IMAGING METHODOLOGY Rapid Communications
Magnetic Resonance in Medicine 71:1676–1681 (2014)
Novel MRI Contrast Development by Lock-In Suppression Yu-Wen Chen,1 Chou-Hsiung Hsu,2,3 and Dennis W. Hwang1,4* induction decay (FID) signal to the detection hardware. However, a major issue with MRI lies in the lack of contrast in in vivo images between tissues with small variations in magnetic susceptibility or relaxation rates and not the amount of signal that can be detected. To resolve this problem, using feedback from a radiation damping (RD) field to selectively drive magnetization toward equilibrium much faster than the inherent relaxation rate, as described by Lenz’s law, is proposed. RD is shown to amplify the contrast when compared with standard MR sequences (1–15). Furthermore, nonlinear effects triggered by the associative coupling to RD fields amplify the effects of nonuniformity (9,16), creating a macroscopic signal to enhance ex vivo MRI contrast between gray and white matter at 14 Tesla (T) in epileptic patients with symptomatic lesions (1). On the basis of a concept similar to RD, we designed an active feedback device. Our goal is to identify and amplify a highly selective feedback field frequency to gain a significant contrast enhancement. This study focuses on a concept of frequency lock-in suppression and shows that the feedback RF field can image resonant frequency differences between magnetizations with small chemical shifts or frequency distributions.
Purpose: The goal of this study is to develop novel MR contrast by frequency lock-in technique. Methods: An electronic feedback device that can control the frequency and bandwidth of the feedback RF field is presented. In this study, the effects of lock-in suppressed imaging are discussed both theoretically and experimentally. Results: Two important imaging experiments were performed. The first experiment used magnetizations with the same central frequency but different frequency distributions and was compared with MR images obtained with T2 contrast agents. Lock-in suppressed images showed an improvement in contrast relative to the conventional imaging method. The second experiment used magnetizations with small shifts in frequency and a broad frequency distribution. This is helpful for differentiating between small structural variations in biological tissues. The contrast achieved in in vivo tumor imaging using the lock-in suppressed technique provide higher spatial resolutions and discriminate the regimes of necrosis and activation consistent with pathologic results. Conclusion: Lock-in suppressed imaging introduces a conceptually new approach to MRI. Heightened sensitivity to underlying susceptibility variations and their relative contribution to total magnetization may thus be achieved to yield new and enhanced contrast. Magn Reson Med 71:1676–1681, C 2014 Wiley Periodicals, Inc. 2014. V Key words: suppression
MRI;
feedback
field;
radiation
damping;
THEORY
INTRODUCTION MRI is an important noninvasive technique widely used in medical diagnosis. In MR experiments, the application of a radiofrequency (RF) pulse tips magnetization toward the transverse plane; subsequently precessing magnetization induces a current in the probe coil that carries the free 1
Department of Chemistry and Biochemistry, National Chung Cheng University, Min-Hsiung Township Chiayi, Taiwan. 2 Molecular Science and Technology Program, Taiwan International Graduate Program, Institute of Atomic and Molecular Sciences, Academia Sinica, Taipei, Taiwan. 3 Department of Chemistry, National Tsing-Hua University, Hsinchu, Taiwan. 4 Center for Nano Bio-detection, AIM-HI, National Chung Cheng University, Min-Hsiung Township Chiayi, Taiwan. Grant sponsor: the National Science Council of the Republic of China; Grant number: NSC 101–2113-M-194-006-MY2. *Correspondence to: Dennis W. Hwang, Ph.D., National Chung Cheng University, Department of Chemistry and Biochemistry, 168 University Road, Min-Hsiung Township Chiayi, Taiwan 62102. E-mail: [email protected] Additional Supporting Information may be found in the online version of this article. Received 17 October 2013; revised 12 January 2014; accepted 14 January 2014 DOI 10.1002/mrm.25162 Published online 6 March 2014 in Wiley Online Library (wileyonlinelibrary. com). C 2014 Wiley Periodicals, Inc. V
A feedback RF field is generated by artificially amplifying the FID and feeding the signal back into the coil with a certain phase. This user-generated active feedback field can be expressed by
BF ðv; tÞ ¼
iGeif x þ hMx ðr; tÞi^ y Þ; ðhMy ðr; tÞi^ g
[1]
where G 5 vggQM0 /2 is the amplification gain factor of the feedback RF field. The amplification gain factor is a measure of field strength, where g is the amplification gain by the active feedback circuit, g is the filling factor, and M0 is equilibrium magnetization. Mþ ðv; r; tÞ is the complex transverse magnetization integrated over sample volume with specific frequency, because the feedback field is dependent on net transverse magnetization. u is the phase between the feedback RF field and the FID. For example, when u ¼ 0 , the feedback RF field is in phase with the FID, and the feedback RF field is orthogonal to transverse magnetization, which is determined by the tuning condition of the RF coil circuit and can also be controlled by the feedback circuit. By including Eq. [1] in Bloch equations, dM=dt ¼ gM B, the time evolution of magnetization in the absence of relaxation can be expressed as
1676
Novel MR Contrast by Frequency Lock-In Technique
2
Mx ðr; tÞ
3
2
1677
0
7 6 @6 6 My ðr; tÞ 7 ¼ 6 dvðrÞ 5 4 @t 4 if Mz ðr; tÞ iGe hMx ðr; tÞi
dvðrÞ 0 iGe
if
