Download Biomagnetic sensing and more Lecture notes Anatomy in PDF only on Docsity!
Biomagnetic sensing
Hans-Joachim Krause^1 and Hui Dong 2, 3
(^1) Forschungszentrum Jülich, Institute of Bioelectronics, Peter Grünberg Institute (PGI-8), 52425 Jülich, Germany (corresponding author [email protected]) (^2) State Key Laboratory of Functional Materials for Informatics, Shanghai Institute of Microsystem and Information Technology (SIMIT), Chinese Academy of Sciences (CAS), Shanghai 200050, China (^3) CAS Center for ExcelleNce in Superconducting Electronics (CENSE) , Shanghai 200050, China
Abstract Bio magnetic sensing is a particularly valuable measurement technique because it is non-invasive in nature. Moving ions responsible for the electric activity of cells give rise to a magnetic field surrounding the current flow. Biomagnetic measurements denote the purely passive recording of this magnetic field outside the human body. The challenge is to record these extremely small magnetic fields in the presence of magnetic disturbance fields from the environment. Superconducting Quantum Interference Devices (SQUIDs), the most sensitive magnetic field sensors known to date, are used to measure the minute biomagnetic fields originating from the human heart or brain, in conjunction with attenuation of disturbances from the environment by passive shielding and/or active gradiometric suppression. Magnetic resonance imaging (MRI) is a well-established technique based on exposing the subject to a strong magnetic field, thus allowing to non-destructively measure the distribution of hydrogen atoms within the body. Low-field magnetic resonance im- aging (LF-MRI) is a novel measurement technique requiring more than thou- sandfold lower magnetic fields than conventional MRI, thus allowing to perform imaging with much simpler instrumentation in the presence of metals. Recent ex- periments yielded promising results with respect to distinction of healthy from ma- lignant tissue. Recently, combinatorial devices allowing to simultaneously record biomagnetic signals and perform magnetic resonance imaging of the anatomy of the human body source are developed to facilitate the determination of the biomagnetic sources by solving the three-dimensional magnetic inverse problem.
Keywords: SQUID, Biomagnetism, Magnetocardiography, Magnetoencephalog- raphy, Magnetic resonance imaging, Magnetic immunoassay
1 Introduction
Almost two hundred years ago, Oersted discovered that the flow of electrical cur- rents produces a magnetic field that encircles the current. Inside all animate beings and plants, ions move inside and outside of living cells, as well as from cell to cell. These moving charge carriers give rise to a magnetic field inside and in the vicinity of the creatures. Due to the fact that these natural currents in animals and humans are very small, the ensuing magnetic field is extremely weak. In almost all cases, this so-called biomagnetic field is much smaller than the magnetic field from other environmental sources, such as the earth’s magnetic field and the field from man- made sources such as power lines, electric appliances and moving steel objects. Therefore, it is indispensable to shield these environmental disturbance fields in order to be able to record minute biomagnetic fields. In many cases the fields are so small that they can hardly be measured even with the most sensitive magnetic field sensors known to date. The strongest contributions to the magnetic field intrinsically generated by hu- man beings are from the heart, the so-called magnetocardiogram (MCG), from mus- cles (magnetomyogram), and from the brain, the magnetoencephalogram (MEG). Biomagnetic measurements denote the contact-free registration of this magnetic field emitted from the body of the human or animal subject. This measurement is entirely non-invasive and purely passive. No excitation whatsoever is incident on the subject; just the magnetic field generated by the ongoing electrical action cur- rents are recorded at one or more positions outside the body. Due to the non-mag- netic properties of almost all tissue types, the magnetic field is practically unaffected by the intermediate tissue between source current and measurement location. In par- ticular, the varying electrical conductivities of the different types of tissue between source and sensor do not influence the magnetic field, whereas they do influence the voltages during electrocardiogram (ECG) recordings. Therefore, finding a solu- tion to the inverse problem for noninvasive 3D localization of intracardiac sources is much easier when using MCG data as compared to ECG recordings.
