вход по аккаунту



код для вставкиСкачать
Magnetic Resonance Imaging 53 (2018) 164–172
Contents lists available at ScienceDirect
Magnetic Resonance Imaging
journal homepage:
Original contribution
Evaluation of stacked resonators to enhance the performance of a surface
receive-only array for prostate MRI at 3 Tesla
Jorge Chacon-Calderaa, , Alexander Fischera, Matthias Malzachera, Yannik Vettera,
Mathias Davidsa, Martina Flöserb, Christopher Stumpfc, Lothar R. Schada
Computer Assisted Clinical Medicine, Medical Faculty Mannheim, Heidelberg University, Mannheim, Germany
Medical Physics in Radiology, German Cancer Research Center (DKFZ), Heidelberg, Germany
Work Microwave, Holzkirchen, Germany
RF receive coils
Coil arrays
Stacked resonators
Prostate MRI
Prostate MRI is an important tool to diagnose and characterize cancer. High local sensitivity and good parallel
imaging performance are of paramount importance for diagnostic quality and efficiency. The purpose of this
work was to evaluate stacked resonators as part of a surface receiver array for prostate MRI at 3 Tesla. A base
array of 6-channels consisting of a flexible anterior and a rigid posterior part were built each with three loop
coils. A pair of stacked resonators was added concentrically to the center loops (anterior and posterior) of the
base array. The evaluated stacked resonators were butterflies, composites and dipoles which yielded a total of
three 8-channel arrays. The arrays were compared using noise correlations and single-channel signal-to-noise
ratio maps in a phantom. Combined signal-to-noise ratio maps and parallel imaging performances were measured and compared in vivo in 6 healthy volunteers. The results were compared to the base and a commercial
array. The SNR values in the prostate yielded by all the arrays were not statistically different using fully sampled
k-space. However, significant differences were found in the parallel imaging performance of the arrays. More
specifically, up to 88% geometric factor reduction was found compared to the commercial array and up to 83%
reduction compared to the base array using butterfly coils. Thus, signal-to-noise ratio improvements were observed with stacked resonators when using parallel imaging. The use of stacked elements, in particular butterfly
coils, can improve the performance of a base array consisting solely of single loops when using parallel imaging.
We expect prostate MRI at 3 Tesla to benefit from using combinations of single loops and stacked resonators.
1. Introduction
The high incidence of prostate cancer requires robust and reliable
methods to diagnose and characterize the disease. MRI has long been
recognized as an important diagnostic tool for this purpose.
Furthermore, it has recently been demonstrated that MRI can also be
used to rule out cases where prostate cancer is falsely suspected, thus
avoiding the need for biopsies [1]. The main pre-requisite for prostate
MRI is: high local sensitivity in the prostate with efficient acquisition
times to image the larger abdominal cavity. In terms of sensitivity, the
location of the prostate complicates the signal detection due to the coil
sensitivity drop as the distance from the coil increases. When using
surface coils, this distance is determined by the anatomical location of
the prostate with respect to the surface of the body. An approach to
overcome this limitation is the use of tailored endorectal coils which
can be placed near the prostate using the rectal cavity [2]. However,
there exists controversy in the literature regarding the advantages of
using this type of coils [3–7]. Moreover, they are uncomfortable for
patients. For these reasons, arrays of surface coils are often preferred
despite an arguable reduction in signal-to-noise ratio (SNR) in the
prostate. The limit in SNR has an impact on image quality and it could
compromise the diagnostic quality of the images. Therefore, it is important to maximize sensitivity of the surface arrays. Likewise,
achieving efficient scanning times, is of high clinical relevance. For this
Abbreviations: SNR, Signal-to-noise ratio; G-factor, Geometric factor; REF, Reference/base; COM, Commercial; B3SL, Base array with a pair of butterfly coils; C3SL,
Base array with a pair of composite coils; D3SL, Base array with a pair of dipole antennas; AP, Anterior-Posterior; LR, Left-Right; SD, Standard deviation; SENSE,
Sensitivity encoding; FOV, Field-of-view; ROI, Region-of-interest; A, Anterior; P, Posterior; L, Left; R, Right; S, Stacked; C, Center; Dir, Direction; Acc, Acceleration
Corresponding author at: Computer Assisted Clinical Medicine, Medical Faculty Mannheim, Heidelberg University, Theodor-Kutzer-Ufer 1-3, 68167 Mannheim,
E-mail address: (J. Chacon-Caldera).
Received 26 March 2018; Received in revised form 8 June 2018; Accepted 21 July 2018
0730-725X/ © 2018 Elsevier Inc. All rights reserved.
Magnetic Resonance Imaging 53 (2018) 164–172
J. Chacon-Caldera et al.
Fig. 1. Schematic design of the tested stacked resonators and their reception chain. Their position and dimensions compared to the center loop is shown (Top). The
stacked arrays are shown in a representative design and representative setups for phantom, and in vivo measurements (Bottom).