where dvðrÞ represents the frequency offset at position r. In Eq. [1], because the feedback signal is filtered by a low pass filter, a feedback RF field with a particular frequency bandwidth can be generated. The off-resonance frequency causes oscillation between transverse magnetizations, Mx and My, and becomes less efficient in driving longitudinal magnetization, Mz. Hence the evolution of Mz with a different frequency can be differentiated. Incorporating with simulation of signal filter and timeshare mechanism, the detailed dynamics can be studied by numerical simulation using Matlab (The MathWorks, USA) and given in the supporting information. METHODS Animal Preparation Three-week-old athymic nude mice (BALB/cAnN.CgFoxn1nu/Cr1Narl) were ordered from the National Laboratory Animal Center (Taiwan). All mice were kept under specific pathogen-free conditions using a laminar airflow rack with free access to sterilized food and autoclaved water. All experiments were conducted after obtaining the relevant license from the Animal Experimentation Ethics Committee of the National Chung Cheng University. 2 105 cells of Oral cancer cell (OML1-R50) were resuspended in a mixture of 0.1 mL medium and Matrigel (1:1) (BD Bioscience, San Jose, CA). Cell suspension was then injected subcutaneously into the flank of each 4-
hMy ðr; tÞi
iGeif hMx ðr; tÞi
32
Mx ðr; tÞ
3
7 76 6 7 iGeif hMy ðr; tÞi 7 54 My ðr; tÞ 5 0
[2]
Mz ðr; tÞ
week-old nude mouse. At the end of experiment, all mice were killed by cervical dislocation. In vivo mice studies were conducted under anesthesia with an initial dosage of 2.0% isoflurane in air and using a small animal gating system (SA Instruments Inc., NY). For maintenance, isoflurane was set under 1.0% and gated by means of a respiration trigger sensor. The respiration rate of the mice was maintained under 60 breaths per min. Warm air was supplied at 25 6 2 C to avoid hypothermia and regulated by a rectal temperature probe (SA Instruments Inc., NY). MR Measurements A detailed description of the hardware design is given in supporting information. All MR data were acquired using a 300 MHz (7T) NMR spectrometer (Apollo, Tecmag, Inc.) with micro-imaging capability. The phantom images were obtained using micro-imaging probe heads and an 18 mm I.D. imaging RF coil (Doty Scientific Inc., USA) with self-shielded gradient systems and a maximum strength of 1000 mT/m in the x-, y-, and zdirections (Resonance Research Inc., Billerica, MA). Animal images were obtained using microimaging probe heads and a 35 mm I.D. imaging RF coil with self-shielded gradient systems and a maximum strength of 700 mT/m in the x-, y-, and z-directions (Doty Scientific Inc., USA).
FIG. 1. Block diagram of frequency lock-in feedback circuit. The functions of the system as follows: the RF signal is picked up at the output of the preamplifier receiver by a directional coupler and fed into a demodulator and down-converter to audio frequency range (KHz). Next, the obtained water signal is selectively retained by low-pass filters. The low-frequency water signal is then up-converted back to the original resonance frequency. Then, after phase and gain adjustment, the signal is fed back into the probe by a time-sharing mechanism. The selected feedback signal, which affects the magnetization, then adjusts to the same frequency more efficiently. Consequently, it suppresses the magnetizations selected. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
1678
Chen et al.
FIG. 2. a: Frequency distribution of magnetization A (blue line) and B (red line) (frequency distribution 30 and 100 Hz, respectively). Population ratio is 7/3. For A, T1 ¼ 1 s, T2 ¼ 0.9 s, and for B, T1 ¼ 1 s, T2 ¼ 0.01 s, which simulates fast T2 in the SPIO solution. b: Time evolution of two magnetizations A (blue line) and B (red line) described in (a). Thin and thick lines present the time evolution from 0 to 70 ms and 0 to 40 ms, respectively. It can be found that the two magnetizations move along different trajectories due to the feedback field. The magnetization A is affected by feedback more efficiently and driven faster to þz-direction while the magnetization B is only driven to near null point. The best contrast between magnetization A and B is found at 40 ms. The difference between the two magnetizations at 40 ms is indicated by a red arrow. c–e: Images of the SPIO solution (inner tube)/water (outer tube) phantom. One can see that lockin suppressed images (c) show a positive contrast in the SPIO solution region while both conventional spin-echo images (d) and gradient-echo images (e) present a darker contrast. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
The pulse sequence initially applied a RF pulse with excitation angle 175 to provide an initial feedback field. Following the hard pulse, the sample evolved in the presence of the lock-in frequency feedback field for an evolution time, s, optimized in each study. After the evolution time, a spoiler gradient was applied to transverse magnetization, and the longitudinal component of magnetization was imaged with a conventional spin-echo pulse sequence. The field of view of axial images in the phantom test and in vivo studies were 20 20 and 40 40 mm, respectively, with a slice thickness of 1 mm. For T1-weighted images (T1WI) and active feedback field enhanced images, the relaxation delay was set to 5 T1 (7.5 s); thus, TR values for phantom and in vivo image measurements were 15 s and 7.5 s, respectively. Echo time was set to 10 ms to meet the minimum requirement of spin-echo pulse sequence. The T2-weighted images (T2WI) of the phantom were acquired by spin-echo imaging with TE ¼ 50 ms. The T2*-weighted images (T2*WI) of the phantom were acquired by gradient-echo imaging with TE ¼ 20 ms.
