IEEE TRANSACTIONS ON APPLIED SUPERCONDUCTIVITY, VOL. 21, NO. 3, JUNE 2011
509
Low Field MRI Detection With Tuned HTS SQUID Magnetometer Hui Dong, Yi Zhang, Hans-Joachim Krause, Xiaoming Xie, and Andreas Offenhäusser
Abstract—We set up a low field (LF) magnetic resonance imaging (MRI) system with a tuned high- c (HTS) superconducting quantum interference device (SQUID) consisting of a SQUID magnetometer and a liquid-nitrogen-cooled resonant pickup circuit, which are inductively coupled. Around 9 kHz, the configuration enhances the sensitivity of the system to 6–7 fT Hz. Meanwhile, the relatively large diameter of the coil compensates the magnetometer’s drawback of a small pickup area. We performed our experiments with a prepolarization field of 10 mT in a magnetically shielded room. The measurement field was 212.2 T, corresponding to the Larmor frequency of 9.038 kHz. The compromise between the signal-to-noise ratio and the spatial resolution for our system was studied by recording one-dimensional MRI images of carrots under different gradient fields. For the sample of two carrot slices with center distance of 2.8 cm, a gradient field of 20 Hz/cm was suitable. Two-dimensional images were finally acquired. A HTS system for biological LF MRI detection by a tuned SQUID is demonstrated to be feasible. Index Terms—HTS SQUID magnetometer, LC resonant circuit, low field MRI, tuned SQUID.
I. INTRODUCTION
I
N RECENT years, researchers demonstrated advantages in using superconducting quantum interference device (SQUID) sensors for detection of liquid state low-field (LF) nuclear magnetic resonance (NMR) and magnetic resonance imaging (MRI) [1]–[3]. LF MRI has several advantages compared to the conventional high field MRI, such as moderate demand on field homogeneity, the possibility to perform imaging in the presence of metals [4] and enhanced -contrast which allows one to discriminate different types of tissues [5]. The simultaneous detection of MRI and magnetoencephalography (MEG) is another promising application [3], [6]. However, the main drawback of LF MRI is the low signal-to-noise ratio (SNR). This can be partly alleviated by applying a strong sample prepolarization [7]. For example, the UC Berkeley group applied a prepolarizing field up to Manuscript received August 01, 2010; accepted November 06, 2010. Date of publication December 20, 2010; date of current version May 27, 2011. This work was supported by the International Bureau of the German BMBF at DLR, Grant CHW09/009. Hui Dong is with Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences, Shanghai 200050, P. R. China. He is also with the University of the Chinese Academy of Sciences, Beijing 100049, China (e-mail:
[email protected]). Yi Zhang, Hans-Joachim Krause, and Andreas Offenhäusser are with Institute of Bio- and Nanosystems (IBN-2), Forschungszentrum Jülich, D-52425, Jülich, Germany (e-mail:
[email protected]). Xiaoming Xie is with Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences, Shanghai 200050, China (e-mail:
[email protected]). Digital Object Identifier 10.1109/TASC.2010.2091713
0.3 T, more than 3 orders of magnitude higher than the mea, which determines the Larmor frequency surement field [2]. Besides the enhancement of the , the use of a sensitive detector is also necessary. Most recently published SQUID-based LF NMR & MRI results utilized LTS SQUID sensors, which had a quite high sensitivity of several . The systems with high- (HTS) SQUIDs are also attractive, because liquid helium is not needed for cooling. Schlenga et al. [8] used a HTS SQUID magnetometer to directly detect MRI signals. The SQUID was operated in vacuum and the spatial separation between SQUID at 77 K and sample at room temperature was less than 1 mm. Liao et al. [9] tuned the resonant frequency of a resonant circuit, comprised of one canormal conducting pacitor and two coils, the pickup coil and the input coil, to the of MRI signals. The input coil was inductively coupled to an HTS dc SQUID and immersed in a superconducting vessel whereas the pickup coil coupled to the sample at room temperature. The introduction of the pickup coil improved the coupling between sample and detector. However, the system noise level of 280 was limited by the thermal noise of this coil at room temperature. In the previous work of our group, Qiu et al. applied a tuned SQUID including a liquid nitrogen cooled resonant circuit inductively coupled to an HTS rf SQUID magnetometer [10]. This tuned circuit enhanced the equivalent sensitivity of the system up to a few above several kHz. Furthermore, this configuration significantly increased the effective detecting area of the SQUID to the same level as that attained with the large normal conducting pickup coil , because the coupling between the sensor and the sample was improved [11]. In MRI technique, the SNR is affected by a number of different factors, for example, the field of view (FOV), the slice thickness, the spatial resolution and the receiver bandwidth, etc. The factors are related to each other, e.g. the slice thickness should be increased to improve SNR, but it should be decreased to gain spatial resolution. In this paper, the tradeoff between the SNR and spatial resolution for our system are studied by comparing the one-dimensional (1D) MRI spectra of carrots at gradient field strengths varying from 5 Hz/cm to 50 Hz/cm. Finally, two-dimensional (2D) MRI results at two different gradient fields are presented. II. SETUP Our setup is schematically shown in Fig. 1(a). The sample was surrounded by the solenoid located beneath a fiberglass cryostat, and at the center of a Helmholtz coil pair (coil pair I) generating . The field was perpendicular to . The excitation field coil was wound outside the solenoid.
