- Email: [email protected]
Signal-to-noise ratio comparison of phased-array vs. implantable coil for rat spinal cord MRI
Magnetic Resonance Imaging 25 (2007) 1215 – 1221
Signal-to-noise ratio comparison of phased-array vs. implantable coil for rat spinal cord MRIB Andrew C. Yung4, Piotr Kozlowski University of British Columbia MRI Research Centre, Vancouver, BC Canada V6T 2B5 Received 2 October 2006; revised 8 January 2007; accepted 8 January 2007
Abstract The signal-to-noise ratio (SNR) performance and practicality issues of a four-element phased-array coil and an implantable coil system were compared for rat spinal cord magnetic resonance imaging (MRI) at 7 T. MRI scans of the rat spinal cord at T10 were acquired from eight rats over a 3 week period using both coil systems, with and without laminectomy. The results demonstrate that both the phased array and the implantable coil systems are feasible options for rat spinal cord imaging at 7 T, with both systems providing adequate SNR for 100-Am spatial resolution at reasonable imaging times. The implantable coils provided significantly higher SNR, as compared to the phased array (average SNR gain of 5.3 between the laminectomy groups and 2.5 between the nonlaminectomy groups). The implantable coil system should be used if maximal SNR is critical, whereas the phased array is a good choice for its ease of use and lesser invasiveness. D 2007 Elsevier Inc. All rights reserved. Keywords: Implantable coil; Phased array; Rat spinal cord; MRI
1. Introduction Magnetic resonance imaging (MRI) has been useful in longitudinally tracking pathological changes in rat models of mechanical spinal cord injury and other neurodegenerative diseases involving the spinal cord, using a variety of techniques such as T1- and T2-weighted imaging [1,2], proton spectroscopy [3], diffusion tensor imaging [4] and T2 relaxometry [5]. High signal-to-noise ratio (SNR) is required to meet the stringent data quality requirements of such techniques and to maintain adequate spatial resolution in the spinal cord (typically only a few millimeters square in the transverse plane). The most reported coil strategy to achieve increased SNR in animal spinal cord is the implantable coil system, where an internal coil is surgically placed over the spinal column and inductively coupled to a pickup coil outside the body [3,6–11]. The SNR advantage of implantable coils over single-element surface coils is clearly shown in the literature, with SNR gains ranging from 2.2 to 4.8 [6,7].
B This work was funded by the Rick Hansen Man-in-Motion Research Fund. 4 Corresponding author. Tel.: +1 604 822 6938; fax: +1 604 827 3973. E-mail address: [email protected] (A.C. Yung).
0730-725X/$ – see front matter D 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.mri.2007.01.006
Multiple-element phased arrays are known to produce SNR comparable with small surface coils within a relatively large field of view (FOV) [12]. These coils, commonly used on clinical and research human MRI scanners, are becoming increasingly popular for animal brain and spine MRI [13,14]. The objective of this study was to compare SNR performance of a phased-array coil vs. an implantable coil for rat spinal cord MRI. If phased array coils can achieve comparable SNR in the spinal cord, they may be an attractive alternative to implantable coils due to their relative ease of use and reduced invasiveness. MRI scans of the rat spinal cord at T10 were acquired at 7 T from eight rats over a 3-week period, using both coil systems, with and without laminectomy. The SNR performance and practicality issues between the two coil systems were compared. 2. Methods 2.1. Phased-array coil implementation A four-element receive-only phased array was built for rat brain and spinal cord imaging (National Research Council, Winnipeg, Canada). The coil elements are copper traces etched on a printed circuit board, are rectangular in shape (individually 827 mm) and are overlapped along
1216
A.C. Yung, P. Kozlowski / Magnetic Resonance Imaging 25 (2007) 1215 – 1221
Fig. 1. The phased-array coil for rat brain/spine MRI. (a) The array consists of four decoupled coil elements on a curved shell (bCQ), a floating cable balun (bBQ) and a matching/tuning box (bM/TQ). (Inset) shows coil arrangement. (b) Phased array circuit diagram.
their short axis on a curved plastic shell, covering a lateral width of 20 mm (see Fig. 1A). The length of 27 mm was chosen to ensure coverage of at least two vertebral levels. The 8-mm width was chosen to provide optimal signal reception at roughly 8 mm below the coil, where the spinal cord at T10 is typically located. The circuit diagram of the phased array is shown in Fig. 1B. Neighboring elements are geometrically decoupled through coil overlap, while isolation between nonadjacent elements is achieved by capacitive decoupling. PIN diodes turn on LC traps that detune the array during the transmission pulse, which is provided by an 11-cm internal diameter quadrature volume coil (Bruker Biospin, Ettlingen, Baden-Wuerttemberg, Germany). Variable capacitors situated one wavelength away from the coil elements allow adjustment of the tuning and matching inside the magnet, and a floating cable balun reduces current on the cable shields. The nuclear magnetic resonance (NMR) signals detected by each coil element are fed into the system preamplifiers and combined in a sum-of-squares reconstruction to form the composite phased-array image. 2.2. Implantable coil system implementation The implantable coil system (Fig. 2) consisted of a small internal coil inductively coupled to a pickup coil placed above it on the body surface. The implantable coil (previously described in [4]) consisted of a 1610-mm rectangular loop made from AWG 20 copper wire, with curved arches on each end to allow the implant to fit over the long axis of the spinal column. Series chip capacitors (700A series, American Technical Ceramics, Huntington Station, NY, USA) equaling 8.6 pF were soldered to one of the coil arches to achieve circuit resonance, while insulation and biocompatibility is achieved by coating the coil with polyolefin heatshrink and biocompatible elastomer (MDX4210 from Dow Corning, Midland, MI, USA). The pickup coil was a 3-cm-diameter circular surface coil made from copper tubing, which is tuned and matched to the system preamplifier’s 50-V impedance with a capacitive network. One important consideration is the bresonance splittingQ in the coil system’s frequency response. Terman [15] has
shown that the mutual inductance between two resonant circuits produces two impedance minima, with the amount of peak separation dependent on the severity of sample loading and the degree of mutual inductance. Tuning either peak to the Larmor frequency will enable efficient reception of the NMR signal; however, the spatial B1 profile depends on the choice of resonant peak, and therefore, the coil designer should select the one that best fits the region of interest. Since the low-frequency peak reportedly gives better SNR in the cord region [3], the implantable coil resonance was tuned to 310.5 MHz (in air) so that the in vivo low-frequency resonance is tuned to the Larmor frequency 300.3 MHz. This overcoupled resonance splitting exists even if the self-resonant frequencies of the isolated implant, and isolated pickup is slightly unequal, which provides some capability of retuning the system if the implantable coil resonant frequency drifts over time in vivo [11], especially in longitudinal studies. However, practical experience showed that a mismatch between the selfresonant frequencies of the implant and pickup reduces SNR significantly. Therefore, effort was made to maximize SNR by ensuring that the in vivo self-resonant frequencies were equal, which is indicated by equal peak heights in the pickup coil spectrum, as seen on a network analyzer.