RESULTS Overview of Hardware Design The function of lock-in frequency feedback suppression is similar to other frequency-sensitive MRI methods but it fundamentally differs in that it is sensitive to frequency distributions and the relative contribution of each frequency component to total magnetization. This concept uses the RF field to drive magnetization at specific frequencies to nullify their contribution to the overall signal. Then, the unaffected magnetizations will show a positive contrast in the resultant images. The principle behind the frequency lock-in feedback loop is to filter, phase adjust, and amplify the signal from the spectrometer and then retransmit the modified signal to the RF coil (Fig. 1). The design of the circuit focuses on filtration and amplification, and the feedback signal is extracted from the original FID by means of a directional coupler. Because of signal attenuation due to coupling, a low noise amplifier is used before down-conversion of the audio frequency range (kHz) by quadrature demodulation. Subsequently, the signal is passed through an
Novel MR Contrast by Frequency Lock-In Technique
1679
FIG. 3. a: Frequency distribution of magnetization A (blue line) and B (red line) with a frequency offset of 300 Hz. The population ratio is 7/3. Both magnetizations have T1 ¼ 1 s and T2 ¼ 0.9 s. b: Simulations of signal variations of two magnetizations by gradient-echo imaging. Both signals decay at almost the same rate and cannot be differentiated even over longer timescales. c: Time evolution of magnetization A (blue line) and B (red line) described in (a) by lock-in suppressed feedback imaging. Thin and thick lines present the time evolution from 0 to 60 ms and 0 to 42 ms, respectively. Magnetization A was driven efficiently by the feedback field and moved directly to þz while magnetization B initially moves around z-axis. In the end, both magnetizations spin around in different region of the space. The best contrast between magnetization A and B is found at 42 ms. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
active Bessel-type low pass filter that can be adjusted to a narrow excitation bandwidth, allowing the appropriate frequency to pass through the filter. After filtering, the signal is up-converted to the original frequency (MHz) by a modulator. To achieve frequency selective amplification, the carrier frequency is tuned to provide a feedback field with the frequency of interest. A digitally controlled phase shifter, which enables effective and precise phase control, is incorporated to optimize the feedback field effects. The maximum amplification gain of the feedback loop is estimated to be 41 dB; however, the exact gain depends on experimental conditions. Discrimination of Magnetizations with Same Central Frequency but Different Frequency Distributions The magnetizations with the same central frequency but different frequency distributions are typically used to
produce MR images with T2 contrast agents. Hence, a phantom containing a T2 contrast agent (i.e., superparamagnetic iron oxide (SPIO) nanoparticle solution and deionized [D.I.] water) was prepared to validate the effect of lock-in suppressed imaging. The simulation model indicates that the difference between the SPIO solution and D.I. water resides mainly in their frequency distributions (Fig. 2a). SPIO solution has broader frequency distribution and usually causes faster T2* relaxation and a darker contrast in MR images. By applying lock-in suppressed imaging, the time evolution of longitudinal magnetization, Mz, suggests that the SPIO solution demonstrates stronger signal intensity than D.I. water (Fig. 2b). The dynamics are primarily affected by the filtered feedback field. Only magnetizations near the central frequency are driven. Because magnetization with a broader frequency distribution is less influenced by the feedback field, the magnetizations of these two
1680
Chen et al.
FIG. 4. a: Images of tumors obtained by lock-in suppressed imaging (feedback evolution times set to 18.2 ms, 19.5 ms, and 23.4 ms) show different contrasts in the tumor features at different evolution times. In contrast, (b) conventional T2weighted images of tumors only present the profile of the tumor with very limited detail on the structures. (The echo times (TE) for spin-echo imaging are 15, 30, and 50 ms).
components would gradually separate in the z-direction leading to a stronger observed imaging contrast through z-magnetization. This indicates that lock-in suppressed imaging can efficiently suppress the magnetizations with a narrow frequency distribution at specific times (Fig. 2b). From a time domain perspective, the faster decay signal can be preserved. Phantom images confirm the theoretical simulation prediction. Lock-in suppressed imaging at various recovery times shows a bright contrast with use of the SPIO solution (Fig. 2c). In contrast, conventional T2-weighted (T2WI) (Fig. 2d) and T22*-weighted imaging (T2*WI) (Fig. 2e) demonstrate a darker contrast.
Discrimination of Magnetizations with Small Frequency Shift: Contrast Enhancement of In Vivo Tumor Image The simulation model indicates that the two magnetizations have the same frequency distribution and MR properties but slightly different frequencies (Fig. 3a). However, while gradient-echo imaging fails to distinguish between the two magnetizations (Fig. 3b), lock-in suppressed imaging can separate the longitudinal magnetization contribution, Mz (Fig. 3c). The dynamics shows that magnetizations with different frequencies are driven to two different regions. Magnetizations coinciding with the lock-in frequency are mainly affected by the filtered feedback field and driven almost vertically while the others rotate around the z-axis due to off-resonance as indicated in Eq. [2]. Consequently, magnetization in the frequency-selected regime is returned to the þz-axis comparatively quickly. Compared with conventional MRI, lock-in suppressed MRI may improve the differentiation of lesions, as pathological changes can minimally alter water content, blood supply, and cell morphology. To demonstrate the applicability of lock-in suppressed contrast to in vivo lesion detection, imaging was performed
on mice with embedded tumors. Among conventional methods, T2WI provided the best overall contrast by highlighting the tumors as hyperintense regions against the surrounding muscle tissue. Figures 3a and 3b show representative lock-in suppression enhanced and conventional MR images. Under lock-in feedback (Fig. 4a), variable contrast in the tumor region is evident while conventional T2WI only shows the tumor profile. Contrast in lock-in suppressed images may result from differences in cell morphology at different stages of a developing tumor. Using time evolution mapping of lock-in suppressed images can highlight the regimes of tumor progression (Fig. 5a). The regimes exhibiting higher (more red) and lower intensity (more blue) (Fig. 5a) indicate living cells and necrosis in histopathology (Figs. 5b and c), respectively. Herein, the lock-in suppressed RF field provides selective excitation of specific frequency components, thereby enabling more efficient contrast enhancement.