[29]. This approach has been more explored as a standalone element
[30] and in combination with single loop coils [31, 32]. Gains have
been demonstrated in arrays using both novel approaches combined
with single loop coils e.g. composite coils in simulations and measurements of a cuboid phantom [28] and dipole antennas in body imaging
at 7 T [32]. Moreover, using simulations the dipole has shown to have a
similar performance as the single loop in the target depth of the prostate at a range of magnetic field strengths [30].
The aim of this work is to experimentally investigate three different
stacked elements in a simple setup with low element count. This was
designed to facilitate the characterization of the elements and to test
their performance as part of a base array composed of single loop coils.
Moreover, a comparison to a commercial solution is provided to ultimately obtain gains in SNR when using parallel imaging in prostate MRI
at 3 T.
purpose, parallel imaging techniques are the most common approach
used in clinical. However, parallel imaging techniques yield an intrinsic
SNR penalty determined by the receiver array. This penalty is characterized in terms of the geometric factor (g-factor) and must also be
considered in the design and evaluation of RF coil arrays [8].
Standard approaches to the design of RF coil arrays rely on loop
coils decoupled by overlap [9–12]. The design in that case can be focused on placing the maximum number of single loop coils that fit an
area determined by the application [13–15]. This is feasible due to the
large number of available reception channels in modern MRI systems.
For body imaging, arrays of up to 128 single loop coils have been built
with limited SNR increase in the center of the body but improved
parallel imaging performance [16]. However, this approach entails
several technical limitations such as a reduction in individual performance of the coils due to coupling with next nearest neighboring coils,
higher contribution to noise from coil resistance and reduced penetration depth.
Several works have outlined that the maximum sensitivity in the
center of the body (calculated as a homogeneous cylinder) can be well
approximated with a limited number of single loop coils following the
calculations of ultimate intrinsic SNR [17, 18]. Nevertheless, it has been
proposed that additional resonators could complement the B1-field
components of the single loop coils and contribute further towards the
sensitivity of RF arrays. Therefore, a combination of these resonators
with single loops could yield SNR values closer to the ultimate intrinsic
SNR. The impact has been investigated theoretically by several authors
in simulations and experimental setups [19, 20]. Three main types of
resonators have been investigated: butterfly and composite coils as well
as dipole antennas. In this work we refer to these resonators as stacked
elements to simplify the coil/antenna terminology. These stacked elements also allow geometrical decoupling when placed concentrically
with loop coils. This represents an advantage over single loop coils
which can only be decoupled geometrically using controlled lateral
placement. Butterfly coils have been traditionally used in combination
with single loops in MRI for quadrature detection [21–26]. Composite
coils are single loop coils placed orthogonally with respect to the
standard single loop coils and they have been seldom investigated [27,
28]. Another recently introduced approach is the use of dipole antennas
2. Materials and methods
All methods were performed according to the relevant guidelines
and regulations. The in vivo prostate scans were approved by the local
Institutional Review Board and acquired with prior written informed
2.1.1. Arrays design
An array composed of a flexible anterior and a rigid posterior part
was used for the base of our analysis and also for reference (REF array).
Each part of the REF array was composed of three single loops decoupled by overlap. The dimensions of the three loops were
180 × 180 mm2 for the center loop and 120 × 180 mm2 for the side
loops. This resulted in a total dimension of 360 × 180 mm2 for each
part of the array. The dimensions of the center loop were experimentally optimized for the depth of the prostate. The side loops were reduced in size (x-axis) compared to the center loop to minimize coupling
between anterior and posterior parts of the array. The dimensions were
also chosen to achieve full loading of the loops to maintain sample
dominated noise. Stacked elements were added at the center of the two
Magnetic Resonance Imaging 53 (2018) 164–172
J. Chacon-Caldera et al.
height and 340 mm in maximum width (rcylinder = 110 mm,
Wcuboid = 120 mm). The phantom contained homogeneous material
properties with the dielectric parameters: ε′ = 50.3, ε″ = 66.6 and
σ = 0.46 S/m. Additionally, the measurements included in vivo scans of
healthy volunteers using a FOV of 200 × 200 mm2. 6 healthy male
volunteers aged 23 to 31 years (mean ± SD = 27.3 ± 3.1) participated in this study. The height and weight of the volunteers was
180.8 ± 4.6 cm and 75.5 ± 10.45 kg (mean ± SD), respectively. The
performance evaluation was carried out in the following three stages:
parts of the base array with a distance of approximately 2 mm from the
center loop coils. Both, anterior and posterior parts were analyzed using
the same type of stacked element. The three resultant 8-channel arrays
tested were: with butterflies (B3SL), with composites (C3SL) and with
dipoles (D3SL).
In the design of the stacked elements, each butterfly had a distance
of 3 mm between the lobes and a total dimension of 160 × 160 mm2 to
achieve overlap decoupling with the side loop coils (see Fig. 1 – Butterfly). The composites were built with a length of 180 mm. However,
the height of the two composites was different given the limited space
in the bore and the height of the housing. The anterior composite had a
height of 70 mm and the posterior 40 mm (see Fig. 1 – Composite). For
the dipole, a fractionated approach was taken [33]. Namely, the dipoles
were shortened using 2 inductors (air solenoids) which were inserted
equidistantly in each of the two segments. Solenoids with an inductance
of 237 ± 32 nF were chosen to tune the element to the resonance
frequency. This resulted in four segments each with a dimension of
30 × 75 mm2 (see Fig. 1 – Dipole).
