). A similar measurement of the S12 parameter between two decoupled probes was performed to get the preamplifier decoupling state and the state without preamplifier decoupling (i.e., with the preamplifier removed); 20 different PIN diode detuning values were measured.

The output of each preamplifier was connected by a 1.18-mm-diameter silver coax (Leoni Special Cables Gmbh, Friesoythe, Germany) to an interconnection board located at the head end of the coil. The interconnect board was used to distribute the system’s 64 PIN diode bias voltages, which were used to detune the elements during transmission. The bias lines were taken from the standard coil connectors on the patient table of the scanner (P1–P7 in Fig. 1c) near the subjects feet and head (each controlling 32 channels). The bias lines were connected in pairs to double their current driving capability to 200 mA and then used to bias four elements in parallel. Thus the system had the ability to turn on or off elements in groups of four.

After the interconnection board, the RF signals from 96-coil elements were routed out the back of the bore to the second stage amplifiers using the same 1.18-mm-diameter coaxial cables, while 32 RF channels were connected to the standard coil plugs (P1–P5 in Fig. 1c) at the head end of the table. To avoid common mode voltages on the RF path, the coaxial cables were led in groups of 14 to 18 coaxes shielded by a copper braid wound into a balun (B1 in Fig. 1c). To achieve this, the cable assembly (18 coaxial cables and braid) were wound into an 1½turn inductor and tuned to resonance with a parallel capacitor to form a resonant trap circuit to reject common mode currents on the ground braid line. This trap assembly was shielded from direct interaction with the body RF coil by a copper plated box. Additionally, common mode rejection was achieved near the coil elements by leading groups of 16 –18 cables shielded by copper braid through additional bazooka baluns (Siemens Medical Solutions, Erlangen, Germany). Thus, each RF cable entered at least two and as many as three common mode rejection traps between the preamplifier and the second stage amplifier (B2 in Fig. 1c).

After all coils were mounted on the former with a loading phantom, three tuning adjustments were performed on each coil element using a S12 measurement between two untuned probes loosely coupled to the coil element. First, the frequency tuning of each coil was adjusted via the tuning capacitor (Fig. 1). Active detuning of each element was then optimized by monitoring the S12 coupling between a pair of decoupled broadband pickup loops held near the coil with and without the PIN diode forward biased. The inductance L_t was adjusted by physically deforming the inductor to maximize the S12 difference between the tuned and detuned state. Finally, the tuning of the preamplifier decoupling was verified by noting the position of the minimum in the S12 measurement. Further fine-tuning of the preamplifier decoupling was achieved by changing the input impedance of the preamplifier by adjusting the variable capacitor value on the input tank circuit of the individual preamplifier.

Phantom and Volunteer Imaging

All MRI experiments were performed on a 3T whole-body scanner (Trio, A Tim System; Siemens Medical Solutions, Erlangen, Germany) extended to accommodate 128 receive channels (15). This allowed the signal of each individual coil element to be detected with a single RF receive channel. Two different phantoms with a torso shape complementary to the coil former were constructed for evaluation of the coil: one phantom was filled with an agar solution at physiological saline concentration and the other was filled with vegetable oil to avoid the transmit inhomogeneities associated with wavelength shortening in high dielectric material at 3T (7,17).

Initial measurements were performed to evaluate the influence of the surrounding coil elements on the SNR achieved with one coil element in the 128-coil array. A single coil element, constructed with the same diameter and design as used in the 128-channel array, was placed alone on the water phantom and the SNR from this element was evaluated from the mean signal in two different ROIs in the phantom. The noise was evaluated as the mean signal in an ROI from a second measurement with the RF amplitude set to 0 V. For both measurements a gradient recalled echo (GRE) protocol (TR/TE/α = 5000 ms/4.03 ms/90°, FOV = 500 mm, matrix size = 256 × 256, bandwidth (BW) = 300 Hz/pixel, slice thickness = 3 mm) was utilized. The same measurements were repeated with the array present and using the whole array of 128 elements for acquisition of the data. The SNR from a single element in the 128-element array was then evaluated. Care was taken to select an element in the 128-channel array with approximately the same position on the phantom as used in the single-coil measurement.