DISCUSSION AND CONCLUSIONS Lock-in suppressed imaging introduces a conceptually new approach to MRI. Evolution under the reaction fields allows the spins to play an active role in determining and differentiating subsequent evolution, thereby improving the distinction between regions with different MR properties. Furthermore, by adjusting the filter bandwidth, phase, and amplitude of the feedback field dynamically adds a new line to the pulse sequence and allows additional control over the spin degrees of freedom. The dynamics of both of the imaging experiments presented above show that the magnetization locked by a lock-in RF feedback field is driven effectively. In contrast, off-resonance magnetizations with respect to the lock-in RF feedback field are, relatively, only mildly affected. Consequently, the dynamics of two magnetizations move along with different trajectory (Figs. 2b and
Novel MR Contrast by Frequency Lock-In Technique
1681
for differentiating between small structural variations in biological tissues. The contrast in in vivo tumor lock-in suppressed images provides higher spatial discrimination to assist in distinguishing between the regimes of necrosis and activation consistent with pathologic results. Heightened sensitivity to underlying susceptibility variations and their relative contribution to total magnetization may thus be achieved to yield new and enhanced contrast. However, there are also some limitations for this technique. Because the feedback field originates from the transverse magnetizations, weak transverse magnetizations cause feedback field to be inefficient. Therefore, good field homogeneity and high proton density in the imaging regime are required. Nevertheless, comparing with other suppression methods, such as presaturating methods, the frequency lock-in suppression technique still demonstrates a convenient and more efficient way for contrast enhancement because its frequency distribution is determined by the magnetization itself. ACKNOWLEDGMENT This study was supported by the National Science Council of the Republic of China and the College of Science of the National Chung Cheng University. REFERENCES
FIG. 5. a: Time evolution mapping of lock-in suppressed images and (b) corresponding histopathology. c: Magnified view of the picture in (b). One can see the highlighted contrast in MR images (a) agrees well with the regime of living cells and necrosis which and $ (c), respectively. are marked by
*
3c). Then, by acquiring the signal at specific times, enhanced contrast can be obtained. In the first experiment, water in the SPIO solution usually has broader frequency distribution. When applying the lock-in suppressed imaging, the magnetizations outside the lockin frequency moves more slowly than other magnetizations. Hence, it gives rise the positive contrast in SPIO solution phantom, in contrast to what would be expected from a conventional imaging method. In addition, the second experiment using magnetizations with a small shift in frequency and broad frequency distributions is useful
1. Huang SY, Furuyama JK, Lin YY. Designing feedback-based contrast enhancement for in vivo imaging. Magn Reson Mater Phys Biol Med 2006;19:333–346. 2. Huang SY, Wolahan SM, Mathern GW, Chute DJ, Akhtari M, Nguyen ST, Huynh MN, Salamon N, Lin YY. Improving MRI differentiation of gray and white matter in epileptogenic lesions based on nonlinear feedback. Magn Reson Med 2006;56:776–786. 3. Price WS, Walchli M. NMR diffusion measurements of strong signals: the PGSE-Q-switch experiment. Magn Reson Chem 2002;40:S128–S132. 4. Bloom S. Effects of radiation damping on spin dynamics. J Appl Phys 1957;28:800–805. 5. Szoke A, Meiboom S. Radiation damping in nuclear magnetic resonance. Phys Rev 1959;113:585–586. 6. Warren WS, Hammes SL, Bates JL. Dynamics of radiation damping in nuclear magnetic resonance. J Chem Phys 1989;91:5895–5904. 7. Vlassenbroek A, Jeener J, Broekaert P. Radiation damping in highresolution liquid NMR-a simulation study. J Chem Phys 1995;103: 5886–5897. 8. Szantay C, Demeter A. Radiation damping diagnostics. Concepts Magn Reson 1999;11:121–145. 9. Lin YY, Lisitza N, Ahn SD, Warren WS. Resurrection of crushed magnetization and chaotic dynamics in solution NMR spectroscopy. Science 2000;290:118–121. 10. Huang SY, Lin YY, Lisitza N, Warren WS. Signal interferences from turbulent spin dynamics in solution nuclear magnetic resonance spectroscopy. J Chem Phys 2002;116:10325–10337. 11. Huang SY, Walls JD, Wang Y, Warren WS, Lin YY. Signal irreproducibility in high-field solution magnetic resonance experiments caused by spin turbulence. J Chem Phys 2004;121:6105–6109. 12. Abergel D, LouisJoseph A, Lallemand JY. Amplification of radiation damping in a 600-MHz NMR spectrometer: application to the study of water-protein interactions. J Biomol NMR 1996;8:15–22. 13. Broekaert P, Jeener J. Suppression of radiation damping in NMR in liquids by active electronic feedback. J Magn Reson Ser A 1995;113:60–64. 14. Louisjoseph A, Abergel D, Lallemand JY. Neutralization of radiation damping by selective feedback on a 400-MHz NMR spectrometer. J Biomol NMR 1995;5:212–216. 15. Louis-Joseph A, Abergel D, Lebars I, Lallemand JY. Enhancement of water suppression by radiation damping-based manipulation of residual water in Jump and Return NMR experiments. Chem Phys Lett 2001;337:92–96. 16. Abergel D. Chaotic solutions of the feedback driven Bloch equations. Phys Lett A 2002;302:17–22.