Fig. 2. The implantable coil system. The implantable coil (foreground) is inductively coupled with a 3-cm-diameter pickup coil (background) outside the body.
A.C. Yung, P. Kozlowski / Magnetic Resonance Imaging 25 (2007) 1215 – 1221
1217
Fig. 3. Combined phased-array image generated by sum of squares reconstruction (far left) and individual images generated by the array elements.
Prior to implantation, the implantable coil was sterilized in ethanol for at least 2 h and rinsed in sterile saline. A midline exposure of the spinal column and adjoining rib junctions was first made at the T9-T11 level. The implantable coil was then placed over the column, and each corner was sutured to the nearest rib. This step was taken to ensure long-term immobilization of the implantable coil, since our initial experience showed that unsutured implants had a tendency to migrate away from the spinal cord because of swelling and formation of a tissue capsule around the implant during long-term implantation. The overlying paraspinal muscles and skin were then sutured to close the wound. 2.3. Experiments Quality assurance tests were performed on the coil systems by measuring quality factor (unloaded and loaded) using a network analyzer. The cross-talk between phasedarray elements was also measured. All animal experimental procedures were carried out in compliance with the guidelines of the Canadian Council for Animal Care and were approved by the institutional Animal Care Committee. The spinal cords of female Sprague– Dawley rats were individually imaged four times over a period of 3 weeks (time points at approximately 1, 7, 14 and 21 days) using a 7-T, 30-cm bore MR scanner (Bruker, Germany) for comparing SNR performance of the two coil systems. Four rats were imaged for each coil system. For each coil system, two of the rats underwent laminectomy while the remaining two did not. Laminectomies were
Fig. 4. Typical FLASH images of rat spinal cord at T10 level at 1001001000-Am resolution, acquired with (A) implantable coil and (B) phased array.
performed to mimic the increased invasiveness and heavier loading that may occur after an experimental spinal cord injury; the spatial patterns produced by actual mechanical trauma to the cord were deemed to be too variable to allow consistent comparison of SNR. Two implantable coils were constructed and first implanted into the nonlaminectomized rats. After this group’s time series was completed, the same implantable coils were recoated, resterilized and reused in the laminectomized rats. Before reimplantation, the quality factors of the implantable coils were verified to be unchanged. For the MRI experiments, the rats were placed supine on a plastic halfshell with the T10 region centered on the phasedarray elements or pickup coil. Anesthesia was maintained with isoflurane in air (4% induction, 1–2% maintenance), body temperature was maintained and monitored with a heated water pad and rectal probe, respectively and the respiration pattern was monitored with a pneumatic pressure sensor (SA Instruments, Stony Brook, NY, USA). A series of scout images were used to prescribe axial slices centered about the T10 level of the spinal cord. All rats were imaged with a multislice gradient-echo Fast Low Angle SHot (FLASH) sequence (TR/TE = 250/8 ms, FOV= 2.56 cm, 256256 matrix, slice thickness =1 mm, NA=10, 10 slices), resulting in a voxel size of 1001001000 Am. The sequence was respiration-triggered to reduce motion artifact; the resultant TR varied between 850 and 1000 ms due to
Fig. 5. Time evolution of in vivo SNR in individual rats over a 3-week period. The implantable coil measurements show higher SNR than the phased array measurements. Note that there is a significant difference in the implantable coil groups (with and without laminectomy), while there are no differences in the phased-array groups.
1218
A.C. Yung, P. Kozlowski / Magnetic Resonance Imaging 25 (2007) 1215 – 1221
Table 1 Group statistics of SNR measurements Group
Mean SNR
jSNR
Variability (jSNR/mean SNR)
Phased array without laminectomy Phased array with laminectomy Implant without laminectomy Implant with laminectomy
37 28 94 148
5.3 6.7 11.0 25.7
14% 24% 12% 17%
spontaneous changes in breathing rate, resulting in a total acquisition time of 36–43 min. The transmission power for the phased-array scans was automatically set to achieve a flip angle of 608 (previously determined as the Ernst angle). For the implantable coils, the transmitter power was manually adjusted to provide maximum signal in the centre of the cord, since the scanner’s automatic algorithm does not provide accurate flip angles for inhomogeneous transmit fields. SNR was estimated at the T10 level as the ratio between the signal mean in the spinal cord and the standard deviation of a noise region. Care was taken to exclude pixels in the cord that may suffer from partial volume effects with surrounding structures. In cases where the slice centered at T10 was corrupted by susceptibility artifacts at the cord periphery (a result of the cord/bone interface), an adjacent slice free of these artifacts was selected for SNR analysis. The SNR values require multiplication by correction factors to account for the noise bias introduced by the magnitude reconstruction. The amount of noise bias is slightly different for the phased array due to the sum-of-squares reconstruction of the four individual coil elements [16]. True SNR values were calculated by multiplying the implantable coil SNR values by 0.655 [17], while the phased-array SNR values required a correction factor of 0.695 [16]. Each of the four experimental groups (i.e., with and without laminectomy using the implantable coil, with and without laminectomy using the phased array) included two rats, with four measurements each over a 3-week period. All time measurements in each experimental group were pooled together (N = 8) to generate group means and standard deviations. A Tukey-Kramer test [18] was performed on the four groups to determine if any of the differences between means were statistically significant. To test whether SNR was independent of time, a straight line was first fitted through each group’s SNR vs. time data, with a slope of zero considered to be evidence that SNR was independent
from time. Random variability of the data can confound an observation of zero slope, so a t test was performed to calculate the probability that any nonzero slope was due to chance alone [19]. 3. Results Network analyzer measurements showed an interelement isolation for the phased array in the range of 16.8–29 dB, with only one measurement below 20 dB, thereby indicating acceptable levels of noise correlation between elements. Q air/Q in vivo was 178/84 and 210/112 for the phased-array and implantable coil system, respectively. The rats seemed to tolerate the implantable coils reasonably well and exhibited no behavioral deficit. In one rat, the implantation resulted in swelling around the surgery site, which disappeared within the first week after the surgery. Fig. 3 shows a representative phased-array image generated by sum-of-squares reconstruction, along with the images individually generated by each coil element. The phased array SNR is 41 in the spinal cord, whereas the individual elements alone generated cord SNR values of 20, 27, 20 and 15 from left to right, respectively. It is clear that the combination of four 827-mm coil elements provides a higher SNR than is achievable by a single 827-mm coil alone. Fig. 4 compares typical axial FLASH images acquired at T10 level using the implantable coil (left) and the phased array (right). Both images are at 100-Am pixel resolution and clearly elucidate structures within the spinal cord, such as blood vessels and the characteristic butterfly pattern of the central gray matter. For the implantable coil measurements, the mean distance between the centre of the cord and the implantable coil wires was 1.8F0.2 mm, whereas the mean distance between the cord centre and the phased array elements was 11.2F0.9 mm. The serial SNR measurements for each rat are presented in Fig. 5, while the overall results are summarized in Table 1. Fig. 6 shows images at different time points (t = 1, 5, 13, 19 days) acquired from the rat that showed the greatest SNR variability when using the implantable coil. The SNR variability (ratio between the standard deviation of SNR and mean SNR) ranged from 12% to 24%; none of the groups showed sufficient evidence that the variability was caused
Fig. 6. Longitudinal series of images on laminectomized rat using implantable coil (t = 1, 5, 13, 19 days). Images do not show SNR degradation due to longterm implantation of coil. Intensity levels are scaled identically in all images.