2.2. Performance in phantom
Using phantom measurements, the individual-channel SNR yielded
by the stacked elements and the center loops were analyzed as part of
the arrays. This was done to investigate the sensitivity of the elements
and their effect on the performance of the center loop. For this, SNR
maps were calculated and profiles were plotted along the y-axis in sagittal slices. Noise datasets were used to generate the correlation matrix
of each array. These matrices were compared to scattering parameters
Flip angle maps were also plotted to assess the detuning of elements
during RF transmission. These tests included measurements without
receiver arrays for reference and with every tested array. For this, five
scans using a gradient echo sequence with flip angles of 0, 10, 30, 60
and 90° were acquired using TR = 5*T1 (2000 ms). To compute the flip
angle map, a sine function was fitted to the signal of the images obtained with the different flip angles in each pixel. The sine of the actual
flip angle was then used to assess the percentage of signal expected
considering the flip angle errors.
2.1.2. Arrays construction
The coils of the base array were built using copper tape of 3 mm in
width and 0.1 mm in height. Similar to the composites, three fixed
capacitors were used for splitting each coil; one additional variable
capacitor was used for the fine tuning of the coil. Split capacitors were
also used to connect and balance the coils. Fuses were added to all coils
and stacked elements for the safety of our volunteers. Each coil and
stacked element had its own receiver channel composed of a capacitive
matching network, a cable trap, and a low input impedance and low
noise preamplifier. An active detuning network was also incorporated
in each coil and element. Given the larger size of the butterfly due to the
extra lobe, an additional passive detuning circuit was also added for
these stacked elements [34]. In the case of the dipole, active detuning
was achieved by using a diode in parallel with the solenoid which
shorted the segments when forward biased. A fuse was also added in
series with the inductor between the segments in each of the two
conductors of the element. For all preamplifiers, each output was individually connected to the scanner using a standard ODU cable with
three cable traps. In terms of the arrays, the anterior part was made
flexible to adapt to the shape of the phantom and volunteers. This was
achieved by employing three layers. In the first layer, foam with 5 mm
thickness was used for the comfort of the volunteers. Two layers of FR-4
material with 0.5 mm in thickness were added to support the coils. The
posterior array was fixed and embedded in a box made of acrylic glass
tightly fitting the patient bed. The top layer was 5 mm in thickness and
made of polystyrene. Foam was additionally employed for comfort.
Thus, the minimum distance between the nearest coil and the surface of
a sample was ca. 10 mm for the anterior part and 15 mm for the posterior. A commercial solution (COM array) for prostate imaging was
also used for reference in the in vivo measurements. This setup included
18 channels divided into 12 channels of a spine array and 6 channels of
a flexible body array (Siemens Healthineers, Erlangen, Germany).
2.3. Array performance (in vivo)
The adaptive combination [36] of all the channels in the stacked,
the REF and the COM arrays was respectively used to evaluate array
sensitivity. SNR was calculated in the prostate using in vivo scans of 6
healthy volunteers with each of the stacked and reference arrays. A
manual 2D selection of the prostate was determined to evaluate SNR
and g-factors. Furthermore, g-factors were calculated using transversal
slices with both, the anterior-posterior (AP) and left-right (LR) direction
for undersampling. For this calculation, coil sensitivities were obtained
using the eigenvalue approach [37]. G-factor maps were calculated
based on the SENSE formalism [8]. Mean g-factor values were then
measured in the prostate. After the quantitative evaluation of g-factors
in the prostate, the g-factor maps were multiplied by a mask. The mask
was obtained by thresholding magnitude images and the multiplication
was used for qualitative evaluation only. Finally, the overall performance of each array at acceleration factors 2, 3 and 4 and both encoding directions was evaluated in terms of SNR. For this, mean SNR
values in the prostate were calculated, following (8), as:
SNRreduced = SNRfull/(g − factor∗√ (acceleration factor)).
Representative magnitude images for qualitative assessment were
also reconstructed using SENSE with a sum-of-square combination for
acceleration factors 2 and 3 in both encoding directions. ANOVA was
used to test the statistical differences between the arrays in terms of
SNR using fully sampled k-spaces and g-factors with acceleration factor
2.1.3. MRI measurements
The measurements were performed using a 3 T Siemens MAGNETOM Trio. A 2D turbo-spin-echo sequence was used for all acquisitions
with the following parameters: TE/TR = 95/3500 ms, flip
angle = 137°,
thickness = 3 mm,
size = 320 × 320,
bandwidth = 200 Hz/px, turbo factor = 27, echo trains per slice = 12
and echo spacing = 10.6 ms. Per measurement, two scans were acquired using this protocol with full k-space sampling with the second
being a noise scan (same parameters but with no excitation RF pulse).