Magnetic Resonance in Medicine 71:1676–1681 (2014)
Novel MRI Contrast Development by Lock-In Suppression Yu-Wen Chen,1 Chou-Hsiung Hsu,2,3 and Dennis W. Hwang1,4* induction decay (FID) signal to the detection hardware. However, a major issue with MRI lies in the lack of contrast in in vivo images between tissues with small variations in magnetic susceptibility or relaxation rates and not the amount of signal that can be detected. To resolve this problem, using feedback from a radiation damping (RD) field to selectively drive magnetization toward equilibrium much faster than the inherent relaxation rate, as described by Lenz’s law, is proposed. RD is shown to amplify the contrast when compared with standard MR sequences (1–15). Furthermore, nonlinear effects triggered by the associative coupling to RD fields amplify the effects of nonuniformity (9,16), creating a macroscopic signal to enhance ex vivo MRI contrast between gray and white matter at 14 Tesla (T) in epileptic patients with symptomatic lesions (1). On the basis of a concept similar to RD, we designed an active feedback device. Our goal is to identify and amplify a highly selective feedback field frequency to gain a significant contrast enhancement. This study focuses on a concept of frequency lock-in suppression and shows that the feedback RF field can image resonant frequency differences between magnetizations with small chemical shifts or frequency distributions.
Purpose: The goal of this study is to develop novel MR contrast by frequency lock-in technique. Methods: An electronic feedback device that can control the frequency and bandwidth of the feedback RF field is presented. In this study, the effects of lock-in suppressed imaging are discussed both theoretically and experimentally. Results: Two important imaging experiments were performed. The first experiment used magnetizations with the same central frequency but different frequency distributions and was compared with MR images obtained with T2 contrast agents. Lock-in suppressed images showed an improvement in contrast relative to the conventional imaging method. The second experiment used magnetizations with small shifts in frequency and a broad frequency distribution. This is helpful for differentiating between small structural variations in biological tissues. The contrast achieved in in vivo tumor imaging using the lock-in suppressed technique provide higher spatial resolutions and discriminate the regimes of necrosis and activation consistent with pathologic results. Conclusion: Lock-in suppressed imaging introduces a conceptually new approach to MRI. Heightened sensitivity to underlying susceptibility variations and their relative contribution to total magnetization may thus be achieved to yield new and enhanced contrast. Magn Reson Med 71:1676–1681, C 2014 Wiley Periodicals, Inc. 2014. V Key words: suppression
MRI;
feedback
field;
radiation
damping;
THEORY
INTRODUCTION MRI is an important noninvasive technique widely used in medical diagnosis. In MR experiments, the application of a radiofrequency (RF) pulse tips magnetization toward the transverse plane; subsequently precessing magnetization induces a current in the probe coil that carries the free 1
Department of Chemistry and Biochemistry, National Chung Cheng University, Min-Hsiung Township Chiayi, Taiwan. 2 Molecular Science and Technology Program, Taiwan International Graduate Program, Institute of Atomic and Molecular Sciences, Academia Sinica, Taipei, Taiwan. 3 Department of Chemistry, National Tsing-Hua University, Hsinchu, Taiwan. 4 Center for Nano Bio-detection, AIM-HI, National Chung Cheng University, Min-Hsiung Township Chiayi, Taiwan. Grant sponsor: the National Science Council of the Republic of China; Grant number: NSC 101–2113-M-194-006-MY2. *Correspondence to: Dennis W. Hwang, Ph.D., National Chung Cheng University, Department of Chemistry and Biochemistry, 168 University Road, Min-Hsiung Township Chiayi, Taiwan 62102. E-mail: [email protected] Additional Supporting Information may be found in the online version of this article. Received 17 October 2013; revised 12 January 2014; accepted 14 January 2014 DOI 10.1002/mrm.25162 Published online 6 March 2014 in Wiley Online Library (wileyonlinelibrary. com). C 2014 Wiley Periodicals, Inc. V
A feedback RF field is generated by artificially amplifying the FID and feeding the signal back into the coil with a certain phase. This user-generated active feedback field can be expressed by
BF ðv; tÞ ¼
iGeif x þ hMx ðr; tÞi^ y Þ; ðhMy ðr; tÞi^ g
[1]
where G 5 vggQM0 /2 is the amplification gain factor of the feedback RF field. The amplification gain factor is a measure of field strength, where g is the amplification gain by the active feedback circuit, g is the filling factor, and M0 is equilibrium magnetization. Mþ ðv; r; tÞ is the complex transverse magnetization integrated over sample volume with specific frequency, because the feedback field is dependent on net transverse magnetization. u is the phase between the feedback RF field and the FID. For example, when u ¼ 0 , the feedback RF field is in phase with the FID, and the feedback RF field is orthogonal to transverse magnetization, which is determined by the tuning condition of the RF coil circuit and can also be controlled by the feedback circuit. By including Eq. [1] in Bloch equations, dM=dt ¼ gM B, the time evolution of magnetization in the absence of relaxation can be expressed as
1676
Novel MR Contrast by Frequency Lock-In Technique
2
Mx ðr; tÞ
3
2
1677
0
7 6 @6 6 My ðr; tÞ 7 ¼ 6 dvðrÞ 5 4 @t 4 if Mz ðr; tÞ iGe hMx ðr; tÞi
dvðrÞ 0 iGe
if