A.C. Yung, P. Kozlowski / Magnetic Resonance Imaging 25 (2007) 1215 – 1221
by a linear dependence between SNR and time, suggesting that SNR and time are independent. Overall, the implantable coils in our study produced better SNR than the phased array. The implantable coils with and without laminectomy performed better than the phased array by a factor of 5.3 and 2.5, respectively; these differences were statistically significant ( P b.001). There was no statistically significant difference between the phased-array groups ( P N.25). Conversely, there was a statistically significant difference in the implant coil groups ( P b.001): the average SNR of the laminectomy group was 1.6 times higher than that of the nonlaminectomy group. 4. Discussion Both the phased-array and implantable coil systems produced good in vivo anatomical images of the T10 spinal cord at 100-Am pixel resolution and 1-mm slice thickness at 7 T. Variability in each group ranged from 12% to 24%, which seems reasonable, considering the experimental variability in repeated in vivo scans. Although both coil systems have proven to be viable alternatives to highresolution imaging of the spinal cord, there are distinct differences between the coil systems that may lead an investigator to choose one system over the other. The phased-array coil does suffer from lower SNR in the T10 spinal cord, primarily as a result of lower proximity to the region of interest. However, one major advantage of the phased array is that it does not require surgery or long-term implantation, thereby reducing invasiveness and possible interference to the experimental data. The phased array is also more versatile in the sense that it can be moved from session to session to image different parts of the spinal cord, whereas the implantable coil’s location is fixed by the implantation. In fact, relative performance of the phased array may improve for regions in the lumbar spine, where the spinal cord lies closer to the body surface. Finally, the use of a multielement phased array opens up the possibility of parallel imaging, which can be used to accelerate data acquisition. Improvements to the phased array may increase its SNR performance; however, the phased-array coil used in this study represents a practical implementation that an investigator might typically construct for general rat spine and brain imaging. Integrated preamplifiers on the receive channels may reduce noise correlation between elements; however, resultant SNR increase may be marginal since the measured interelement isolation already appears adequate ( N20 dB for all but one pair). The element geometries could be optimized to achieve the best SNR at a certain depth below the array; however, the flexibility of the coil to image other parts of the spine and brain may be reduced. There was no significant difference in phased-array SNR between the laminectomy and nonlaminectomy groups. This indicates that laminectomy (and, by suggestion, spinal cord injury) does not add any significant long-term loading
1219
effects on the phased array. The phased-array SNR may possibly be lower in the immediate phase following surgery or injury because of increased loading that comes with initial inflammation. However, measurements at these early time points were not taken in order to allow the animals to recover from anesthesia. The most significant advantage of the implantable coil system is its higher SNR [9], outperforming the phased array by a factor of 5.3 and 2.5 for the laminectomy and nonlaminectomy groups, respectively. The results are on the same order of magnitude as other published improvement factors between implantable coils and single-element surface coils, as reported by Wirth et al. [7] (gain of 2.2) and Ford et al. [6] (gain of 4.8). The in vivo SNR comparisons presented here also correlate with a preliminary study, which showed that the implantable coil performed 2.4 better than the phased array in a dead rat [20]. Although the in situ comparison is not exactly equivalent to the in vivo case (e.g., different relaxation properties, untriggered acquisition and the absence of motion), it may reflect coil performance under more favourable imaging conditions. The higher implantable coil SNR highlights the advantage of using a small coil near its region of interest, which limits the volume of reception sensitivity but also decreases the amount of noise detected. Because of the ability to place the receiver coil very close to the spinal cord regardless of its depth below the surface, the relative SNR advantage over the phased array may increase for deep cord regions such as the cervical spine, which is an important site used in experimental spinal cord injury. The SNR gain is somewhat reduced by the increased dielectric loss due to enveloping the implant coil inside the conductive sample at high fields; however, this loss was ameliorated by insulating the wires with polyolefin heatshrink. Addition of a second capacitor junction would provide more distributed capacitance [8,9]; however, initial experiments showed that the long-term mechanical ruggedness in the coils was greatly reduced, while the SNR gain due to the higher dielectric isolation was only 10–15% (data not shown). Other practical advantages of the implantable coil system include easier localization of the region of interest and less problems with image wraparound due to a limited field of sensitivity. The major drawback of the implantable coil system is its invasiveness. However, the implantation does not add too much additional insult if an experimental spinal cord injury is already being performed. In addition, experience from our research and others [7,8] indicate that the implants are generally well-tolerated without significant inflammation. Furthermore, the implantable coil system requires significant technique refinement in order to optimize the SNR. Automatic power adjustment fails for the implantable coil system, and manual adjustment of the transmission power requires user experience and is approximate, at best. Mechanical ruggedness must be ensured by strong solder joints and uniform application of the biocompatible elastomer, since repair is extremely impractical after
1220
A.C. Yung, P. Kozlowski / Magnetic Resonance Imaging 25 (2007) 1215 – 1221
implantation. Suturing the corners to the ribs must be done carefully in order to prevent migration of the coil away from the cord, since we noticed that the SNR in the spinal cord can drop by as much as 50% for a 1-mm increase in separation distance (data not shown), even after appropriate power adjustments to optimize the SNR in the spinal cord. Figs. 5 and 6 illustrate that these precautions were successful in preventing the systematic degradation in implantable coil SNR over time, which has been reported by other investigators [8]. The implantable coil system may also suffer from higher signal inhomogeneity as compared to the phased array, since the implantable coil is small in size and acts as a receive– transmit system (the phased array is a receive-only device and relies on the volume coil to provide a homogeneous excitation). This effect is more noticeable in spin-echo images [11]. Fig. 7 compares phased array performance with implantable coil performance using spin-echo Rapid Acquisition with Relaxation Enhancement data from a separate MRI study on contusion-injured rats (RARE factor = 8, TR/TE = 5000/ 37 ms, FOV= 4 cm, 256256 matrix, slice thickness = 1 mm, NA=4, 10 slices). A signal dropoff was observed at the end of the implantable coil (Slice 1), which is not seen in the phased array image series. The rest of implantable coil slices show good uniformity within the spinal cord, with a clear advantage in SNR over the phased array.