SNR comparisons between the 128-channel coil and two different commercially available coils were performed in the following way: Using both phantoms, axial, coronal, and sagittal images were acquired with a proton density (PD)-weighted GRE protocol (TR/TE/α = 200 ms/4.03 ms/20°, FOV = 500 mm, matrix size = 256 × 256, BW = 300 Hz/pixel, slice thickness = 10 mm). Moreover, a noise measurement was acquired with the same GRE protocol, but with no RF excitation. The images in each channel were used to determine the individual signal and noise levels in that channel, and SNR was then evaluated by computing the SNR corresponding to the array coil combination method yielding the optimal image SNR. This “optimal SNR” (SNR_opt) reconstruction method utilizes the coil sensitivity maps and noise correlation statistics (2). The same images were used to compute the G-factor maps on a pixel-by-pixel basis (4). The G-factor maps were calculated for 1D acceleration factors ranging from 2 to 8 in both the head-foot (H-F) and left-right (L-R) directions. All analysis was done using a custom MATLAB program (MATLAB 7.1, The Mathworks, Inc.). The SNR and G-factor map measurements on the phantoms were repeated with a commercially available 32-channel dedicated cardiac coil (In Vivo Corporation, Gainesville, FL, USA). We also compared these images to images acquired with the scanner’s Spine Matrix array in combination with two anteriorly placed body matrix arrays (Siemens Medical Solutions), thus using a total of 24-coil elements.

In vivo measurements were performed in four healthy subjects (three men, one woman) after obtaining informed consent in compliance with our institutions regulatory committee. Cardiac imaging SNR comparisons were made from prospectively triggered, segmented PD-weighted GRE images, acquired in the four-chamber view with the following parameters: FOV = 360 mm, slice thickness = 7 mm, 256 × 256, nine segments per RR interval, TE = 4.03 ms, BW = 300 Hz/pixel.

Moreover, we also measured the impact of the 128-channel coil on the transmit field of the body RF coil in one volunteer (subject #1). In this measurement, while the subject was lying in the scanner, the transmitted RF field needed to achieve a 180° pulse with the body coil was measured, with and without the 128-channel coil present. Care was taken to use the same positioning of the volunteer in the magnet for both experiments (with and without the 128-channel coil).

Image quality with the 128-channel coil was demonstrated by acquiring cine images using steady-state free precession (SSFP) (2D-true fast imaging with steady precession [TrueFISP]). The images were acquired in the short-axis view with acceleration factors of up to 7. A total of 10 consecutive 2D slices could be imaged with R = 6 or 7 in approximately 30 s, allowing cine images of the whole heart (10-mm slice thickness) to be performed in a single breathhold. Other image parameters were as follows: FOV = 360 mm, TE/α/BW = 1.4 ms/35°/965 Hz/pixel, matrix size = 101 × 192, in-plane matrix and voxel size = 2.9 mm × 1.9 mm, and a temporal resolution of 16 frames per RR interval. Parallel imaging reconstruction was performed using the generalized autocalibrating partially parallel acquisitions (GRAPPA) algorithm with 24 reference lines for R < 7 and 28 reference lines for R = 7. Cine images, with the identical parameters, were also acquired at one slice location per breathhold with acceleration factors of 1, 2, 4, 6, and 7. In addition, the single-slice protocol was repeated with a slice thickness of 6 mm. Image reconstruction was performed on the scanner using the standard image reconstruction system.