All SNR calculations were performed following the standard method
presented by Kellman et al. [35]. Stacked elements were investigated as
part of the array in a first instance using a phantom using a FOV of
300 × 300 mm2. The phantom was composed of a cuboid with two half
cylinders on two sides. This structure was 500 mm in length, 220 mm in
3. Results
3.1. Performance (phantom)
3.1.1. Sensitivity
In regards to the individual channels of the array, the center single
loop coils showed higher SNR than any of the tested stacked elements in
Magnetic Resonance Imaging 53 (2018) 164–172
J. Chacon-Caldera et al.
Fig. 2. 2D sagittal views of single-channel SNR maps measured in the phantom (Top). The maps correspond to the center loop and the stacked element in each tested
array with full channel count (8-channel B3SL, C3SL and D3SL, 6-channel REF see Subsection 2.1.1). The lines in the SNR maps indicate the position where the SNR
plots were taken. Line plots illustrate the SNR variations of the tested stacked elements of each of the arrays (B3SL, C3SL and D3SL) and center loops including the
reference array (REF) according to the depth (Bottom). The depth of interest was zoomed for improved visualization.
the prescribed flip was approximately 8° when the COM setup was in
place. Of the stacked arrays, the dipole showed the maximum deviation
ca. 3%. Moreover, the reduction in signal due to errors in flip angle
maps was < 1.1% for all tested arrays (see Fig. 4).
the phantom at the depth of interest (100 mm). For instance, the
minimum SNR value of a center loop was 9.1 yielded by the posterior
one in the D3SL. The butterflies and the dipoles showed minimum SNR
values of 5.7 and 7.1 respectively. The composites showed the lowest
sensitivity at the depth of interest with a minimum SNR value of 4.4
and 5 (anterior and posterior, respectively). The difference in SNR between stacked elements and loop coils was more clearly observed in the
anterior part of the arrays. Additionally, the SNR values yielded by the
center loops were consistently higher for the B3SL compared to the rest
of the arrays in both anterior and posterior arrays at the depth of interest. Compared to the loop in the REF array, the SNR values of the
center loops in C3SL and D3SL were similar in the anterior array but
lower in the posterior array (see Fig. 2).
3.2. Array performance (in vivo)
3.2.1. Sensitivity
In our in vivo evaluation, the ANOVA test showed that the differences between all tested arrays were not significant (P = 0.77).
However, the mean SNR values of the stacked arrays were consistently
higher than those of the REF array. The array that yielded the closest
SNR to the REF was the C3SL which yielded a 2.7% higher mean SNR.
The greatest difference was 11% yielded by the D3SL compared to the
REF (mean SNRs: 9.43 and 8.53, respectively; see Fig. 5a). A greater
variability in SNR values was found amongst volunteers than amongst
arrays e.g. the maximum difference was 42% between volunteer 2 and 3
(mean SNRs: 10.72 and 5.51 respectively; see Fig. 5b). The highest
variability in SNR values of a single array was measured in the B3SL
(SD = 2.06). In contrast, the minimum variability was yielded by the
COM array (SD = 1.21). Qualitative prostate images yielded by all the
tested arrays can be observed in Fig. 5c–d.
3.1.2. Noise correlation
Between stacked elements and center loops, the lowest correlation
was found in the C3SL. The values were 5.6% for the anterior and 4.5%
for the posterior part of the array. The noise correlations in the D3SL
were 17.3%, and 30.4% (anterior and posterior). The highest overall
correlation was found between the center loop and the butterfly in the
posterior part of the array. This correlation value was 66.5%. However,
in the anterior part of the array, the correlation between the single loop
and the butterfly was considerably lower (28.9%). Likewise, the B3SL
showed higher overall coupling between the rest of the coils compared
to the D3SL and C3SL (see Fig. 3). The S-parameters showed higher
coupling of the center loops with the adjacent loops (S21 ≈ −12 dB)
than with any of the stacked arrays. Here, the maximum coupling was
the center loop with the butterfly in the anterior array of the B3SL
(S21 ≈ −19 dB) and the minimum was found similarly in the anterior
composite and dipole with the center loop (S21 ≈ −34 dB both). All Sparameters from the tested arrays are summarized in Table 1.
3.2.2. Parallel imaging
In this evaluation, statistical significance was found in the different
g-factors yielded by the tested arrays (e.g. P = 0.009/P = 0.0031 AP/
LR both with acceleration factor 2). Moreover, the stacked arrays
yielded g-factors in the prostate which were lower than the ones from
the REF array for all tested undersampling directions and acceleration
factors. The B3SL performed consistently better than the rest of the
tested arrays aside from one single case (acceleration factor 2 in the AP
direction) where its g-factor was only 0.2% higher than the one from
the C3SL. For low accelerations, the performances of all the arrays were
closer to each other than for higher acceleration factors. For instance,
with an acceleration factor 2, the mean g-factors of the tested stacked
3.1.3. Flip angle maps
For a prescribed 90° excitation, the actual flip angle without coils
was 94.4° at the region-of-interest (ROI). A maximum deviation from
Magnetic Resonance Imaging 53 (2018) 164–172
J. Chacon-Caldera et al.
Fig. 3. Noise correlations found in phantom for each
tested array. The coils and elements in the anterior
(A) and posterior (P) arrays were: left loop (L), right
loop (R), stacked element (S) and center loop (C).