Another issue that adversely affects the implantable coil performance is the difficulty in choosing the unloaded (air) resonant frequency of the implant that would achieve the best SNR in vivo. It is true that overcoupling the implantable coil to the pickup coil gives some flexibility to the choice of self-resonant frequencies, since the tuning and matching circuit of the external coil can be used to adjust the position and height of the coupled resonances, therefore maintaining imaging performance in longitudinal studies even when the resonant frequency drifts [11]. However, as mentioned in a previous section, the best performance occurs when the amplitudes of the two resonant peaks are equal, with the low-frequency peak resonating at the Larmor frequency. In order to perfectly tune the implant on the bench to achieve this condition, the peak separation in vivo would need to be known in advance. Since this a priori knowledge is generally not available (due to variability in experiment geometries and loading conditions), the unloaded implant frequency is chosen as a best guess based on previous experience. As a result, the system resonance was not optimal in many of our experiments, primarily due to electronic mismatch (i.e., response at Larmor frequency not matched to 50 V) with an estimated SNR decrease of 15% or less in most cases. In our study, the SNR measurements for the implantable coils with laminectomy showed significantly higher values
Fig. 7. RARE spin-echo images from contusion-injured rats showing differences in signal homogeneity when using implantable coil (top) and phased array (bottom).
A.C. Yung, P. Kozlowski / Magnetic Resonance Imaging 25 (2007) 1215 – 1221
than those without laminectomy. This result is somewhat surprising and counterintuitive. One would expect that foregoing the laminectomy would produce SNR that was similar to or even higher than the laminectomy case, which was the case with the phased-array measurements. We do not believe that this result reflects a real effect of the laminectomy on SNR, but rather, this SNR difference is indicative of systematic differences that are difficult to control between sets of measurements using the implantable coil system. The uncertainty in pretuning the implantable coil on the bench has already been mentioned; this and other differences in coil preparation between individual implantations are especially prone to bias the SNR in a set of measurements in one rat relative to another. The level of expertise in the coil implantation may play a role as well; the laminectomy group implantations were performed by a more experienced animal technician than the nonlaminectomy group. We also noticed that the split resonant peaks in the nonlaminectomy measurements were generally broader and closer together than for the laminectomized rats, suggesting that the mutual coupling between coil elements was consistently different. It remains to be determined whether this phenomenon reflects an underlying cause that decreases SNR. Clearly, more understanding into the electromagnetic interactions in the implantable coil system is needed to explain the difference between the implantable coil measurements. However, these differences in SNR between implantable coils do not invalidate the general conclusion from our study that the implantable coils perform better than phased-array coils in terms of SNR.
5. Conclusion Our results demonstrate that both the phased-array and implantable coil systems are feasible options for rat spinal cord imaging at 7 T, with both systems providing adequate SNR for 100-Am spatial resolution at reasonable imaging times. The implanted coils provided significantly higher SNR as compared to the phased array. The phased array is a good choice if less invasiveness and more simplicity is desired, whereas the implantable coil can be used if maximal SNR is critical.
Acknowledgments The phased array was designed and constructed by Vyacheslav Volotovskyy (NRC Institute for Biodiagnostics). The authors would like to thank Caroline Hall, Katie Carr and Jie Liu for expertise in coil implantation and surgeries.
1221
References [1] Fraidakis M, Klason T, Cheng H, Olson L, Spenger C. Highresolution MRI of intact and transected rat spinal cord. Exp Neurol 1998;153(2):299 – 312. [2] Weirich SD, Cotler HB, Narayana PA, Hazle JD, Jackson EF, Coupe KJ, et al. Histopathologic correlation of magnetic resonance imaging signal patterns in a spinal cord injury model. Spine 1990;15(7): 630 – 8. [3] Silver X, Ni WX, Mercer EV, Beck BL, Bossart EL, Inglis B, et al. In vivo 1H magnetic resonance imaging and spectroscopy of the rat spinal cord using an inductively-coupled chronically implanted RF coil. Magn Reson Med 2001;46(6):1216 – 22. [4] Ford JC, Hackney DB, Alsop DC, Jara H, Joseph PM, Hand CM, et al. MRI characterization of diffusion coefficients in a rat spinal cord injury model. Magn Reson Med 1994;31(5):488 – 94. [5] Yung AC, Choo A, Liu J, Tetzlaff W, Oxland T, Kozlowski P. Measurements of myelin water in rat spinal cord in vivo. Book of abstracts: Twelfth Annual Meeting of the International Society of Magnetic Resonance in Medicine, vol. 1. Kyoto, Japan7 ISMRM; 2004. p. 1537. [6] Ford JC, Hackney DB, Joseph PM, Phelan M, Alsop DC, Tabor SL, et al. A method for in vivo high resolution MRI of rat spinal cord injury. Magn Reson Med 1994;31(2):218 – 23. [7] Wirth III ED, Mareci TH, Beck BL, Fitzsimmons JR, Reier PJ. A comparison of an inductively coupled implanted coil with optimized surface coils for in vivo NMR imaging of the spinal cord. Magn Reson Med 1993;30(5):626 – 33. [8] Bilgen M, Elshafiey I, Narayana PA. In vivo magnetic resonance microscopy of rat spinal cord at 7 T using implantable RF coils. Magn Reson Med 2001;46(6):1250 – 3. [9] Bilgen M. Simple, low-cost multipurpose RF coil for MR microscopy at 9.4 T. Magn Reson Med 2004;52(4):937 – 40. [10] Bilgen M. Magnetic resonance microscopy of spinal cord injury in mouse using a miniaturized implantable RF coil. J Neurosci Methods 2006;159(1):93 – 7. [11] Bilgen M. Inductively-overcoupled coil design for high resolution magnetic resonance imaging. Biomed Eng Online 2006;5:3. [12] Roemer PB, Edelstein WA, Hayes CE, Souza SP, Mueller OM. The NMR phased array. Magn Reson Med 1990;16(2):192 – 225. [13] Beck BL, Blackband SJ. Phased array imaging on a 4.7T/33 cm animal research system. Rev Sci Instrum 2001;72(11):4292 – 4. [14] Ullmann P, Junge S, Nauerth A, Panagiotelis I, Ruhm W, Hennig J. High field parallel imaging in rats. Book of abstracts: Twelfth Annual Meeting of the International Society of Magnetic Resonance in Medicine, vol. 1. Kyoto, Japan7 ISMRM; 2004. p. 1610. [15] Terman F. Radio engineers’ handbook. McGraw-Hill; 1943. p. 148. [16] Constantinides CD, Atalar E, McVeigh ER. Signal-to-noise measurements in magnitude images from NMR phased arrays. Magn Reson Med 1997;38(5):852 – 7. [17] Henkelman RM. Measurement of signal intensities in the presence of noise in MR images. Med Phys 1985;12(2):232 – 3. [18] Kramer CY. Extension of multiple range tests to group means with unequal numbers of replications. Biometrics 1956;12:309 – 10. [19] Bewick V, Cheek L, Ball J. Statistics review 7: correlation and regression. Crit Care 2003;7(6):451 – 9. [20] Yung AC, Carr K, Kozlowski P. Comparison of phased array versus implantable coil for rat spinal cord MRI. Book of abstracts: Fourteenth Annual Meeting of the International Society of Magnetic Resonance in Medicine, vol. 1. Seattle, USA7 ISMRM; 2006. p. 886.