2 Superconducting Quantum Interference Devices
A SQUID consists of a superconducting loop with one or two Josephson junctions. It combines the effects of quantization of magnetic flux in units of the magnetic flux
quantum Φ 0 = 2.07×10 –15^ Vs and the dependence of the supercurrent circulating in
the loop on the magnetic flux threading the loop. The SQUID is an extremely sen- sitive converter of magnetic flux to output voltage. With a feedback electronics that compensates the measured flux with counter-flux to maintain a stable operating point, the device can be used as a linear null detector. SQUIDs are capable of re-
Φ/ A , with A denoting the area over which the flux is collected. To measure very
small magnetic fields B , the pickup area A needs to be of the order of mm^2 or even cm^2. Hence, SQUID magnetometers must include flux pickup structures having a sufficiently large effective area, A (^) eff. The low-inductance SQUID loop alone has a very small A (^) eff of the order of 10 –3^ mm^2 , which is not sufficient in most cases. For low SQUID noise, its inductance L (^) s should be as small as possible, but for good field sensitivity, a large A (^) eff yields a large L (^) s. This conflict can be resolved by col-
lecting the magnetic flux Φ over a larger area and coupling it more or less effec-
tively to the very small SQUID hole. The simplest solution is to make the outer dimensions of the SQUID loop much larger than the loop’s hole in order to concen- trate into that hole a fraction of the flux expelled from the superconductor due to the Meissner effect. This scheme is called washer SQUID. Much higher effective areas A (^) eff can be obtained using flux transformers. They consist of a pick-up coil, a pair of interconnecting leads, and a multiturn input coil inductively coupled to the SQUID, see Fig. 2 a. In order not to deteriorate performance by thermal Johnson noise, the whole flux transformer circuit is superconducting.
A pickup coil
coupling coil
superconducting shield
to SQUID electronics
SQUID
a b c
Fig. 2. a Flux transformer, consisting of a pickup coil and a coupling coil made from supercon- ducting wire, inductively coupled to a SQUID, b axial gradiometer, consisting of two counter- wound pickup coils above each other, c planar gradiometer, with counter-wound pickup coils next to each other.
In biomagnetic applications, very weak signals from localized sources such as heart or brain have to be measured against a background of magnetic disturbances which are orders of magnitude stronger, but more uniformly distributed in space because they stem from far sources, like power lines or cars. In such cases, the use of gradiometers is an alternative to magnetic shielding. For example, in the simplest case of a single first-order gradient component, two coils having identical areas A and a spacing b called baseline, are connected in series-opposite, as shown sche-
matically in Fig. 2 b for ∂ Bz / ∂ z (axial gradiometer) and Fig. 2 c for ∂ Bz / ∂ x (planar
gradiometer). Higher-order gradiometers that measure second and higher order spa- tial gradients of the magnetic field are made similarly using more pickup coils. Pla- nar gradiometers can be fabricated in thin-film technology. However, the axial ver- sion involves 3D superconducting wire structures, and thus is feasible only with low-temperature superconductor technology. High-T (^) c axial gradiometers require the usage of two or more SQUID sensors and electronic subtraction of the SQUID signals.
Low-T (^) c SQUID sensors are usually fabricated in thin-film niobium technology using magnetron sputtering on oxidized silicon or quartz wafers. The SQUID struc- ture and the junctions are photolithograhically patterned from Nb/AlO (^) x/Nb trilayers, where the aluminium oxide barrier is formed by thermally oxidizing a nanometer- thin aluminum layer. Additional metallization is used for bonding pads and for fab- rication of junction shunts. The insulation between Nb and conducting layers is usu- ally obtained by depositing silicon oxide films and by edge anodization to form an insulating NbO (^) x oxide. SQUID and input coil of the flux transformer are usually integrated into one monolithic thin-film structure. For biomagnetic applications, the complete magnetometers and gradiometers are often made with 3D Nb-Ti wire- wound pickup coils bonded to Nb pads of the input coil. High-T (^) c SQUID sensors are typically fabricated from YBa (^) 2Cu (^) 3O (^) 7-x (YBCO) ep- itaxial thin films deposited on single crystal SrTiO3, MgO or LaAlO 3 substrates with surface polished to epitaxial quality. The most popular Josephson junction type is the grain-boundary junction. The grain boundary can for instance be formed by de- positing YBCO on a bicrystal substrate assembled from two single crystals which are glued together with twisted crystal orientations. Another approach is to etch a ditch into the substrate and to grow the YBCO film across this so-called step edge. The YBCO SQUID structures are patterned by optical photolithography of films followed by argon ion beam or wet-chemical etching.