The white square indicates the noise correlation
between the center loop and the stacked element.
The commercial array (COM) was composed of 6
channels of a surface body array (BO6) and 12
channels of the spine array (SP12).
Table 1
Scattering parameters of all tested arrays. The coils and elements were: left loop
(L), right loop (R), center loop (C) and stacked element (S).
arrays were within 6% difference (AP/LR: B3SL = 1.063/1.034,
C3SL = 1.061/1.037, D3SL = 1.071/1.039, REF = 1.116/1.055 and
COM = 1.092/1.076). However, for acceleration factor 3, the g-factor
of the worst performing array in AP (REF, g-factor = 2.42) was 44%
higher than the best performing one (B3SL, g-factor = 1.68). This trend
continued for acceleration factor 4 where the REF yielded a g-factor
61% higher than the B3SL in AP and 83% in LR. Furthermore, in LR, the
COM yielded the worst performance which peaked at an acceleration
factor 4 where the g-factor was 88% higher than the one of the B3SL
array. Between the other two stacked arrays, the C3SL performed better
than the D3SL in AP, particularly for an acceleration factor 3 where the
D3SL yielded a 1.11-fold higher mean g-factor. However, in LR, the
D3SL outperformed the C3SL by a maximum factor of 1.06 in acceleration factor 4. This can also be appreciated in representative g-factor
maps and mean g-factor plots (see Fig. 6). Mean g-factors in the prostate
and their standard deviations for all tested arrays, acceleration factors
and directions are summarized in Table 2.
In the calculation of SNR at different acceleration factors, the B3SL
and the D3SL showed similar performance in both AP and LR directions
Fig. 4. Flip angle maps for a prescribed 90° excitation and the signal expected
due to the actual flip angle in percentage (100% corresponds to optimal flip
angle). The tested setups were A: without receiver arrays, B: COM, C: REF, D:
B3SL, E: C3SL, F: D: D3SL.
for an acceleration factor 2 (SNR difference < 1%). However, the gains
in SNR were clearly higher for the B3SL for acceleration factors 3 (SNR
gains = 28 and 16%, AP and LR) and 4 (SNR gains = 24 and 30%, AP
and LR). Compared to the REF array, the maximum SNR gain (yielded
by the B3SL) was calculated to be larger than a factor of 2. The COM
array was also consistently outperformed by the B3SL. The range of the
gain in this case was from 2% (acceleration factor 2, AP) to up to 94%
(acceleration factor 4, LR; see Fig. 7). Sample reconstructions using
SENSE can be observed in Fig. 8.
Magnetic Resonance Imaging 53 (2018) 164–172
J. Chacon-Caldera et al.
Fig. 5. a) SNR performance of the arrays as measured in 6 volunteers. Mean and standard deviation (SD) values yielded by the arrays are displayed on the top. The
mean SNR values are also marked on the plot (+). b) Bar plot of SNR values found in the prostate of each of the scanned volunteers with the tested arrays. c)
Representative transversal images of the prostate acquired with all the tested arrays showing d) a zoom of the prostate.
Fig. 6. Representative g-factor maps yielded by all the tested arrays are displayed for an acceleration factor of 3 (Left). Mean g-factors calculated in the prostate of the
6 volunteers using anterior-posterior (AP) and left-right (LR) undersampling for different acceleration factors are also shown (Right).
Table 2
Mean g-factor values in the prostate of our 6 volunteers for all tested arrays.
Undersampling direction.
Acceleration factor.
Magnetic Resonance Imaging 53 (2018) 164–172
J. Chacon-Caldera et al.
part of the array where it lays on the volunteer. Air solenoids were
useful to shorten the dipole's length. The noise added by these inductors
is given by the effective series resistance [38] which is determined by
the copper of the inductor and was negligible in our case (< 0.01 Ω).
Our first evaluation of sensitivity was carried in phantom experiments using individual channels. Here, all the stacked elements showed
sensitivity at the depth of interest. This was observed in the SNR profile
plots which were clearly above the noise floor. This was expected from
previous works [28, 30]. However, their sensitivities were found to be
consistently lower than the one from the single loop coils. One different
dominant reason for this can be provided for each element. In the case
of the butterfly, the penetration depth is reduced as the number of lobes
increases [22]. The composite is a single loop coil which is well-known
to have its main B1-field component normal to its plane [19]. Lastly, the
dipole is mainly a radiative element (curl-free) but at the tested frequency, the signal detection is based on near-field induction (divergence-free currents dominate the contributions to the SNR [39]). In this
work, only the center loop coil was combined with a stacked element.
Stacked elements can be added to the rest of the single loops at the cost
of increased construction complexity and coupling. Furthermore, considering the longer distance of the side loop coils to the prostate and the
observed limited penetration depth of the stacked elements these additional stacked elements are not expected to have considerable sensitivity. The center loops showed variability with each of the stacked
elements. One potential reason for this is coupling e.g. the highest
coupling between stacked elements and center loops was found to be
between the butterfly and the single loop in the anterior part of the
B3SL using S-parameters. This could increase the mutual inductance
and affect the individual performance of the coils by adding extra
loading. Other possible sources of variations are slight differences in
tuning and matching, array positioning and variations in the position of
the selected line plot.