Signal-to-noise ratio comparison of phased-array vs. implantable coil for rat spinal cord MRIB Andrew C. Yung4, Piotr Kozlowski University of British Columbia MRI Research Centre, Vancouver, BC Canada V6T 2B5 Received 2 October 2006; revised 8 January 2007; accepted 8 January 2007
Abstract The signal-to-noise ratio (SNR) performance and practicality issues of a four-element phased-array coil and an implantable coil system were compared for rat spinal cord magnetic resonance imaging (MRI) at 7 T. MRI scans of the rat spinal cord at T10 were acquired from eight rats over a 3 week period using both coil systems, with and without laminectomy. The results demonstrate that both the phased array and the implantable coil systems are feasible options for rat spinal cord imaging at 7 T, with both systems providing adequate SNR for 100-Am spatial resolution at reasonable imaging times. The implantable coils provided significantly higher SNR, as compared to the phased array (average SNR gain of 5.3 between the laminectomy groups and 2.5 between the nonlaminectomy groups). The implantable coil system should be used if maximal SNR is critical, whereas the phased array is a good choice for its ease of use and lesser invasiveness. D 2007 Elsevier Inc. All rights reserved. Keywords: Implantable coil; Phased array; Rat spinal cord; MRI
1. Introduction Magnetic resonance imaging (MRI) has been useful in longitudinally tracking pathological changes in rat models of mechanical spinal cord injury and other neurodegenerative diseases involving the spinal cord, using a variety of techniques such as T1- and T2-weighted imaging [1,2], proton spectroscopy [3], diffusion tensor imaging [4] and T2 relaxometry [5]. High signal-to-noise ratio (SNR) is required to meet the stringent data quality requirements of such techniques and to maintain adequate spatial resolution in the spinal cord (typically only a few millimeters square in the transverse plane). The most reported coil strategy to achieve increased SNR in animal spinal cord is the implantable coil system, where an internal coil is surgically placed over the spinal column and inductively coupled to a pickup coil outside the body [3,6–11]. The SNR advantage of implantable coils over single-element surface coils is clearly shown in the literature, with SNR gains ranging from 2.2 to 4.8 [6,7].
B This work was funded by the Rick Hansen Man-in-Motion Research Fund. 4 Corresponding author. Tel.: +1 604 822 6938; fax: +1 604 827 3973. E-mail address: [email protected] (A.C. Yung).
0730-725X/$ – see front matter D 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.mri.2007.01.006
Multiple-element phased arrays are known to produce SNR comparable with small surface coils within a relatively large field of view (FOV) [12]. These coils, commonly used on clinical and research human MRI scanners, are becoming increasingly popular for animal brain and spine MRI [13,14]. The objective of this study was to compare SNR performance of a phased-array coil vs. an implantable coil for rat spinal cord MRI. If phased array coils can achieve comparable SNR in the spinal cord, they may be an attractive alternative to implantable coils due to their relative ease of use and reduced invasiveness. MRI scans of the rat spinal cord at T10 were acquired at 7 T from eight rats over a 3-week period, using both coil systems, with and without laminectomy. The SNR performance and practicality issues between the two coil systems were compared. 2. Methods 2.1. Phased-array coil implementation A four-element receive-only phased array was built for rat brain and spinal cord imaging (National Research Council, Winnipeg, Canada). The coil elements are copper traces etched on a printed circuit board, are rectangular in shape (individually 827 mm) and are overlapped along
1216
A.C. Yung, P. Kozlowski / Magnetic Resonance Imaging 25 (2007) 1215 – 1221
Fig. 1. The phased-array coil for rat brain/spine MRI. (a) The array consists of four decoupled coil elements on a curved shell (bCQ), a floating cable balun (bBQ) and a matching/tuning box (bM/TQ). (Inset) shows coil arrangement. (b) Phased array circuit diagram.
their short axis on a curved plastic shell, covering a lateral width of 20 mm (see Fig. 1A). The length of 27 mm was chosen to ensure coverage of at least two vertebral levels. The 8-mm width was chosen to provide optimal signal reception at roughly 8 mm below the coil, where the spinal cord at T10 is typically located. The circuit diagram of the phased array is shown in Fig. 1B. Neighboring elements are geometrically decoupled through coil overlap, while isolation between nonadjacent elements is achieved by capacitive decoupling. PIN diodes turn on LC traps that detune the array during the transmission pulse, which is provided by an 11-cm internal diameter quadrature volume coil (Bruker Biospin, Ettlingen, Baden-Wuerttemberg, Germany). Variable capacitors situated one wavelength away from the coil elements allow adjustment of the tuning and matching inside the magnet, and a floating cable balun reduces current on the cable shields. The nuclear magnetic resonance (NMR) signals detected by each coil element are fed into the system preamplifiers and combined in a sum-of-squares reconstruction to form the composite phased-array image. 2.2. Implantable coil system implementation The implantable coil system (Fig. 2) consisted of a small internal coil inductively coupled to a pickup coil placed above it on the body surface. The implantable coil (previously described in [4]) consisted of a 1610-mm rectangular loop made from AWG 20 copper wire, with curved arches on each end to allow the implant to fit over the long axis of the spinal column. Series chip capacitors (700A series, American Technical Ceramics, Huntington Station, NY, USA) equaling 8.6 pF were soldered to one of the coil arches to achieve circuit resonance, while insulation and biocompatibility is achieved by coating the coil with polyolefin heatshrink and biocompatible elastomer (MDX4210 from Dow Corning, Midland, MI, USA). The pickup coil was a 3-cm-diameter circular surface coil made from copper tubing, which is tuned and matched to the system preamplifier’s 50-V impedance with a capacitive network. One important consideration is the bresonance splittingQ in the coil system’s frequency response. Terman [15] has
shown that the mutual inductance between two resonant circuits produces two impedance minima, with the amount of peak separation dependent on the severity of sample loading and the degree of mutual inductance. Tuning either peak to the Larmor frequency will enable efficient reception of the NMR signal; however, the spatial B1 profile depends on the choice of resonant peak, and therefore, the coil designer should select the one that best fits the region of interest. Since the low-frequency peak reportedly gives better SNR in the cord region [3], the implantable coil resonance was tuned to 310.5 MHz (in air) so that the in vivo low-frequency resonance is tuned to the Larmor frequency 300.3 MHz. This overcoupled resonance splitting exists even if the self-resonant frequencies of the isolated implant, and isolated pickup is slightly unequal, which provides some capability of retuning the system if the implantable coil resonant frequency drifts over time in vivo [11], especially in longitudinal studies. However, practical experience showed that a mismatch between the selfresonant frequencies of the implant and pickup reduces SNR significantly. Therefore, effort was made to maximize SNR by ensuring that the in vivo self-resonant frequencies were equal, which is indicated by equal peak heights in the pickup coil spectrum, as seen on a network analyzer.