1051-8223/$26.00 © 2010 IEEE
510
IEEE TRANSACTIONS ON APPLIED SUPERCONDUCTIVITY, VOL. 21, NO. 3, JUNE 2011
TABLE I PARAMETERS OF THE COIL
Fig. 1. (a) Setup scheme and orientation of magnetic fields. Coil pairs I, II, and III represent the Helmholtz coil pair (B ), the Maxwell coil pair (G ) and the planar gradient coil pair (G & G ), respectively. (b) Equivalent sensitivity of tuned SQUID. For comparison, the dashed line indicates the sensitivity of the SQUID magnetometer alone. The inset depicts a schematic diagram of the tuned SQUID with Q switch.
The gradient field was produced by a Maxwell coil pair (coil pair II) whereas the other two gradient fields, & , were generated by a planar gradient coil pair (coil pair III). The construction and performance of the coil pair III were described in [12]. The tuned SQUID which consisted of an rf SQUID magnetometer [13] and a resonant circuit was positioned at the bottom of the cryostat. The copper coil surrounded the SQUID magnetometer and inductively coupled to it via a mutual inductance . Both the resonant circuit and the SQUID were immersed into liquid nitrogen. The distance between the tuned SQUID and the sample center was about 2.7 cm. The resonant frequency of the circuit was adjusted to by changing the value of the capacitor , leading to a resonant signal enhancement at [10]. The parameters of the coil wound with copper wire (diameter 0.15 mm) are listed in Table I. is the intrinsic resonance frequency which is determined by the inductance and the parasitic capacitance of the coil. The dependence of the measured equivalent sensitivity of the tuned SQUID on frequency is shown in Fig. 1(b). From 8 kHz to 25 kHz, the sensitivity of the tuned SQUID is about 6–7 . The dashed line indicates the frequency-independent sensitivity of the SQUID magnetometer of 40–50 . The of our MRI experiments was selected to be about 9 kHz, which made a
Fig. 2. The 10-times averaged FID signals of carrots (a) and 20 ml water (b).
good compromise between the sensitivity of the sensor, the stability and homogeneity of the measurement field , and the minimization of environmental disturbances. All experiments were performed in a magnetically shielded room. Each measurement started with a prepolarizing field of 10 mT for a polarizing time of 5 s. The field was switched off nonadiabatically and then the free induction decay (FID) signal appeared. After a variable delay time , a pulse was applied by coil to record the spin echo signal. Because the quality factor of the circuit at 9 kHz is about one hundred, transient ringing was induced in the circuit after the sudden switch-off of and pulses. This undesirable ringing may cover the MRI signals because the duration of the ringing is comparable to that of the signals. To effectively damp the ringing, we developed a switch circuit [14]. The circuit was shunted by a field-effect-transistor (FET). By modifying the and logic pulse with an integrating module within the switch circuit, the gate voltage, the drain-source resistance and consequently the value of the circuit gradually increased from low value to its maximum after the end of the pulses. The connection between the tuned SQUID and the switch is schematically shown in the inset of Fig. 1(b). III. EXPERIMENTS AND RESULTS The basic principle of MRI is that by applying magnetic field gradients, spatial coordinates of magnetic resonance signals can be determined. However, the SNR decreases with increasing
DONG et al.: LOW FIELD MRI DETECTION WITH TUNED HTS SQUID MAGNETOMETER
511
Fig. 3. (a) The 10-times averaged FID and spin echo signals recorded by the tuned SQUID. The inset shows the FID signal measured with the SQUID magnetometer. (b) and (c) are the FFT spectra obtained from the FID and spin echo signals of (a), respectively. (d) shows the spectrum of the FID signal obtained with the SQUID magnetometer.
