Physics, P-DO
|
Ultra-Low-Field Nuclear Magnetic Resonance and Magnetic Resonance Imaging
A.N. Matlachov, P.L. Volegov, M.A. Espy, J.C. Mosher, J.S. George, R.H. Kraus, Jr. (P-21) IntroductionA primary thrust in clinical nuclear magnetic resonance (NMR) spectroscopy and magnetic resonance imaging (MRI) has been towards ever-higher magnetic-field strengths. For example, clinical images increasingly are moving from 1.5 T to 3–4 T and research instruments for humans are operating at 7 T or higher. This is largely motivated by the enhanced sensitivity at high fields due to increased polarization and increasing detection efficiency at higher frequencies. Nevertheless, very low field (VLF, in the mT range) and ultra-low-field (ULF, in the μT range) NMR and MRI are areas of active interest. Cost and size of systems could be significantly reduced. VLF and ULF systems could be easily portable and the sample need not be restricted to the interior of a magnet bore (ex situ or "inside out" imaging). Other interests are driven by the complications of using high-field MRI with samples containing metal (i.e., subjects with metal pins or implants), which are minimized at low fields. Furthermore, we recently demonstrated that samples contained entirely inside metallic shells may also be imaged at ULF.1 Measurement fields do not have to be highly homogeneous to achieve narrow NMR line widths at ULF. Moreover, for a fixed relative homogeneity, the NMR line width scales linearly with the strength of the measurement field, allowing the possibility of very narrow NMR lines with high signal-to-noise at ULF.2 Susceptibility artifacts caused by coupling between the applied magnetic field and different sample materials broaden resonance lines at high fields but are significantly reduced at ULF. The absence of such artifacts may provide opportunities for novel forms of functional imaging at ULF. For example, it may be possible to manipulate T1 (longitudinal relaxation time of the spin polarization) contrast at low field strength to provide significant contrast not realizable at high fields. SQUIDs and Magnetic ScanningNMR spectroscopy detects the magnetic signature of nuclear spins precessing in the measurement magnetic field. At low field strength, signals become increasingly difficult to measure with conventional detectors. Superconducting quantum interference devices (SQUIDs) are magnetic-flux-to-voltage converters of exquisite sensitivity with a response that is independent of frequency. For this reason, a number of low-field NMR systems have employed SQUID sensors at measurement fields below 10 mT, using both high-Tc (liquid-nitrogen cooled) and low-Tc (liquid-helium cooled) SQUIDs. Low-Tc SQUIDs provide higher sensitivity (due primarily to lower thermal noise) and greater reliability and robustness than presently available in high-Tc devices. The frequency-independent response of SQUID detectors also enables one to simultaneously detect the signature from multiple different nuclei, even though their NMR frequencies can differ by factors of 2 or more.3 We have recently demonstrated this in our own system as well.4 Various investigators, including ourselves, have already demonstrated that ULF MRI is possible (see for example, References 1, 3, and 5). In addition, we recently completed the first-ever demonstration of the feasibility of magnetic resonance (MR) measurements with simultaneous recordings of biomagnetic signals from the brain (magnetoencephalography, MEG), heart (magnetocardiography, MCG) and muscle (magnetomyography, MMG), using the same detectors.1,4,6,7 SQUID-based biomagnetic measurements are noninvasive techniques that measure magnetic fields outside the body, for example at the surface of the head in MEG. These fields arise as a direct consequence of electrical activity (neurons or nerves) in the living body. MEG requires the use of SQUID sensors to measure the extraordinarily low-level magnetic fields, usually in the range from 10-15 to 10-12 T, produced by neuronal activity in the brain. While other functional imaging modalities, such as functional MRI, depend on the relatively slow and indirect hemodynamic response of the brain, MEG (and electroencephalography, or EEG) can provide measurement of the electromagnetic fields arising from the actual neuronal currents with submillisecond temporal resolution. Acquiring MR and biomagnetic data simultaneously will reduce most of the sources of error in colocalization of bioelectric sources with anatomy. This will be particularly