In terms of noise correlation, high values were found between the
stacked elements and the single loops despite lower coil coupling than
between the adjacent loops as observed in the S-parameters. This is
because noise correlation is not only dependent on coil inductance
(cross-talk between coils) but also on coil overlap as demonstrated by
Roemer et al. [40]. Considering the increased overlap, the noise correlation of the butterfly and the single loop coils in the B3SL was higher
than between the other stacked elements and their corresponding
center loop coils.
On the other hand, the composite showed the lowest overall noise
correlations and mutual inductances with the center loops. The dipoles
Fig. 7. Mean SNR values in the prostate of the 6 volunteers calculated for
different acceleration factors are plotted for both undersampling directions
anterior-posterior (AP) and left-right (LR).
4. Discussion
In this work, we evaluated three stacked elements added to a base
array of single loop coils. This evaluation was carried out in order to
test, in the first instance, whether SNR gains in the prostate could be
obtained from a) detecting B1-field components orthogonal to single
loop coils, b) increasing coil density without decreasing individual coil
size and c) from placing additional elements nearest to the prostate as
possible. Their parallel imaging performance and thus, their SNR with
parallel imaging was tested in the second instance. The stacked elements included in the base array were butterfly and composite coils as
well as dipole antennas. In the design process of the arrays, different
constraints regarding the dimensions of the elements had to be taken
into account. For the butterfly, the size of the lobes was determined by
the coils adjacent to the center loop which had a fixed position to
achieve overlap decoupling with the center loop. Here, the x-axis dimensions of the butterflies were chosen to achieve overlap decoupling
with the side coils. The height was the geometrical limitation imposed
by the composite coils. This was due to the limited space inside of the
bore and the housing of the posterior part of the array. The design of the
dipole was not as restrictive but its length could not be increased
without adding complexity to its positioning, especially in the anterior
Fig. 8. Representative parallel imaging reconstructions of all arrays are shown. Images were reconstructed using SENSE with acceleration factors of 2 and 3 using
both undersampling directions: anterior-posterior (AP) and left-right (LR).
Magnetic Resonance Imaging 53 (2018) 164–172
J. Chacon-Caldera et al.
demonstrated a similar behavior to that of the composites in terms of
noise correlations with the center loops. However, higher noise correlations were observed between the dipoles and the adjacent loops.
The impact of the higher noise correlations on SNR was minimized
by using a matched filter coil combination [18] in which the noise
correlation matrix is used to coherently weigh the coil sensitivities to
maximize SNR. In terms of parallel imaging, it has been shown in
previous works that the butterfly coils yields distinct phase information
[41] which seems to outweigh the higher noise correlations. Moreover,
it has been shown that noise correlations have limited impact on the
parallel imaging reconstruction [42].
In the flip angle maps, some coupling in the periphery was found
between the arrays and the transmission coil leading to flip angle deviations. However, at the region-of-interest, the signal decrease due to
flip angle errors was negligible in all tested arrays.
In the in vivo measurements, despite the lack of statistical differences, the SNR of the arrays that included stacked elements were
consistently higher than the base array. In comparison to the COM
array, the D3SL and B3SL showed similar sensitivity despite the higher
channel count of the COM array.
Unlike sensitivity for fully sampled k-spaces, the parallel imaging
performance was significantly different for each array especially as the
acceleration factor increased. In this evaluation, the B3SL showed the
best performance. The improved performance using butterfly coils
compared to a reference array of single loops is consistent with previous
studies [22]. Furthermore, in g-factors, the advantage of using stacked
elements was found to be even higher than the choice between the
types of element. For example, the g-factor difference between the B3SL
with acceleration factor 3 and REF with acceleration factor 2 was less
than the difference between B3SL and D3SL both with acceleration
factor 2 (20% and 29%, respectively).
The calculation of mean g-factors in the prostate allowed the evaluation of image quality degradation due to the tested arrays when
using parallel imaging. Here, it was found that the SNR of the B3SL was
between 12 and 15% higher than the one from the REF array for an
acceleration factor 2. In comparison to the COM array, both the B3SL
and the D3SL showed improved performances but the B3SL showed
higher gains. The parallel imaging reconstructions were performed
using standard methods; more advanced methods could improve the
image quality further with the benefit of the reduced g-factor found by
adding the stacked elements.
[1] Ahmed HU, El-Shater Bosaily A, Brown LC, Gabe R, Kaplan R, Parmar MK, et al.
Diagnostic accuracy of multi-parametric MRI and TRUS biopsy in prostate cancer
(PROMIS): a paired validating confirmatory study. Lancet
[2] Erturk MA, Tian J, Van de Moortele PF, Adriany G, Metzger GJ. Development and
evaluation of a multichannel endorectal RF coil for prostate MRI at 7T in combination with an external surface array. J Magn Reson Imaging 2016;43(6):1279–87.