where dvðrÞ represents the frequency offset at position r. In Eq. [1], because the feedback signal is filtered by a low pass filter, a feedback RF field with a particular frequency bandwidth can be generated. The off-resonance frequency causes oscillation between transverse magnetizations, Mx and My, and becomes less efficient in driving longitudinal magnetization, Mz. Hence the evolution of Mz with a different frequency can be differentiated. Incorporating with simulation of signal filter and timeshare mechanism, the detailed dynamics can be studied by numerical simulation using Matlab (The MathWorks, USA) and given in the supporting information. METHODS Animal Preparation Three-week-old athymic nude mice (BALB/cAnN.CgFoxn1nu/Cr1Narl) were ordered from the National Laboratory Animal Center (Taiwan). All mice were kept under specific pathogen-free conditions using a laminar airflow rack with free access to sterilized food and autoclaved water. All experiments were conducted after obtaining the relevant license from the Animal Experimentation Ethics Committee of the National Chung Cheng University. 2 105 cells of Oral cancer cell (OML1-R50) were resuspended in a mixture of 0.1 mL medium and Matrigel (1:1) (BD Bioscience, San Jose, CA). Cell suspension was then injected subcutaneously into the flank of each 4-
hMy ðr; tÞi
iGeif hMx ðr; tÞi
32
Mx ðr; tÞ
3
7 76 6 7 iGeif hMy ðr; tÞi 7 54 My ðr; tÞ 5 0
[2]
Mz ðr; tÞ
week-old nude mouse. At the end of experiment, all mice were killed by cervical dislocation. In vivo mice studies were conducted under anesthesia with an initial dosage of 2.0% isoflurane in air and using a small animal gating system (SA Instruments Inc., NY). For maintenance, isoflurane was set under 1.0% and gated by means of a respiration trigger sensor. The respiration rate of the mice was maintained under 60 breaths per min. Warm air was supplied at 25 6 2 C to avoid hypothermia and regulated by a rectal temperature probe (SA Instruments Inc., NY). MR Measurements A detailed description of the hardware design is given in supporting information. All MR data were acquired using a 300 MHz (7T) NMR spectrometer (Apollo, Tecmag, Inc.) with micro-imaging capability. The phantom images were obtained using micro-imaging probe heads and an 18 mm I.D. imaging RF coil (Doty Scientific Inc., USA) with self-shielded gradient systems and a maximum strength of 1000 mT/m in the x-, y-, and zdirections (Resonance Research Inc., Billerica, MA). Animal images were obtained using microimaging probe heads and a 35 mm I.D. imaging RF coil with self-shielded gradient systems and a maximum strength of 700 mT/m in the x-, y-, and z-directions (Doty Scientific Inc., USA).
FIG. 1. Block diagram of frequency lock-in feedback circuit. The functions of the system as follows: the RF signal is picked up at the output of the preamplifier receiver by a directional coupler and fed into a demodulator and down-converter to audio frequency range (KHz). Next, the obtained water signal is selectively retained by low-pass filters. The low-frequency water signal is then up-converted back to the original resonance frequency. Then, after phase and gain adjustment, the signal is fed back into the probe by a time-sharing mechanism. The selected feedback signal, which affects the magnetization, then adjusts to the same frequency more efficiently. Consequently, it suppresses the magnetizations selected. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
1678
Chen et al.
FIG. 2. a: Frequency distribution of magnetization A (blue line) and B (red line) (frequency distribution 30 and 100 Hz, respectively). Population ratio is 7/3. For A, T1 ¼ 1 s, T2 ¼ 0.9 s, and for B, T1 ¼ 1 s, T2 ¼ 0.01 s, which simulates fast T2 in the SPIO solution. b: Time evolution of two magnetizations A (blue line) and B (red line) described in (a). Thin and thick lines present the time evolution from 0 to 70 ms and 0 to 40 ms, respectively. It can be found that the two magnetizations move along different trajectories due to the feedback field. The magnetization A is affected by feedback more efficiently and driven faster to þz-direction while the magnetization B is only driven to near null point. The best contrast between magnetization A and B is found at 40 ms. The difference between the two magnetizations at 40 ms is indicated by a red arrow. c–e: Images of the SPIO solution (inner tube)/water (outer tube) phantom. One can see that lockin suppressed images (c) show a positive contrast in the SPIO solution region while both conventional spin-echo images (d) and gradient-echo images (e) present a darker contrast. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
The pulse sequence initially applied a RF pulse with excitation angle 175 to provide an initial feedback field. Following the hard pulse, the sample evolved in the presence of the lock-in frequency feedback field for an evolution time, s, optimized in each study. After the evolution time, a spoiler gradient was applied to transverse magnetization, and the longitudinal component of magnetization was imaged with a conventional spin-echo pulse sequence. The field of view of axial images in the phantom test and in vivo studies were 20 20 and 40 40 mm, respectively, with a slice thickness of 1 mm. For T1-weighted images (T1WI) and active feedback field enhanced images, the relaxation delay was set to 5 T1 (7.5 s); thus, TR values for phantom and in vivo image measurements were 15 s and 7.5 s, respectively. Echo time was set to 10 ms to meet the minimum requirement of spin-echo pulse sequence. The T2-weighted images (T2WI) of the phantom were acquired by spin-echo imaging with TE ¼ 50 ms. The T2*-weighted images (T2*WI) of the phantom were acquired by gradient-echo imaging with TE ¼ 20 ms.