strength of the gradient field while the spatial resolution increases. Therefore, we first studied 1D images of a sample of two carrots to find a compromise between the spatial resolution and the SNR for our system. A. 1D MRI Measurements Two carrot slices in the 1D MRI measurements were separated by a distance of 1.3 cm. Each piece of carrot had a diameter of 2.4 cm and a length of 1.5 cm. The center distance between the two carrots is 2.8 cm (see the inset of Fig. 2(a)). The FID signals of the carrots and 20 ml tap water are compared at of about 212.2 . The two FID signals are mixed down to several hundred Hz when recorded by a HP 3562A dynamic signal analyser (see Figs. 2(a) and 2(b)). The duration of carrot FID signal is about 0.5 s whereas that of tap water lasts about 1 s. The corresponding spectra linewidths of carrot and water signals are 3 Hz and 1.8 Hz at . Typically, echo signals instead of FID signals are used in MRI measurements. Fig. 3(a) presents the FID and spin echo signals of the carrots recorded by the tuned SQUID at a gradient field of 1.18 , corresponding to 5 Hz/cm (the gyromagnetic ratio ). The pulse is applied at . The inset displays the FID signal recorded by the SQUID magnetometer alone. Because of the SNR, the spin echo signal is not well resolved by the SQUID magnetometer without resonant circuit. When performing a fast Fourier transform (FFT) of the FID and of the spin echo signals, similar spectra are obtained (see Figs. 3(b) and 3(c)). The two peaks in both spectra with a frequency difference of 14 Hz represent the centers of the two carrots and indicate a distance of 2.8 cm, which is in accordance with the geometric distance. The peak amplitude of the
echo is evidently reduced compared to the initial amplitude of FID. However, the SNRs of the two spectra are almost the same. In contrast, the two peaks detected by the bare SQUID magnetometer are just barely visible (see Fig. 3(d)). With the assiscircuit, the SNR is improved by about tance of cooled tuned a factor of 5. A detailed comparison between tuned SQUID and SQUID magnetometer is given in [13]. From the results above, it can be concluded that 1D MRI measurements recorded by the tuned SQUID at 5 Hz/cm are feasible. However, the SQUID magnetometer with lower sensitivity and smaller detecting area is not appropriate. To find a compromise between spatial resolution and SNR for our tuned SQUID system, 1D MRI experiments have been performed under different gradient fields. Fig. 4(a) shows the measured amplitude-frequency characteristics of the tuned SQUID. The resonant frequency is 9.038 kHz and the -value reaches 100. These two parameters imply that the resonant bandwidth is about 90 Hz, which is indicated by the frequency range between the two dashed lines in Fig. 4. The signals of the sample should be limited to this band, or they will be seriously attenuated. Figs. 4(b)–4(f) show the 1D MRI results under different gradient fields varying from 10 Hz/cm to 50 Hz/cm with a step of 10 Hz/cm. With increasing gradient field, the frequency difference between the two peaks representing the centers of the two carrot pieces becomes greater and the linewidths of the two peaks become broader. The spatial resolution is significantly improved while the SNR is reduced. From the experiments above, the choice of the gradient field strength depends on the size of the sample, the SNR and on the bandwidth of the resonant circuit. The determination of an optimal compromise between the spatial resolution and SNR for our system should take into account all the three pa-
512
IEEE TRANSACTIONS ON APPLIED SUPERCONDUCTIVITY, VOL. 21, NO. 3, JUNE 2011
Fig. 5. The reconstructed images of the carrot sample acquired at a gradient field of 20 Hz/cm (a) and 30 Hz/cm (b). The geometric diagram of the sample is shown in the inset of Fig. 2(a).
, the image is distorted (see Fig. 5(b)), increased to 70.4 because the SNR is low and a part of the signals locates out of circuit (see Fig. 4(d)). the resonant bandwidth of the For our tuned SQUID system and the sample of two carrot slices with the geometric shape described above, a gradient field of 20 Hz/cm is optimal when considering both SNR and spatial resolution. This result is in accordance with that obtained in 1D MRI experiments. IV. CONCLUSIONS Fig. 4. (a) The amplitude-frequency characteristics of the tuned SQUID circuit. The frequency region between the dashed lines indicates the resonant bandwidth. (b)–(f) are 30-times averaged 1D MRI spectra of the two carrot slices under different gradient fields varying from 10 Hz/cm to 50 Hz/cm with a step of 10 Hz/cm.