[3] Costa DN, Yuan Q, Xi Y, Rofsky NM, Lenkinski RE, Lotan Y, et al. Comparison of
prostate cancer detection at 3-T MRI with and without an endorectal coil: a prospective, paired-patient study. Urol Oncol 2016;34(6):255.e7–255.e13.
[4] Heijmink SWTPJ, Fütterer JJ, Hambrock T, Takahashi S, Scheenen TWJ, Huisman
HJ, et al. Prostate cancer: body-array versus endorectal coil MR imaging at 3
T—comparison of image quality, localization, and staging performance. Radiology
[5] Kim BS, Kim T-H, Kwon TG, Yoo ES. Comparison of pelvic phased-array versus
endorectal coil magnetic resonance imaging at 3 Tesla for local staging of prostate
cancer. Yonsei Med J 2012;53(3):550–6.
[6] Barth BK, Cornelius A, Nanz D, Eberli D, Donati OF. Comparison of image quality
and patient discomfort in prostate MRI: pelvic phased array coil vs. endorectal coil.
Abdom Radiol (NY) 2016;41(11):2218–26.
[7] Gawlitza J, Reiss-Zimmermann M, Thörmer G, Schaudinn A, Linder N, Garnov N,
et al. Impact of the use of an endorectal coil for 3 T prostate MRI on image quality
and cancer detection rate. Sci Rep 2017;7:40640.
[8] Pruessmann KP, Weiger M, Scheidegger MB, Boesiger P. SENSE: sensitivity encoding for fast MRI. Magn Reson Med 1999;42(5):952–62.
[9] Wright SM, Wald LL. Theory and application of array coils in MR spectroscopy.
NMR Biomed 1997;10(8):394–410.
[10] Malzacher M, Kalayciyan R, Konstandin S, Haneder S, Schad LR. Sodium-23 MRI of
whole spine at 3 Tesla using a 5-channel receive-only phased-array and a wholebody transmit resonator. Z Med Phys 2016;26(1):95–100.
[11] Wiggins GC, Triantafyllou C, Potthast A, Reykowski A, Nittka M, Wald LL. 32channel 3 Tesla receive-only phased-array head coil with soccer-ball element geometry. Magn Reson Med 2006;56(1):216–23.
[12] Edelstein WA, Glover GH, Hardy CJ, Redington RW. The intrinsic signal-to-noise
ratio in NMR imaging. Magn Reson Med 1986;3(4):604–18.
[13] Schmitt M, Potthast A, Sosnovik DE, Polimeni JR, Wiggins GC, Triantafyllou C, et al.
A 128-channel receive-only cardiac coil for highly accelerated cardiac MRI at
3 Tesla. Magn Reson Med 2008;59(6):1431–9.
[14] Wiggins GC, Polimeni JR, Potthast A, Schmitt M, Alagappan V, Wald LL. 96-channel
receive-only head coil for 3 Tesla: design optimization and evaluation. Magn Reson
Med 2009;62(3):754–62.
[15] Keil B, Wald LL. Massively parallel MRI detector arrays. Massively parallel MRI
detector arrays. J Magn Reson 2013;229:75–89.
[16] Hardy CJ, Giaquinto RO, Piel JE, Rohling KW, Marinelli L, Blezek DJ, et al. 128channel body MRI with a flexible high-density receiver-coil array. J Magn Reson
Imaging 2008;28(5):1219–25.
[17] Lattanzi R, Sodickson DK. Ideal current patterns yielding optimal signal-to-noise
ratio and specific absorption rate in magnetic resonance imaging: computational
methods and physical insights. Magn Reson Med 2012;68(1):286–304.
[18] Schnell W, Renz W, Vester M, Ermert H. Ultimate signal-to-noise-ratio of surface
and body antennas for magnetic resonance imaging. IEEE Trans Antennas Propag
[19] Wang ZJ. Towards a complete coil array. Magn Reson Imaging 2008;26(9):1310–5.
[20] Guerin B, Villena JF, Polimeridis AG, Adalsteinsson E, Daniel L, White JK, et al. The
ultimate signal-to-noise ratio in realistic body models. Magn Reson Med 2016.
[21] Sapkota N, Thapa B, Lee Y, Kim T, Bisson EF, Shah LM, et al. Eight-channel decoupled array for cervical spinal cord imaging at 3 T: six-channel posterior and twochannel anterior array coil. Concepts Magn Reson Part B 2016;46B(2):90–9.
[22] Ohliger MA, Greenman RL, Giaquinto R, McKenzie CA, Wiggins G, Sodickson DK.
Concentric coil arrays for parallel MRI. Magn Reson Med 2005;54(5):1248–60.
[23] Versluis MJ, Tsekos N, Smith NB, Webb AG. Simple RF design for human functional
and morphological cardiac imaging at 7 tesla. J Magn Reson 2009;200(1):161–6.
[24] Ouhlous M, Moelker A, Flick HJ, Wielopolski PA, de Weert TT, Pattynama PM, et al.