RESULTS Overview of Hardware Design The function of lock-in frequency feedback suppression is similar to other frequency-sensitive MRI methods but it fundamentally differs in that it is sensitive to frequency distributions and the relative contribution of each frequency component to total magnetization. This concept uses the RF field to drive magnetization at specific frequencies to nullify their contribution to the overall signal. Then, the unaffected magnetizations will show a positive contrast in the resultant images. The principle behind the frequency lock-in feedback loop is to filter, phase adjust, and amplify the signal from the spectrometer and then retransmit the modified signal to the RF coil (Fig. 1). The design of the circuit focuses on filtration and amplification, and the feedback signal is extracted from the original FID by means of a directional coupler. Because of signal attenuation due to coupling, a low noise amplifier is used before down-conversion of the audio frequency range (kHz) by quadrature demodulation. Subsequently, the signal is passed through an
Novel MR Contrast by Frequency Lock-In Technique
1679
FIG. 3. a: Frequency distribution of magnetization A (blue line) and B (red line) with a frequency offset of 300 Hz. The population ratio is 7/3. Both magnetizations have T1 ¼ 1 s and T2 ¼ 0.9 s. b: Simulations of signal variations of two magnetizations by gradient-echo imaging. Both signals decay at almost the same rate and cannot be differentiated even over longer timescales. c: Time evolution of magnetization A (blue line) and B (red line) described in (a) by lock-in suppressed feedback imaging. Thin and thick lines present the time evolution from 0 to 60 ms and 0 to 42 ms, respectively. Magnetization A was driven efficiently by the feedback field and moved directly to þz while magnetization B initially moves around z-axis. In the end, both magnetizations spin around in different region of the space. The best contrast between magnetization A and B is found at 42 ms. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
active Bessel-type low pass filter that can be adjusted to a narrow excitation bandwidth, allowing the appropriate frequency to pass through the filter. After filtering, the signal is up-converted to the original frequency (MHz) by a modulator. To achieve frequency selective amplification, the carrier frequency is tuned to provide a feedback field with the frequency of interest. A digitally controlled phase shifter, which enables effective and precise phase control, is incorporated to optimize the feedback field effects. The maximum amplification gain of the feedback loop is estimated to be 41 dB; however, the exact gain depends on experimental conditions. Discrimination of Magnetizations with Same Central Frequency but Different Frequency Distributions The magnetizations with the same central frequency but different frequency distributions are typically used to
produce MR images with T2 contrast agents. Hence, a phantom containing a T2 contrast agent (i.e., superparamagnetic iron oxide (SPIO) nanoparticle solution and deionized [D.I.] water) was prepared to validate the effect of lock-in suppressed imaging. The simulation model indicates that the difference between the SPIO solution and D.I. water resides mainly in their frequency distributions (Fig. 2a). SPIO solution has broader frequency distribution and usually causes faster T2* relaxation and a darker contrast in MR images. By applying lock-in suppressed imaging, the time evolution of longitudinal magnetization, Mz, suggests that the SPIO solution demonstrates stronger signal intensity than D.I. water (Fig. 2b). The dynamics are primarily affected by the filtered feedback field. Only magnetizations near the central frequency are driven. Because magnetization with a broader frequency distribution is less influenced by the feedback field, the magnetizations of these two
1680
Chen et al.
FIG. 4. a: Images of tumors obtained by lock-in suppressed imaging (feedback evolution times set to 18.2 ms, 19.5 ms, and 23.4 ms) show different contrasts in the tumor features at different evolution times. In contrast, (b) conventional T2weighted images of tumors only present the profile of the tumor with very limited detail on the structures. (The echo times (TE) for spin-echo imaging are 15, 30, and 50 ms).
components would gradually separate in the z-direction leading to a stronger observed imaging contrast through z-magnetization. This indicates that lock-in suppressed imaging can efficiently suppress the magnetizations with a narrow frequency distribution at specific times (Fig. 2b). From a time domain perspective, the faster decay signal can be preserved. Phantom images confirm the theoretical simulation prediction. Lock-in suppressed imaging at various recovery times shows a bright contrast with use of the SPIO solution (Fig. 2c). In contrast, conventional T2-weighted (T2WI) (Fig. 2d) and T22*-weighted imaging (T2*WI) (Fig. 2e) demonstrate a darker contrast.
Discrimination of Magnetizations with Small Frequency Shift: Contrast Enhancement of In Vivo Tumor Image The simulation model indicates that the two magnetizations have the same frequency distribution and MR properties but slightly different frequencies (Fig. 3a). However, while gradient-echo imaging fails to distinguish between the two magnetizations (Fig. 3b), lock-in suppressed imaging can separate the longitudinal magnetization contribution, Mz (Fig. 3c). The dynamics shows that magnetizations with different frequencies are driven to two different regions. Magnetizations coinciding with the lock-in frequency are mainly affected by the filtered feedback field and driven almost vertically while the others rotate around the z-axis due to off-resonance as indicated in Eq. [2]. Consequently, magnetization in the frequency-selected regime is returned to the þz-axis comparatively quickly. Compared with conventional MRI, lock-in suppressed MRI may improve the differentiation of lesions, as pathological changes can minimally alter water content, blood supply, and cell morphology. To demonstrate the applicability of lock-in suppressed contrast to in vivo lesion detection, imaging was performed
on mice with embedded tumors. Among conventional methods, T2WI provided the best overall contrast by highlighting the tumors as hyperintense regions against the surrounding muscle tissue. Figures 3a and 3b show representative lock-in suppression enhanced and conventional MR images. Under lock-in feedback (Fig. 4a), variable contrast in the tumor region is evident while conventional T2WI only shows the tumor profile. Contrast in lock-in suppressed images may result from differences in cell morphology at different stages of a developing tumor. Using time evolution mapping of lock-in suppressed images can highlight the regimes of tumor progression (Fig. 5a). The regimes exhibiting higher (more red) and lower intensity (more blue) (Fig. 5a) indicate living cells and necrosis in histopathology (Figs. 5b and c), respectively. Herein, the lock-in suppressed RF field provides selective excitation of specific frequency components, thereby enabling more efficient contrast enhancement.