RESULTS

The unloaded Q to loaded Q was approximately 170/57 = 3.0 ± 0.3 for the top 10 coils studied and 169/44 = 3.9 ± 0.6 for the bottom coils. Active PIN diode detuning, measured in 20 coils, was 41.6 ± 2.0 dB. A similar measurement of the S₁₂ parameter between two decoupled probes on 10 different coils in the array showed a mean difference of 31.5 ± 1.9 dB (with a maximum of 35.3 dB and a minimum of 29.5 dB) between the active preamplifier decoupling state and the state without preamplifier decoupling. Figure 2 shows a representative noise correlation matrix from a noise scan in a healthy volunteer. Excluding the normalized diagonal elements (i.e., the self-correlation of the coil elements), the correlation values ranged between 0.02% and 86% with an average of 5.5%. A total of five element pairs showed correlation coefficients above 50% in all volunteers. Three of these element pairs were located on the anterior section of the coil and two of the pairs on the posterior part. All of these coupling element pairs were adjacent to each other and the anterior pairs were located at the neck and near the arms of the volunteers.

FIG. 2

Noise correlation matrix from images acquired in a normal volunteer using the 128-channel coil. Self-correlation of the elements along the diagonal of the matrix has been normalized to 100%. While five coil pairs exhibited relatively high coupling (>50%), (more ...)

The average phantom SNR, evaluated in two different ROIs, with the single isolated coil element was between 41% and 47% better than the SNR from a single element within the 128-element coil, reflecting some signal loss from the presence of the copper in the other coil elements and the wiring. The voltage necessary to achieve a 180° excitation pulse was 602 V with the subject lying in the scanner without the 128-channel coil and 572 V with the subject lying in the 128-channel coil. Therefore, with the coil in the scanner 5% less voltage was needed to achieve the same pulse amplitude.

Optimal SNR maps of the oil phantom in the axial plane are shown in Fig. 3 for each of the arrays. Both the 32- and 128-channel coils showed a significant gain in SNR closest to the surface of the phantom compared to the 24-channel array. Optimal SNR maps in the four-chamber view from two normal volunteers are shown in Fig. 4. In these two volunteers, for whom the anterior part of the rigid 128-channel coil fit closely against the chest wall, the average SNR gains for the 128-channel coil relative to the 32-channel coil were a factor of 1.2, 1.0, and 1.1 in the apex, mid-septum, and mid-lateral wall, respectively. For the other two subjects studied (one male, one female), the rigid coil form did not fit against the chest wall. For these subjects the 32-channel coil SNR exceeded that of the 128-channel array by a factor of 1.3, 1.2, and 1.2 in the apex, mid-septum, and lateral wall, respectively. The SNR values evaluated in the phantom and the volunteers are summarized in Table 1.

FIG. 3

SNR maps of the oil phantom produced with the optimum reconstruction method, which utilizes coil sensitivity maps and noise correlation information. Maps calculated from images acquired with (a) the 24-channel thoracoabdominal coil, (b) a 32-channel cardiac (more ...)

FIG. 4

SNR maps, using the optimum reconstruction method, acquired in the four-chamber view in two normal volunteers. Volunteer 1: (a) 24 channel, (b) 32-channel coil, and (c) 128-channel coil. Volunteer 2: (d) 24 channel, (e) 32-channel coil, and (f) 128-channel (more ...)

Table 1

Mean SNR Values [arbitrary units] for Unaccelerated Imaging Evaluated in Different Locations in the Phantom and Volunteers

G-factor values (mean with SD, minimum, maximum) for the water phantom scanned in coronal slice orientation are summarized in Table 2. The corresponding G-factor maps are shown in Figs. 5 and and6.6. The evaluation of the G-factors revealed a significant benefit for the 128-channel coil in comparison to both the 32- and 24-channel coils. Maximum G-factor values at higher acceleration factors were a factor of 2 or more lower than that of the 32-channel coil. For example, at rate 5, acceleration in the H-F direction, the maximum G-factor of the 128-channel coil reached only 1.2, and was 2.5 for the 32-channel coil. Rate 7 acceleration in the L-R direction produced a maximum G-factor value of 1.7 with the 128-channel array compared to 3.4 for the 32-channel array.

Table 2

1/G Values for the Coronal Water Phantom Images, Scanned With the 128-Channel Coil, the 32-Channel Coil, and the 24-Channel Coil, Respectively

FIG. 5

Coronal 1/G-factor maps of images of the water phantom. The maps were calculated using images from a PD-weighted GRE sequence and noise correlation information. 1D acceleration was studied with the phase encoding direction along the H-F axis of the phantom/coil. (more ...)