Fig. 2. The implantable coil system. The implantable coil (foreground) is inductively coupled with a 3-cm-diameter pickup coil (background) outside the body.
A.C. Yung, P. Kozlowski / Magnetic Resonance Imaging 25 (2007) 1215 – 1221
1217
Fig. 3. Combined phased-array image generated by sum of squares reconstruction (far left) and individual images generated by the array elements.
Prior to implantation, the implantable coil was sterilized in ethanol for at least 2 h and rinsed in sterile saline. A midline exposure of the spinal column and adjoining rib junctions was first made at the T9-T11 level. The implantable coil was then placed over the column, and each corner was sutured to the nearest rib. This step was taken to ensure long-term immobilization of the implantable coil, since our initial experience showed that unsutured implants had a tendency to migrate away from the spinal cord because of swelling and formation of a tissue capsule around the implant during long-term implantation. The overlying paraspinal muscles and skin were then sutured to close the wound. 2.3. Experiments Quality assurance tests were performed on the coil systems by measuring quality factor (unloaded and loaded) using a network analyzer. The cross-talk between phasedarray elements was also measured. All animal experimental procedures were carried out in compliance with the guidelines of the Canadian Council for Animal Care and were approved by the institutional Animal Care Committee. The spinal cords of female Sprague– Dawley rats were individually imaged four times over a period of 3 weeks (time points at approximately 1, 7, 14 and 21 days) using a 7-T, 30-cm bore MR scanner (Bruker, Germany) for comparing SNR performance of the two coil systems. Four rats were imaged for each coil system. For each coil system, two of the rats underwent laminectomy while the remaining two did not. Laminectomies were
Fig. 4. Typical FLASH images of rat spinal cord at T10 level at 1001001000-Am resolution, acquired with (A) implantable coil and (B) phased array.
performed to mimic the increased invasiveness and heavier loading that may occur after an experimental spinal cord injury; the spatial patterns produced by actual mechanical trauma to the cord were deemed to be too variable to allow consistent comparison of SNR. Two implantable coils were constructed and first implanted into the nonlaminectomized rats. After this group’s time series was completed, the same implantable coils were recoated, resterilized and reused in the laminectomized rats. Before reimplantation, the quality factors of the implantable coils were verified to be unchanged. For the MRI experiments, the rats were placed supine on a plastic halfshell with the T10 region centered on the phasedarray elements or pickup coil. Anesthesia was maintained with isoflurane in air (4% induction, 1–2% maintenance), body temperature was maintained and monitored with a heated water pad and rectal probe, respectively and the respiration pattern was monitored with a pneumatic pressure sensor (SA Instruments, Stony Brook, NY, USA). A series of scout images were used to prescribe axial slices centered about the T10 level of the spinal cord. All rats were imaged with a multislice gradient-echo Fast Low Angle SHot (FLASH) sequence (TR/TE = 250/8 ms, FOV= 2.56 cm, 256256 matrix, slice thickness =1 mm, NA=10, 10 slices), resulting in a voxel size of 1001001000 Am. The sequence was respiration-triggered to reduce motion artifact; the resultant TR varied between 850 and 1000 ms due to
Fig. 5. Time evolution of in vivo SNR in individual rats over a 3-week period. The implantable coil measurements show higher SNR than the phased array measurements. Note that there is a significant difference in the implantable coil groups (with and without laminectomy), while there are no differences in the phased-array groups.
1218
A.C. Yung, P. Kozlowski / Magnetic Resonance Imaging 25 (2007) 1215 – 1221
Table 1 Group statistics of SNR measurements Group
Mean SNR
jSNR
Variability (jSNR/mean SNR)
Phased array without laminectomy Phased array with laminectomy Implant without laminectomy Implant with laminectomy
37 28 94 148
5.3 6.7 11.0 25.7
14% 24% 12% 17%
spontaneous changes in breathing rate, resulting in a total acquisition time of 36–43 min. The transmission power for the phased-array scans was automatically set to achieve a flip angle of 608 (previously determined as the Ernst angle). For the implantable coils, the transmitter power was manually adjusted to provide maximum signal in the centre of the cord, since the scanner’s automatic algorithm does not provide accurate flip angles for inhomogeneous transmit fields. SNR was estimated at the T10 level as the ratio between the signal mean in the spinal cord and the standard deviation of a noise region. Care was taken to exclude pixels in the cord that may suffer from partial volume effects with surrounding structures. In cases where the slice centered at T10 was corrupted by susceptibility artifacts at the cord periphery (a result of the cord/bone interface), an adjacent slice free of these artifacts was selected for SNR analysis. The SNR values require multiplication by correction factors to account for the noise bias introduced by the magnitude reconstruction. The amount of noise bias is slightly different for the phased array due to the sum-of-squares reconstruction of the four individual coil elements [16]. True SNR values were calculated by multiplying the implantable coil SNR values by 0.655 [17], while the phased-array SNR values required a correction factor of 0.695 [16]. Each of the four experimental groups (i.e., with and without laminectomy using the implantable coil, with and without laminectomy using the phased array) included two rats, with four measurements each over a 3-week period. All time measurements in each experimental group were pooled together (N = 8) to generate group means and standard deviations. A Tukey-Kramer test [18] was performed on the four groups to determine if any of the differences between means were statistically significant. To test whether SNR was independent of time, a straight line was first fitted through each group’s SNR vs. time data, with a slope of zero considered to be evidence that SNR was independent
from time. Random variability of the data can confound an observation of zero slope, so a t test was performed to calculate the probability that any nonzero slope was due to chance alone [19]. 3. Results Network analyzer measurements showed an interelement isolation for the phased array in the range of 16.8–29 dB, with only one measurement below 20 dB, thereby indicating acceptable levels of noise correlation between elements. Q air/Q in vivo was 178/84 and 210/112 for the phased-array and implantable coil system, respectively. The rats seemed to tolerate the implantable coils reasonably well and exhibited no behavioral deficit. In one rat, the implantation resulted in swelling around the surgery site, which disappeared within the first week after the surgery. Fig. 3 shows a representative phased-array image generated by sum-of-squares reconstruction, along with the images individually generated by each coil element. The phased array SNR is 41 in the spinal cord, whereas the individual elements alone generated cord SNR values of 20, 27, 20 and 15 from left to right, respectively. It is clear that the combination of four 827-mm coil elements provides a higher SNR than is achievable by a single 827-mm coil alone. Fig. 4 compares typical axial FLASH images acquired at T10 level using the implantable coil (left) and the phased array (right). Both images are at 100-Am pixel resolution and clearly elucidate structures within the spinal cord, such as blood vessels and the characteristic butterfly pattern of the central gray matter. For the implantable coil measurements, the mean distance between the centre of the cord and the implantable coil wires was 1.8F0.2 mm, whereas the mean distance between the cord centre and the phased array elements was 11.2F0.9 mm. The serial SNR measurements for each rat are presented in Fig. 5, while the overall results are summarized in Table 1. Fig. 6 shows images at different time points (t = 1, 5, 13, 19 days) acquired from the rat that showed the greatest SNR variability when using the implantable coil. The SNR variability (ratio between the standard deviation of SNR and mean SNR) ranged from 12% to 24%; none of the groups showed sufficient evidence that the variability was caused
Fig. 6. Longitudinal series of images on laminectomized rat using implantable coil (t = 1, 5, 13, 19 days). Images do not show SNR degradation due to longterm implantation of coil. Intensity levels are scaled identically in all images.