DISCUSSION AND CONCLUSIONS Lock-in suppressed imaging introduces a conceptually new approach to MRI. Evolution under the reaction fields allows the spins to play an active role in determining and differentiating subsequent evolution, thereby improving the distinction between regions with different MR properties. Furthermore, by adjusting the filter bandwidth, phase, and amplitude of the feedback field dynamically adds a new line to the pulse sequence and allows additional control over the spin degrees of freedom. The dynamics of both of the imaging experiments presented above show that the magnetization locked by a lock-in RF feedback field is driven effectively. In contrast, off-resonance magnetizations with respect to the lock-in RF feedback field are, relatively, only mildly affected. Consequently, the dynamics of two magnetizations move along with different trajectory (Figs. 2b and
Novel MR Contrast by Frequency Lock-In Technique
1681
for differentiating between small structural variations in biological tissues. The contrast in in vivo tumor lock-in suppressed images provides higher spatial discrimination to assist in distinguishing between the regimes of necrosis and activation consistent with pathologic results. Heightened sensitivity to underlying susceptibility variations and their relative contribution to total magnetization may thus be achieved to yield new and enhanced contrast. However, there are also some limitations for this technique. Because the feedback field originates from the transverse magnetizations, weak transverse magnetizations cause feedback field to be inefficient. Therefore, good field homogeneity and high proton density in the imaging regime are required. Nevertheless, comparing with other suppression methods, such as presaturating methods, the frequency lock-in suppression technique still demonstrates a convenient and more efficient way for contrast enhancement because its frequency distribution is determined by the magnetization itself. ACKNOWLEDGMENT This study was supported by the National Science Council of the Republic of China and the College of Science of the National Chung Cheng University. REFERENCES
FIG. 5. a: Time evolution mapping of lock-in suppressed images and (b) corresponding histopathology. c: Magnified view of the picture in (b). One can see the highlighted contrast in MR images (a) agrees well with the regime of living cells and necrosis which and $ (c), respectively. are marked by
*
3c). Then, by acquiring the signal at specific times, enhanced contrast can be obtained. In the first experiment, water in the SPIO solution usually has broader frequency distribution. When applying the lock-in suppressed imaging, the magnetizations outside the lockin frequency moves more slowly than other magnetizations. Hence, it gives rise the positive contrast in SPIO solution phantom, in contrast to what would be expected from a conventional imaging method. In addition, the second experiment using magnetizations with a small shift in frequency and broad frequency distributions is useful
1. Huang SY, Furuyama JK, Lin YY. Designing feedback-based contrast enhancement for in vivo imaging. Magn Reson Mater Phys Biol Med 2006;19:333–346. 2. Huang SY, Wolahan SM, Mathern GW, Chute DJ, Akhtari M, Nguyen ST, Huynh MN, Salamon N, Lin YY. Improving MRI differentiation of gray and white matter in epileptogenic lesions based on nonlinear feedback. Magn Reson Med 2006;56:776–786. 3. Price WS, Walchli M. NMR diffusion measurements of strong signals: the PGSE-Q-switch experiment. Magn Reson Chem 2002;40:S128–S132. 4. Bloom S. Effects of radiation damping on spin dynamics. J Appl Phys 1957;28:800–805. 5. Szoke A, Meiboom S. Radiation damping in nuclear magnetic resonance. Phys Rev 1959;113:585–586. 6. Warren WS, Hammes SL, Bates JL. Dynamics of radiation damping in nuclear magnetic resonance. J Chem Phys 1989;91:5895–5904. 7. Vlassenbroek A, Jeener J, Broekaert P. Radiation damping in highresolution liquid NMR-a simulation study. J Chem Phys 1995;103: 5886–5897. 8. Szantay C, Demeter A. Radiation damping diagnostics. Concepts Magn Reson 1999;11:121–145. 9. Lin YY, Lisitza N, Ahn SD, Warren WS. Resurrection of crushed magnetization and chaotic dynamics in solution NMR spectroscopy. Science 2000;290:118–121. 10. Huang SY, Lin YY, Lisitza N, Warren WS. Signal interferences from turbulent spin dynamics in solution nuclear magnetic resonance spectroscopy. J Chem Phys 2002;116:10325–10337. 11. Huang SY, Walls JD, Wang Y, Warren WS, Lin YY. Signal irreproducibility in high-field solution magnetic resonance experiments caused by spin turbulence. J Chem Phys 2004;121:6105–6109. 12. Abergel D, LouisJoseph A, Lallemand JY. Amplification of radiation damping in a 600-MHz NMR spectrometer: application to the study of water-protein interactions. J Biomol NMR 1996;8:15–22. 13. Broekaert P, Jeener J. Suppression of radiation damping in NMR in liquids by active electronic feedback. J Magn Reson Ser A 1995;113:60–64. 14. Louisjoseph A, Abergel D, Lallemand JY. Neutralization of radiation damping by selective feedback on a 400-MHz NMR spectrometer. J Biomol NMR 1995;5:212–216. 15. Louis-Joseph A, Abergel D, Lebars I, Lallemand JY. Enhancement of water suppression by radiation damping-based manipulation of residual water in Jump and Return NMR experiments. Chem Phys Lett 2001;337:92–96. 16. Abergel D. Chaotic solutions of the feedback driven Bloch equations. Phys Lett A 2002;302:17–22.