rameters. For a gradient field below 20 Hz/cm (see Fig. 4(b)), the spectrum shape does not clearly reflect the geometric form of the two carrot slices, because the spatial resolution is low. At the gradient field of 20 Hz/cm (see Fig. 4(c)), all the frequency components covering the sample still lie in the resonant bandwidth, whereas the SNR is acceptable. In this case, the geometric shape of two carrot slices is resolved in the spectrum. If the gradient field strength is enhanced to above 30 Hz/cm (see Figs. 4(d)–4(f)), the SNR is too low. Furthermore, part of frequency components resides beyond the resonant bandwidth. Therefore, the optimal gradient field for this sample is about 20 Hz/cm, representing good compromise between the spatial resolution and SNR. B. 2D MRI Measurement The 2D MRI images of two carrot slices were acquired based on the filtered back-projection reconstruction. The direction of the gradient field was rotated from 0 to 180 with 12 steps. 30 spin-echo signals were averaged for each projection. The reconstructed images at gradient fields of 47 (20 Hz/cm) and 70.4 (30 Hz/cm) are shown in Figs. 5(a) and 5(b), respectively. The dotted lines indicate the geometric contours of the sample. From Fig. 5(a), it can be observed that the center of the carrot images is bright but the edge is blurred. This may be caused by the lack of water content at the edge of carrots. According to the relation , a linewidth of 3 Hz and a gradient field of 47 correspond to a spatial resolution of 1.5 mm. However when the gradient field is
One- and two-dimensional LF MRI measurements were demonstrated using a tuned high- rf SQUID system. The determination of the gradient field strength depends on the size of the sample, the SNR and the resonant bandwidth of the tuned SQUID. For the carrot sample with a total length of 4.3 cm, the gradient field of 20 Hz/cm makes a good compromise between SNR and spatial resolution. A 2D image with spatial resolution of 1.5 mm 1.5 mm was presented. A tuned HTS SQUID system for biological LF MRI measurement is demonstrated to be feasible. To improve the performance of this system, both the enhancement of field strength and the replacement of solenoid by a coil pair could further increase the SNR. In the latter case, the distance between the sample and the sensor will be shortened whereas the sample can still be kept at room temperature. ACKNOWLEDGMENT The authors thank Prof. Dr. Alex I. Braginski for a careful reading of the paper. REFERENCES [1] R. McDermott et al., “Liquid-state NMR and scalar couplings in microtesla magnetic fields,” Science, vol. 295, pp. 2247–2249, 2002. [2] R. McDermott et al., “Microtesla MRI with a superconducting quantum interference device,” Proc. Natl. Acad. Sci., vol. 101, no. 21, pp. 7857–7861, 2004. [3] V. S. Zotev et al., “Multi-channel SQUID system for MEG and ultralow-field MRI,” IEEE Trans. Appl. Superconduct., vol. 17, no. 2, pp. 839–842, 2007. [4] M. Mößle et al., “SQUID-detected microtesla MRI in the presence of metal,” J. Magn. Reson., vol. 179, pp. 146–151, 2006. [5] S. K. Lee et al., “SQUID-detected MRI at 132 T with T1-weighted contrast established at 10 T–300 mT,” Magn. Reson. Med., vol. 53, pp. 9–14, 2005. [6] V. S. Zotev et al., “SQUID-based instrumentation for ultralow-field MRI,” Supercond. Sci. Technol., vol. 20, pp. S367–S373, 2007. [7] M. E. Packard and R. Varian, “Free nuclear induction in the earth’s magnetic field,” Phys. Rev., vol. 93, p. 941, 1954.
DONG et al.: LOW FIELD MRI DETECTION WITH TUNED HTS SQUID MAGNETOMETER
[8] K. Schlenga et al., “Low-field magnetic resonance imaging with a highT dc superconducting quantum interference device,” Appl. Phys. Lett., vol. 75, pp. 3695–3697, 1999. [9] S. H. Liao, H. C. Yang, H. E. Horng, and S. Y. Yang, “Characterization of magnetic nanoparticles as contrast agents in magnetic resonance imaging using high-Tc superconducting quantum interference devices in microtesla magnetic fields,” Supercond. Sci. Technol., vol. 22, p. 025003, 2009. [10] L. Q. Qiu, Y. Zhang, H.-J. Krause, A. I. Braginski, and A. Usoskin, “High-temperature superconducting quantum interference device with cooled LC resonant circuit for measuring alternating magnetic fields with improved signal-to-noise ratio,” Rev. Sci. Instrum., vol. 78, p. 054701, 2007.
513
[11] H. Dong, Y. Zhang, H.-J. Krause, X. M. Xie, A. I. Braginski, and A. Offenhäusser, “Comparison of different detectors in low field NMR measurements,” J. Phy.: Conf. Ser., vol. 234, p. 042008, 2010. [12] Y. Zhang et al., “Planar HTS gradiometers with large baseline,” IEEE Trans. Appl. Superconduct., vol. 2, no. 2, pp. 2866–2896, 1997. [13] Y. Zhang, J. Schubert, N. Wolters, M. Banzet, W. Zander, and H.-J. Krause, “Substrate resonator for HTS rf SQUID operation,” Physica C, vol. 372–376, pp. 282–286, 2002. [14] H. Dong et al., “Suppression of ringing in the tuned input circuit of a SQUID detector used in low-field NMR measurements,” Supercon. Sci. Technol., vol. 22, p. 125022, 2009.