valuable for MEG and EEG where data typically must be superimposed on conventionally acquired MRI images, introducing significant error. Perhaps the most compelling application of NMR/MRI at ULF is to acquire direct evidence of biomagnetic signals with a tomographic imaging modality. Specifically, it is speculated that the bioelectric currents that produce the signals we measure in MEG, MCG, or MMG may produce changes in the proton precession signal, in either frequency, phase, T2* (total relaxation time of the spin polarization), etc. If such a signature could be detected, it would allow one to tomographically localize the effect of those currents using MR techniques, thereby eliminating the complications of the inverse problem inherent in biomagnetic source localization. Our team has recently begun investigating this phenomenon.4,7 Measurements of ULF NMR and MRI
We have also been able to acquire simple images with our system. Figure 3(a) is a photograph of a 60 mm diam by 52 mm high cylindrical plastic phantom with seven 10 mm diam by 48 mm deep wells. Four of the wells were filled with water (shown filled with colored water for visibility). Figure 3(b) illustrates a two-dimensional (2?]D) image of the phantom constructed from a series of gradient-encoded FID spectra acquired at various angles by rotating the sample within a fixed measurement field. The measurement field was 7.8 μT and the gradient was ~ 7 μT/m. Simultaneous Measurements of Biomagnetic Signals and NMR
The Quest for a Signature or Direct Neuronal Measurements with ULF MROne very compelling application of NMR/MRI at ULF is to acquire direct evidence of biomagnetic responses producing changes in the proton precession signal, in either frequency, phase, T2*, etc. If such a signature could be detected, it would allow one to tomographically localize the effect of those currents using MR techniques, thereby eliminating the complications of the inverse problem inherent in biomagnetic source localization. This would open up a whole new imaging modality. To study these effects we looked at T2* values for NMR signals that were simultaneously acquired in the presence of biomagnetic signals such as those from heart (MCG) and muscle (MMG). Our hypothesis was that the value of T2* in the presence of bioelectric currents would be shorter than that measured when such currents were absent, because inhomogeneity in the local fields should produce a dephasing of spin signals.
We focused on the values of T2* for NMR during the "T" peak (200–250 ms after the "R" peak) and for NMR during the resting phase of the heartbeat (400–700 ms after the "R" peak). These values were calculated by the direct exponential curve resolution algorithm (DECRA)10 and the estimated T2* was extracted from the damped exponential found at the Larmor (precession) frequency. T2* values varied among subjects with the shortest being 76 ms and the longest being 123 ms. We hypothesize that physiological differences among subjects such as fat content in the chest wall contribute to the spread in values. We observed that the values of T2* for the resting phase appeared longer than those for the "T" peak by 2–9 ms for four of the five subjects (as we would expect), however was shorter by 5–7 ms for one subject. Our uncertainty in T2* for water phantoms with varying concentrations of copper sulfate was found to be ± 1 ms. However, it is very difficult to estimate the uncertainty in T2* for measurements where the sample geometry is so much more complicated and variable. Because of the complexity of the muscle responses in the cardiac system, we chose to use MMG to investigate the effect of the bioelectric currents on the NMR signal T2*. The bioelectric currents during MMG are much larger than neuronal currents, and unlike MCG, effects such as motion, blood flow, and blood volume of the sample being measured are significantly reduced. The experiments provide data with interleaved epochs of NMR recorded while the muscles of the forearm were either stressed or relaxed. This protocol was chosen to try and reduce any hemodynamic or metabolic effects. The probability density functions (PDFs) of T2* were then inferred from the data for both stressed and relaxed conditions using a "bootstrap" method.4 The results are shown in Figure 6. The same analysis approach was then applied to random permutations of the stressed and relaxed sets. The inferred PDF for the randomly mixed data is shown in Figure 7. The permutation test (Figure 7) shows no inherent preference in