Quadrature coil design for high-resolution carotid artery imaging scores better than
a dual phased-array coil design with the same volume coverage. J Magn Reson
Imaging 2007;25(5):1079–84.
[25] Alfonsetti M, Clementi V, Iotti S, Placidi G, Lodi R, Barbiroli B, et al. Versatile coil
design and positioning of transverse-field RF surface coils for clinical 1.5-T MRI
applications. Magn Reson Mater Phys 2005;18(2):69–75.
[26] Kumar A, Bottomley PA. Optimized quadrature surface coil designs. Magn Reson
Mater Phys 2008;21(1–2):41–52.
[27] Wang ZJ. Improving SNR of RF coils using composite coil elements. NMR Biomed
[28] Maunder A, Fallone BG, Daneshmand M, De Zanche N. Experimental verification of
SNR and parallel imaging improvements using composite arrays. NMR Biomed
[29] Raaijmakers AJ, Ipek O, Klomp DW, Possanzini C, Harvey PR, Lagendijk JJ, et al.
Design of a radiative surface coil array element at 7 T: the single-side adapted dipole antenna. Magn Reson Med 2011;66(5):1488–97.
[30] Raaijmakers AJ, Luijten PR, van den Berg CA. Dipole antennas for ultrahigh-field
body imaging: a comparison with loop coils. NMR Biomed 2016;29:1122–30.
[31] Shajan G, Mirkes C, Buckenmaier K, Hoffmann J, Pohmann R, Scheffler K. Three-
5. Conclusions
In conclusion, our results demonstrated that the use of stacked
elements can improve the performance of a base array consisting solely
of single loops. Moreover, it was shown that the choice of stacked
elements between dipoles and butterflies yields small differences in
SNR but could have considerable impact on parallel imaging performance. The complexity of the design was negligible and the advantages
were clear when using parallel imaging methods. Considering SNR with
parallel imaging reconstructions, the B3SL showed the best performance overall. However, the other elements showed advantages in
design and decoupling that could also be exploited. Moreover, novel
methods to combine elements while minimizing the distance to the ROI
could provide further advantages [43]. These approaches must also be
investigated and put into perspective compared to traditional designs.
In any case, stacked elements could become a consideration in the
design of coil arrays for prostate imaging at 3 T.
The authors wish to thank Andreas Bitz and Nicolas Behl for their
assistance. This research did not receive any specific grant from funding
agencies in the public, commercial, or not-for-profit sectors.
Magnetic Resonance Imaging 53 (2018) 164–172
J. Chacon-Caldera et al.
layered radio frequency coil arrangement for sodium MRI of the human brain at
9.4 Tesla. Magn Reson Med 2016;75(2):906–16.
Erturk MA, Raaijmakers AJ, Adriany G, Ugurbil K, Metzger GJ. A 16-channel
combined loop-dipole transceiver array for 7 Tesla body MRI. Magn Reson Med
Raaijmakers AJ, Italiaander M, Voogt IJ, Luijten PR, Hoogduin JM, Klomp DW,
et al. The fractionated dipole antenna: a new antenna for body imaging at 7 Tesla.
Magn Reson Med 2016;75(3):1366–74.
Edelstein WA, Hardy CJ, Mueller OM. Electronic decoupling of surface-coil receivers for NMR imaging and spectroscopy. J Magn Reson 1986;67(1):156–61.
Kellman P, McVeigh ER. Image reconstruction in SNR units: a general method for
SNR measurement. Magn Reson Med 2005;54(6):1439–47.
Walsh DO, Gmitro AF, Marcellin MW. Adaptive reconstruction of phased array MR
imagery. Magn Reson Med 2000;43(5):682–90.
Uecker M, Lai P, Murphy MJ, Virtue P, Elad M, Pauly JM, et al. ESPIRiT–an
eigenvalue approach to autocalibrating parallel MRI: where SENSE meets GRAPPA.
Magn Reson Med 2014;71(3):990–1001.
Kumar A, Edelstein WA, Bottomley PA. Noise figure limits for circular loop MR
coils. Magn Reson Med 2009;61(5):1201–9.
Lattanzi R, Wiggins GC, Zhang B, Duan Q, Brown R, Sodickson DK. Approaching
ultimate intrinsic signal-to-noise ratio with loop and dipole antennas. Magn Reson
Med 2018;79(3):1789–803.
Roemer PB, Edelstein WA, Hayes CE, Souza SP, Mueller OM. The NMR phased
array. Magn Reson Med 1990;16(2):192–225.
Hajnal J, Larkman D, Herlihy D. An array that exploits phase for SENSE imaging.
Proc Intl Sot Mag Reson Med 2000;8:1719.
Ohliger MA, Ledden P, McKenzie CA, Sodickson DK. Effects of inductive coupling
on parallel MR image reconstructions. Magn Reson Med 2004;52(3):628–39.
Chacon-Caldera J, Malzacher M, Schad LR. Partially orthogonal resonators for
magnetic resonance imaging. Sci Rep 2017;7:42347.
Без категории
Размер файла
2 968 Кб
mri, 010, 2018
Пожаловаться на содержимое документа