A.C. Yung, P. Kozlowski / Magnetic Resonance Imaging 25 (2007) 1215 – 1221
by a linear dependence between SNR and time, suggesting that SNR and time are independent. Overall, the implantable coils in our study produced better SNR than the phased array. The implantable coils with and without laminectomy performed better than the phased array by a factor of 5.3 and 2.5, respectively; these differences were statistically significant ( P b.001). There was no statistically significant difference between the phased-array groups ( P N.25). Conversely, there was a statistically significant difference in the implant coil groups ( P b.001): the average SNR of the laminectomy group was 1.6 times higher than that of the nonlaminectomy group. 4. Discussion Both the phased-array and implantable coil systems produced good in vivo anatomical images of the T10 spinal cord at 100-Am pixel resolution and 1-mm slice thickness at 7 T. Variability in each group ranged from 12% to 24%, which seems reasonable, considering the experimental variability in repeated in vivo scans. Although both coil systems have proven to be viable alternatives to highresolution imaging of the spinal cord, there are distinct differences between the coil systems that may lead an investigator to choose one system over the other. The phased-array coil does suffer from lower SNR in the T10 spinal cord, primarily as a result of lower proximity to the region of interest. However, one major advantage of the phased array is that it does not require surgery or long-term implantation, thereby reducing invasiveness and possible interference to the experimental data. The phased array is also more versatile in the sense that it can be moved from session to session to image different parts of the spinal cord, whereas the implantable coil’s location is fixed by the implantation. In fact, relative performance of the phased array may improve for regions in the lumbar spine, where the spinal cord lies closer to the body surface. Finally, the use of a multielement phased array opens up the possibility of parallel imaging, which can be used to accelerate data acquisition. Improvements to the phased array may increase its SNR performance; however, the phased-array coil used in this study represents a practical implementation that an investigator might typically construct for general rat spine and brain imaging. Integrated preamplifiers on the receive channels may reduce noise correlation between elements; however, resultant SNR increase may be marginal since the measured interelement isolation already appears adequate ( N20 dB for all but one pair). The element geometries could be optimized to achieve the best SNR at a certain depth below the array; however, the flexibility of the coil to image other parts of the spine and brain may be reduced. There was no significant difference in phased-array SNR between the laminectomy and nonlaminectomy groups. This indicates that laminectomy (and, by suggestion, spinal cord injury) does not add any significant long-term loading
1219
effects on the phased array. The phased-array SNR may possibly be lower in the immediate phase following surgery or injury because of increased loading that comes with initial inflammation. However, measurements at these early time points were not taken in order to allow the animals to recover from anesthesia. The most significant advantage of the implantable coil system is its higher SNR [9], outperforming the phased array by a factor of 5.3 and 2.5 for the laminectomy and nonlaminectomy groups, respectively. The results are on the same order of magnitude as other published improvement factors between implantable coils and single-element surface coils, as reported by Wirth et al. [7] (gain of 2.2) and Ford et al. [6] (gain of 4.8). The in vivo SNR comparisons presented here also correlate with a preliminary study, which showed that the implantable coil performed 2.4 better than the phased array in a dead rat [20]. Although the in situ comparison is not exactly equivalent to the in vivo case (e.g., different relaxation properties, untriggered acquisition and the absence of motion), it may reflect coil performance under more favourable imaging conditions. The higher implantable coil SNR highlights the advantage of using a small coil near its region of interest, which limits the volume of reception sensitivity but also decreases the amount of noise detected. Because of the ability to place the receiver coil very close to the spinal cord regardless of its depth below the surface, the relative SNR advantage over the phased array may increase for deep cord regions such as the cervical spine, which is an important site used in experimental spinal cord injury. The SNR gain is somewhat reduced by the increased dielectric loss due to enveloping the implant coil inside the conductive sample at high fields; however, this loss was ameliorated by insulating the wires with polyolefin heatshrink. Addition of a second capacitor junction would provide more distributed capacitance [8,9]; however, initial experiments showed that the long-term mechanical ruggedness in the coils was greatly reduced, while the SNR gain due to the higher dielectric isolation was only 10–15% (data not shown). Other practical advantages of the implantable coil system include easier localization of the region of interest and less problems with image wraparound due to a limited field of sensitivity. The major drawback of the implantable coil system is its invasiveness. However, the implantation does not add too much additional insult if an experimental spinal cord injury is already being performed. In addition, experience from our research and others [7,8] indicate that the implants are generally well-tolerated without significant inflammation. Furthermore, the implantable coil system requires significant technique refinement in order to optimize the SNR. Automatic power adjustment fails for the implantable coil system, and manual adjustment of the transmission power requires user experience and is approximate, at best. Mechanical ruggedness must be ensured by strong solder joints and uniform application of the biocompatible elastomer, since repair is extremely impractical after
1220
A.C. Yung, P. Kozlowski / Magnetic Resonance Imaging 25 (2007) 1215 – 1221
implantation. Suturing the corners to the ribs must be done carefully in order to prevent migration of the coil away from the cord, since we noticed that the SNR in the spinal cord can drop by as much as 50% for a 1-mm increase in separation distance (data not shown), even after appropriate power adjustments to optimize the SNR in the spinal cord. Figs. 5 and 6 illustrate that these precautions were successful in preventing the systematic degradation in implantable coil SNR over time, which has been reported by other investigators [8]. The implantable coil system may also suffer from higher signal inhomogeneity as compared to the phased array, since the implantable coil is small in size and acts as a receive– transmit system (the phased array is a receive-only device and relies on the volume coil to provide a homogeneous excitation). This effect is more noticeable in spin-echo images [11]. Fig. 7 compares phased array performance with implantable coil performance using spin-echo Rapid Acquisition with Relaxation Enhancement data from a separate MRI study on contusion-injured rats (RARE factor = 8, TR/TE = 5000/ 37 ms, FOV= 4 cm, 256256 matrix, slice thickness = 1 mm, NA=4, 10 slices). A signal dropoff was observed at the end of the implantable coil (Slice 1), which is not seen in the phased array image series. The rest of implantable coil slices show good uniformity within the spinal cord, with a clear advantage in SNR over the phased array.