our processing between the two sets, while the bootstrap tests (Figure 6) suggest that the two conditions may be distinct, but with statistically low power. Our observed difference in T2* for the two conditions is not statistically significant; however, it is encouraging that the trend of a shorter T2* for the stressed condition is what one would expect if this effect were due to bioelectric currents dephasing the NMR signal. We caution that even if this effect were statistically significant, we are not yet able to rule out that the measured effect was due to a systematic error due to the slight differences in the experimental configuration between the two cases (i.e., slightly different arm position), some other systematic error in our hardware, or a biological effect that is not electrical in nature. Discussion and ConclusionsWe have demonstrated that biomagnetic signals can be acquired simultaneously with NMR data using SQUID sensors at ultralow magnetic fields. We have demonstrated MRI at these low fields for water phantoms. These results provide the basis of performing MR anatomical imaging simultaneously with bioelectric source localization. Such capability will greatly enhance the efficacy and reduce errors over current functional neuroimaging techniques. In addition, we have investigated the possibility of using MR techniques to tomographically image the direct consequence of bioelectric activity in living tissue. Although there is significant work to be done, we are encouraged that the modality of ULF NMR with SQUIDs is going to be able to see these direct effects and open up a whole new way to gain knowledge regarding bioelectric function. If nothing else, we have demonstrated that simultaneous anatomical and bioelectric images are possible, and that the field of ULF NMR/MRI with SQUIDs has an exciting future. References
AcknowledgmentThe authors wish to thank Dr. C.C. Wood for his thoughtful discussions of the topics presented. Technical development of the instrumentation used for this work was supported by LANL and the Defense Advanced Research Projects Agency. Experimental studies received support from the U.S. DOE Office of Biological and Environmental Research. For further information, contact Robert Kraus, Jr., 505-665-1938, rkraus@lanl.gov. |
Figure 1 shows a schematic diagram of the ULF NMR system designed in Biological and Quantum Physics (P-21). This configuration uses a tangential gradiometer and is optimized for the measurement of simultaneous measurements of MCG and NMR.4 Figure 2 shows the free induction decay (FID, the actual signal of spin precession) from a sample of H2O and C10F18 taken simultaneously, and a fast Fourier transform (FFT) of the signal.4 As noted previously, this would not be possible with a conventional MR system, as the receiver coils would not be able to detect two frequencies so far apart. An example of the utility of such measurements is to food science, where conventional high-field NMR has already become a powerful tool for the detection of moisture content, sugar content, adulteration, bacterial spoilage, etc.8 Our methods suggest a strategy for inspecting food inside metallic cans. We measured the hydrogen-1 NMR signal from tomato juice and cola inside unopened aluminum cans and observed very different T2* times: cola ~ 1500 ms, tomato juice ~ 300 ms.
Figure 4 shows data recorded for our measurement of simultaneous MEG and NMR. The blue trace is the proton FID curve and the red trace is the evoked somatosensory response from a region of human cortex. The stimulus was delivered to the median nerve (thumb) at time t = 0.1 s, producing the artifact seen at that time, and the expected N20 response at 20 ms poststimulus and subsequent somatosensory components9 are clearly visible. These measurements, combined with the imaging demonstrated in Figure 3, demonstrate that it is possible to combine the advantages of low-field NMR/MRI with high-temporal-resolution MEG measurements. The possibility of imaging simultaneously with biomagnetic recordings could be useful for cardiac diagnostic testing and may alleviate some of the issues surrounding localization of MEG sources relative to anatomy.
Figure 5 shows the siumultaneously acquired NMR and MCG data. NMR signals were recorded at various times during the heartbeat, as indicated in the figure. Data for approximately 100 heartbeats were averaged after removing specific noise components (power-line harmonics, cryostat demagnetization signal, and eddy current signal), and filtering.