FIG. 6

Coronal 1/G-factor maps, as described in Fig. 5, but with the acceleration in the R-L direction. The maximum G-factor is again significantly lower with the 128-channel than the 32-channel coil, and at rate 7 reaches a maximum value of only 1.7 compared (more ...)

The ability of the 128-channel array to facilitate highly accelerated cardiac imaging with high image quality was demonstrated in all four normal volunteers. Whole heart imaging (ten 10-mm-thick sequential 2D slices) could be performed in a single 30-s breathhold with R = 6 or 7 acceleration (Fig. 7). Only minor differences in image quality were seen between the rate 1, 4, and 6 images (Fig. 7). SNR and contrast-to-noise ratio (CNR) were still sufficient in the rate 7 images to support diagnostic quality imaging (Fig. 7). Figure 8 shows 2D SSFP cine images performed with rate 1, 4, and 7 acceleration on a female volunteer with the 128-channel coil with the identical in-plane resolution as the images in Fig. 7 (1.9 × 2.9 mm) but with a slice thickness of 6 mm. Even at this spatial resolution the rate 1 and rate 4 images remain almost identical (Fig. 8a and b), and the rate 7 image remains diagnostic with well-preserved blood tissue contrast and well-defined fine anatomical features. This subject was also judged qualitatively to be the poorest fit to the 128-channel coil former and showed the lowest SNR measurements relative the other two (flexible) arrays.

FIG. 7

Highly-accelerated parallel imaging (GRAPPA) allowing whole-heart cine imaging to be performed in a single breathhold. The images shown were acquired with the 128-channel cardiac coil and the following acceleration factors: (a) rate 1, (b) rate 4, (c (more ...)

FIG. 8

High-resolution, highly-accelerated 2D cine imaging performed in the female volunteer with the 128-channel cardiac coil. The SSFP cine images shown were acquired with a slice thickness of 6 mm and an in-plane resolution of 1.9 × 2.9 mm. The following (more ...)

DISCUSSION AND CONCLUSION

In this study we present the design and characterization of a 128-channel receive-only cardiac coil for highly accelerated cardiac MRI at 3T. The characterization of the coil performance included the evaluation of noise correlation, SNR, and G-factor maps in a mixture of phantoms and human experiments. In addition, we show that high-quality 2D images can be acquired with this coil with 1D acceleration factors of 7, raising the possibility that high-quality 3D datasets using 2D acceleration with signifi-cantly higher acceleration factors than previously obtained will be able to be acquired with this coil.

The primary rationale for the construction of this experimental 128-channel coil was to explore potential issues in constructing a highly parallel array for cardiac imaging. Our goal was to evaluate the potential gains in SNR and acceleration of increasing the number of channels beyond that currently available in clinical systems. We constructed this proof-of-concept array using a rigid body former in order to allow us to assess the ability of highly parallel reception without some of the complicating issues faced in developing a flexible commercial array.

Measurements of the transmitted RF field with and without the 128-channel coil in the scanner showed a 5% transmit voltage difference. Although one might expect transmit power to be dissipated in the copper used in the elements, cables and cable trap shields, and cable common modes, requiring a higher transmit power when the array is present, we have also observed small decreases in the needed transmit voltage in other arrays. This could reflect an incomplete detuning of the array elements causing surface coil focusing by induced fields in the coil, it could also arise from the an impedance shift in the RF body coil causing the body coil to be slightly better tuned and/or matched with the array present.