Another issue that adversely affects the implantable coil performance is the difficulty in choosing the unloaded (air) resonant frequency of the implant that would achieve the best SNR in vivo. It is true that overcoupling the implantable coil to the pickup coil gives some flexibility to the choice of self-resonant frequencies, since the tuning and matching circuit of the external coil can be used to adjust the position and height of the coupled resonances, therefore maintaining imaging performance in longitudinal studies even when the resonant frequency drifts [11]. However, as mentioned in a previous section, the best performance occurs when the amplitudes of the two resonant peaks are equal, with the low-frequency peak resonating at the Larmor frequency. In order to perfectly tune the implant on the bench to achieve this condition, the peak separation in vivo would need to be known in advance. Since this a priori knowledge is generally not available (due to variability in experiment geometries and loading conditions), the unloaded implant frequency is chosen as a best guess based on previous experience. As a result, the system resonance was not optimal in many of our experiments, primarily due to electronic mismatch (i.e., response at Larmor frequency not matched to 50 V) with an estimated SNR decrease of 15% or less in most cases. In our study, the SNR measurements for the implantable coils with laminectomy showed significantly higher values
Fig. 7. RARE spin-echo images from contusion-injured rats showing differences in signal homogeneity when using implantable coil (top) and phased array (bottom).
A.C. Yung, P. Kozlowski / Magnetic Resonance Imaging 25 (2007) 1215 – 1221
than those without laminectomy. This result is somewhat surprising and counterintuitive. One would expect that foregoing the laminectomy would produce SNR that was similar to or even higher than the laminectomy case, which was the case with the phased-array measurements. We do not believe that this result reflects a real effect of the laminectomy on SNR, but rather, this SNR difference is indicative of systematic differences that are difficult to control between sets of measurements using the implantable coil system. The uncertainty in pretuning the implantable coil on the bench has already been mentioned; this and other differences in coil preparation between individual implantations are especially prone to bias the SNR in a set of measurements in one rat relative to another. The level of expertise in the coil implantation may play a role as well; the laminectomy group implantations were performed by a more experienced animal technician than the nonlaminectomy group. We also noticed that the split resonant peaks in the nonlaminectomy measurements were generally broader and closer together than for the laminectomized rats, suggesting that the mutual coupling between coil elements was consistently different. It remains to be determined whether this phenomenon reflects an underlying cause that decreases SNR. Clearly, more understanding into the electromagnetic interactions in the implantable coil system is needed to explain the difference between the implantable coil measurements. However, these differences in SNR between implantable coils do not invalidate the general conclusion from our study that the implantable coils perform better than phased-array coils in terms of SNR.
5. Conclusion Our results demonstrate that both the phased-array and implantable coil systems are feasible options for rat spinal cord imaging at 7 T, with both systems providing adequate SNR for 100-Am spatial resolution at reasonable imaging times. The implanted coils provided significantly higher SNR as compared to the phased array. The phased array is a good choice if less invasiveness and more simplicity is desired, whereas the implantable coil can be used if maximal SNR is critical.
Acknowledgments The phased array was designed and constructed by Vyacheslav Volotovskyy (NRC Institute for Biodiagnostics). The authors would like to thank Caroline Hall, Katie Carr and Jie Liu for expertise in coil implantation and surgeries.
1221
References [1] Fraidakis M, Klason T, Cheng H, Olson L, Spenger C. Highresolution MRI of intact and transected rat spinal cord. Exp Neurol 1998;153(2):299 – 312. [2] Weirich SD, Cotler HB, Narayana PA, Hazle JD, Jackson EF, Coupe KJ, et al. Histopathologic correlation of magnetic resonance imaging signal patterns in a spinal cord injury model. Spine 1990;15(7): 630 – 8. [3] Silver X, Ni WX, Mercer EV, Beck BL, Bossart EL, Inglis B, et al. In vivo 1H magnetic resonance imaging and spectroscopy of the rat spinal cord using an inductively-coupled chronically implanted RF coil. Magn Reson Med 2001;46(6):1216 – 22. [4] Ford JC, Hackney DB, Alsop DC, Jara H, Joseph PM, Hand CM, et al. MRI characterization of diffusion coefficients in a rat spinal cord injury model. Magn Reson Med 1994;31(5):488 – 94. [5] Yung AC, Choo A, Liu J, Tetzlaff W, Oxland T, Kozlowski P. Measurements of myelin water in rat spinal cord in vivo. Book of abstracts: Twelfth Annual Meeting of the International Society of Magnetic Resonance in Medicine, vol. 1. Kyoto, Japan7 ISMRM; 2004. p. 1537. [6] Ford JC, Hackney DB, Joseph PM, Phelan M, Alsop DC, Tabor SL, et al. A method for in vivo high resolution MRI of rat spinal cord injury. Magn Reson Med 1994;31(2):218 – 23. [7] Wirth III ED, Mareci TH, Beck BL, Fitzsimmons JR, Reier PJ. A comparison of an inductively coupled implanted coil with optimized surface coils for in vivo NMR imaging of the spinal cord. Magn Reson Med 1993;30(5):626 – 33. [8] Bilgen M, Elshafiey I, Narayana PA. In vivo magnetic resonance microscopy of rat spinal cord at 7 T using implantable RF coils. Magn Reson Med 2001;46(6):1250 – 3. [9] Bilgen M. Simple, low-cost multipurpose RF coil for MR microscopy at 9.4 T. Magn Reson Med 2004;52(4):937 – 40. [10] Bilgen M. Magnetic resonance microscopy of spinal cord injury in mouse using a miniaturized implantable RF coil. J Neurosci Methods 2006;159(1):93 – 7. [11] Bilgen M. Inductively-overcoupled coil design for high resolution magnetic resonance imaging. Biomed Eng Online 2006;5:3. [12] Roemer PB, Edelstein WA, Hayes CE, Souza SP, Mueller OM. The NMR phased array. Magn Reson Med 1990;16(2):192 – 225. [13] Beck BL, Blackband SJ. Phased array imaging on a 4.7T/33 cm animal research system. Rev Sci Instrum 2001;72(11):4292 – 4. [14] Ullmann P, Junge S, Nauerth A, Panagiotelis I, Ruhm W, Hennig J. High field parallel imaging in rats. Book of abstracts: Twelfth Annual Meeting of the International Society of Magnetic Resonance in Medicine, vol. 1. Kyoto, Japan7 ISMRM; 2004. p. 1610. [15] Terman F. Radio engineers’ handbook. McGraw-Hill; 1943. p. 148. [16] Constantinides CD, Atalar E, McVeigh ER. Signal-to-noise measurements in magnitude images from NMR phased arrays. Magn Reson Med 1997;38(5):852 – 7. [17] Henkelman RM. Measurement of signal intensities in the presence of noise in MR images. Med Phys 1985;12(2):232 – 3. [18] Kramer CY. Extension of multiple range tests to group means with unequal numbers of replications. Biometrics 1956;12:309 – 10. [19] Bewick V, Cheek L, Ball J. Statistics review 7: correlation and regression. Crit Care 2003;7(6):451 – 9. [20] Yung AC, Carr K, Kozlowski P. Comparison of phased array versus implantable coil for rat spinal cord MRI. Book of abstracts: Fourteenth Annual Meeting of the International Society of Magnetic Resonance in Medicine, vol. 1. Seattle, USA7 ISMRM; 2006. p. 886.