The measured SNR from one element in the 128-channel array was approximately 43% of the SNR measured with only a single-element coil present on the phantom. This SNR difference might be explained by eddy current losses from the copper in the surrounding cables and coils, as has been shown previously in a 96-channel brain array (18). The unloaded-to-loaded Q ratio of 3.0 for the top elements of the array coil indicates that an SNR gain of only 20% would be expected in the case that the isolated coil had no copper loss (infinite unloaded Q). A second source of signal loss might arise from coupling between the elements in the array, essentially making the coil element act like a much larger diameter coil when in the array (when coupling partners are present). Some of the effects of coupling, however, are addressed when the optimal reconstruction (in which the noise covariance due to coupling is taken into account) method is used. Future advances in wireless or optical technology and more generalized availability of optimal reconstruction algorithms will thus be favorable developments for high-element arrays, such as the 128-channel coil.

A comparison of the SNR and G-factor results of our 128-channel coil with the results from commercially available 24- and 32-channel coils suggests that the 128-channel approach can produce significantly more favorable G-factors than either a 24-channel thoracoabdominal coil or a dedicated 32-channel cardiac coil (Figs. 5 and and6).6). Use of the 128-channel coil also provided a secondary benefit by boosting image SNR for unaccelerated imaging in the subjects that fit well in the rigid coil. As shown in Fig. 4, significant portions of the heart lie close enough to the surface of the chest wall to derive a major gain in SNR from the use of high-element arrays such as the 32- and 128-channel coils, both of which were significantly more sensitive than the 24-channel coil, which contains only 12 elements overlying the heart.

The ability of the 128-channel coil to provide increased sensitivity compared to the flexible 32-channel array appears to be correlated with the ability of the rigid former to fit the subject. For the subject used to form the mold (subject #1), the unaccelerated SNR gain in the apex compared to the 32-channel array was 1.3-fold. All other subjects studied showed smaller gains or, in the case of the two subjects who were a poor fit to the 128-channel former, reduced SNR compared to the 32-channel coil. This suggests the importance of a flexible, tight-fitting coil former to insure the elements over the heart are as close to the body as possible. It also suggests that for unaccelerated imaging, 32 appropriately-sized elements overlying the heart will likely produce close to the optimal SNR profile with some benefit possible in the apex if a flexible 128-channel coil could be constructed without loss of performance. Recent results from a 128-channel flexible body array suggests that highly parallel detection can be successfully coupled with a flexible former (11).

Many of the additional elements in the 128-channel coil were distant from the heart and likely contributed little to SNR in unaccelerated myocardial imaging. This is in contrast to the 32-channel array, which compactly covered only the thorax above and below the heart. While the additional elements distal to the heart likely did not benefit the myocardial SNR in unaccelerated imaging, the G-factor maps obtained from the chest-sized phantom suggest a significant role for distal elements in highly-accelerated cardiac imaging. Our results suggest that this effect can yield more than a factor of 2 in SNR in highly-accelerated imaging compared to a coil with elements concentrated on the heart, such as the 32-channel coil used here. The 24-channel coil, in contrast, covered a larger anatomical area, but appeared to suffer from the large element size required.

The SNR maps produced by the 128- and 32-channel coils are in qualitative agreement with theoretical and experimental investigations of the SNR behavior of coils with different number of coil elements (8,9,16). For example, simulation studies predicted that the SNR near the coil elements should increase significantly with the number of RF channels with lower gains distant from the elements (8,9). Moreover, an experimental comparison of a 32-channel head coil with a commercially available 8-channel head coil at 3T found a very high SNR increase of up to a factor of 3.5 near the coil plane (corresponding to the chest-wall in thoracic imaging) and a factor of 1.4 in the center of the head (16).

The noise correlation between the coil elements in the 128-channel array revealed a low mean coupling value of 5.5%, but five pairs showed correlations over 50%. Three of these element pairs were located on the anterior section of the coil and two of them on the posterior part. All of these coupling-element pairs were adjacent to each other and the anterior pairs were located at the neck and near the arms of the volunteers, where the array has high curvature. We hypothesize that the degree of coil overlap needs to be further optimized in these high-curvature areas. The coupling of the two element pairs on the posterior part of the array is likely due to a leak from the output coax of one element into the preamplifier input of another.

Image reconstruction exploiting both coil sensitivity and noise correlation information, the so-called optimum reconstruction method, is able to reduce the impact of such element coupling within the array. It should be noted, however, that the in vivo images were not reconstructed with this method but rather with a standard sum-of-squares reconstruction.

Several studies have previously been published using 32-channel coils for cardiac imaging at 1.5T (12–14). Niendorf and Sodickson (12) investigated the SNR difference in coronary artery and phantom images acquired with a four-channel cardiac coil, an eight-channel cardiac coil, and their 32-channel coil. Acceleration factors up to rate 4 in one direction and 12 in two directions were used, beyond which image quality deteriorated significantly. Wintersperger et al. (14) also used a 32-channel coil for accelerated cine SSFP protocols at 1.5T in combination with TSENSE. Their study showed that accurate volumetric evaluation was possible until an acceleration of R = 4 was reached. The improved ability to accelerate cardiac imaging (up to R = 7 in 1D) shown in the 128-channel cine images compared to the 32-channel results obtained at 1.5T may reflect the benefits of increased field strength as well as the increased number of elements.

Although the 128-channel coil was designed primarily for accelerated 3D imaging, the benefits of the low G-factors it produced could be clearly seen in the 2D cine images in this study. The female volunteer (Fig. 8) showed acceptable high-resolution (6-mm slice) cine image quality at R = 7 acceleration even though this subject was the poorest fit in the anterior coil former (which was molded to a male chest) and showed the lowest myocardial SNR of the four subjects studied in the 128-channel coil. In addition, phase encoding in short axis cardiac images occurs partially along the anterior-posterior (AP) direction, along which the array has the lowest distribution of coil intensities and thus creates conditions least likely to demonstrate the benefits of increased elements. In contrast, 3D imaging accelerated in two phase encode directions with a significant component of one or more phase encode directions along the right-left (R-L) and H-F direction, should produce further acceleration benefits. The experimental 128-channel system used to acquire the images in this study used a modified version of the operating system that did not support some cardiac features such as retrospective gating and spatiotemporal compression schemes such as TSENSE and time-adaptive GRAPPA (TGRAPPA). These spatiotemporal compression algorithms have the potential to produce even further acceleration benefits compared to SENSE or GRAPPA alone.

The results obtained with this prototype coil suggest several improvements to the 128-channel design. A flexible anterior component to more closely fit varying body shapes will likely increase the sensitivity and clinical potential of the coil. Careful attention to the sources of the highly coupled coil pairs could improve SNR in the local regions dominated by these coils. On-going development in cardiac pulse sequence design will likely facilitate full use of the acceleration capabilities of this coil.

In conclusion, we have developed and tested a 128-channel receive-only coil for cardiac MRI at 3T. SNR and G-factor measurements in phantoms and healthy subjects revealed potential SNR gains from the mid-ventricle to the apex and significantly lower G-factors in highly-accelerated imaging compared to commercially available 24- and 32-channel cardiac coils. The ability to acquire high-quality, highly-accelerated images with this coil has been demonstrated in normal volunteers, with diagnostic quality 2D cine images obtained with up to R = 7 with GRAPPA image reconstruction. As 3D cardiac imaging with 2D acceleration becomes more widely available, highly parallel cardiac arrays, such as the 128-channel coil described in this study, will have the potential to prove of benefit in both the research and clinical settings

Acknowledgments

National Institutes of Health; Grant sponsor: National Center for Research Resources (NCRR); Grant number: P41RR14075; Grant sponsor: National Institute of Biomedical Imaging and Bioengineering (NIBIB); Grant numbers: 1R01EB006847; RO1EB000790; Grant sponsor: MIND Institute.

References

1. Barkhausen J, Ruehm SG, Goyen M, Buck T, Laub G, Debatin JF. MR evaluation of ventricular function: true fast imaging with steady-state precession versus fast low-angle shot cine MR imaging: feasibility study. Radiology. 2001;219:264–269. [PubMed]

2. Roemer PB, Edelstein WA, Hayes CE, Souza SP, Mueller OM. The NMR phased array. Magn Reson Med. 1990;16:192–225. [PubMed]

3. Wichmann T, Lanz T, Griswold MA, Greiser A, Nittka M, Jakob PM. Highly accelerated real-time imaging of cardiac function using a 32 channel phased array at 3T. Proceedings of the 14th Annual Meeting of ISMRM; Seattle, WA, USA. 2006. (Abstract 32)

4. Pruessmann KP, Weiger M, Scheidegger MB, Boesiger P. SENSE: sensitivity encoding for fast MRI. Magn Reson Med. 1999;42:952–962. [PubMed]

5. Sodickson DK, Manning WJ. Simultaneous acquisition of spatial harmonics (SMASH): fast imaging with radiofrequency coil arrays. Magn Reson Med. 1997;38:591–603. [PubMed]

6. Hayes CE, Hattes N, Roemer PB. Volume imaging with MR phased arrays. Magn Reson Med. 1991;18:309–319. [PubMed]

7. Yang QX, Wang J, Zhang X, Collins CM, Smith MB, Liu H, Zhu XH, Vaughan JT, Ugurbil K, Chen W. Analysis of wave behavior in lossy dielectric samples at high field. Magn Reson Med. 2002;47:982–989. [PubMed]

8. Wiesinger F, De Zanche N, Pruessmann KP. Approaching ultimate SNR with finite coil arrays. Proceedings of the 13th Annual Meeting of the ISMRM; Miami Beach, FL, USA. 2005. (page 672)

9. McDougall MP, Wright SM. 64-channel array coil for single echo acquisition magnetic resonance imaging. Magn Reson Med. 2005;54:386–392. [PubMed]

10. Wald LL, Moyher SE, Day MR, Nelson SJ, Vigneron DB. Proton spectroscopic imaging of the human brain using phased array detectors. Magn Reson Med. 1995;34:440–445. [PubMed]

11. Hardy CJ, Giaquinto RO, Piel JE, Rohling KW, Marinelli L, Fiveland EW, Rossi CJ, Park KJ, Darrow RD, Watkins RD, Foo TK. 128-channel body MRI with a flexible high-density receiver-coil array. Proceedings of the 15th Annual Meeting of ISMRM; Berlin, Germany. 2007. (page 244)

12. Niendorf T, Sodickson DK. Parallel imaging in cardiovascular MRI: methods and applications. NMR Biomed. 2006;19:325–341. [PubMed]

13. Wichmann T, Lanz T, Griswold MA, Greiser A, Nittka M, Jakob PM. Highly accelerated real-time imaging of cardiac function using a 32 channel phased array at 3T. Proceedings of the 14th Annual Meeting of ISMRM; Seattle, WA, USA. 2006. (Abstract 32)

14. Wintersperger BJ, Reeder SB, Nikolaou K, Dietrich O, Huber A, Greiser A, Lanz T, Reiser MF, Schoenberg SO. Cardiac CINE MR imaging with a 32-channel cardiac coil and parallel imaging: impact of acceleration factors on image quality and volumetric accuracy. J Magn Reson Imaging. 2006;23:222–227. [PubMed]

15. Schmitt M, Potthast A, Wiggins GC, Sosnovik DE, Triantafyllou C, Wald LL. A 128 channel receive-only cardiac coil for 3T. Proceedings of the 15th Annual Meeting of ISMRM; Berlin, Germany. 2007. (Abstract 245)

16. Wiggins GC, Triantafyllou C, Potthast A, Reykowski A, Nittka M, Wald LL. 32-channel 3 Tesla receive-only phased-array head coil with soccer-ball element geometry. Magn Reson Med. 2006;56:216–223. [PubMed]

17. Tropp J. Image brightening in samples of high dielectric constant. J Magn Reson. 2004;167:12–24. [PubMed]

18. Wiggins GC, Alagappan V, Potthast A, Schmitt M, Wiggins CJ, Fischer H, Jahns K, Benner T, Polimeni J, Wald LL. Design optimization and SNR performance of 3T 96 channel phased array head coils. Proceedings of the 15th Annual Meeting of ISMRM; Berlin, Germany. 2007. (page 244)