3 Biomagnetism
The term “biomagnetism” denotes the measurement of the natural magnetic field generated by a living creature due to the movement of electric charge carriers inside the body [ 2 ]. Both intra- and extracellular currents, as well as the exchange of ions between cells contribute to the electrical currents inside the body and thus to the magnetic field surrounding the currents [ 3 ]. MCG and MEG are the most common biomagnetic modalities. Other electrically active organs are also known to produce detectable magnetic fields. The field of the eye is called magnetooculogram, the field of the stomach is the magnetogastrogram.
3.1 Magnetocardiography
The heart of mammals, in particular the human heart, consists of heart muscle cells, so-called myocardial fibers, arranged as series of cells connected with intercalated discs. These discs act as connectors that allow the electrical excitation to propagate successively from cell to cell [ 4 ]. In order to pump blood through the body, all the muscle cells in the atria and in the ventricles have to be excited quasi simultane- ously. This is performed by the natural pacemaker, the so-called sinus node, located close to the entry of the vena cava superior into the right atrium. In ECG as well as
four-channel system to adult magnetocardiography and to fetal magnetocardiog- raphy has been shown in [ 20 ]. The system has been utilized for educational pur- poses in Jülich, measuring the MCG of more than 3000 high-school students.
a b
Fig. 3. a Principle of MCG measurement; b Typical human MCG measured with high-T (^) c rf SQUID.
Numerous clinical trials have been performed regarding the application of MCG to different heart problems. Applications include the detection of myocardial ische- mia [ 21 ] and viability [ 22 ], coronary artery disease [ 23 , 24 , 25 ], arrhythmogenic risk assessment, imaging of arrhythmogenic sites such as the Wolff-Parkinson-White syndrome [ 26 ] or ventricular arrhythmia [ 27 ].
3.2 Fetal Magnetocardiography
The MCG of an unborn baby in the mother’s womb was measured for the first time in 1974 with an unshielded SQUID system [ 28 ]. Fetal magnetocardiography (fMCG) is a reliable method for noninvasive study of fetal cardiac electrophysiol- ogy from the 20 th^ week of gestation on, especially during the third trimester of preg- nancy when the electrically insulating vernix caseosa hampers abdominal recording of fetal ECG [ 29 ]. In addition, maternal ECG obscures fetal ECG. Systematic stud- ies on the analysis of cardiac time intervals have been performed in [ 30 ]. Other common fetal monitoring techniques such as cardiotocography and echocardiog- raphy lack the temporal resolution to extract such information. fMCG has been shown well applicable to the early diagnosis of arrhythmia [ 31 , 32 ]. For the early detection of congenital heart defects, however, only a few fMCG case reports have been reported [ 29 ]. Here, echocardiography appears to be the method of choice. Albeit the sensitivity is not as high as in case of usual helium-cooled SQUIDs, it is possible to record fMCG with high-T (^) c SQUIDs [ 33 ], see Fig. 4. In order to study the electrophysiology, averaging and subtraction of the maternal MCG is needed, in contrast to low-Tc instrumentation that yields sufficient signal-to-noise even in real time. Once that instrumentation becomes more robust and can operate outside mag- netic shielding, it is expected that fMCG finds widespread acceptance because it is the only modality that allows to study the electrophysiology of the fetal heart.
Fig. 4. Typical real-time fetal MCG signal, measured with high-Tc rf SQUID, and fetal signal averaged over 75 s.
3.3 Magnetoencephalography
Measuring the magnetic field of neuronal brain activity with MEG is more chal- lenging than MCG because the magnetic field of the human brain is about hundred- fold smaller than that of the heart [ 34 ]. The electrical ion currents flowing in the human head tissue give rise to a magnetic field that can be observed outside the skull. The currents include both intracellular “impressed” currents by neural activ- ity, and the so-called “volume current” of freely moving ions in the extracellular space. The latter can be modeled as a conductive medium with an electrical con- ductivity depending on the type of tissue, i.e. white matter, gray matter etc. The total magnetic field is determined by a summation over all current elements in the whole head according to Biot-Savart’s law. To a good approximation, the magnetic field generated by the impressed currents is orthogonal to the scalp surface, whereas the contribution from the volume currents is tangential to it [ 35 ]. Thus, the contribution of volume currents can be neglected in the most common measurement configura- tion that registers just the magnetic field component orthogonal to the scalp. In the case of pyramidal neurons with an open-field structure [ 2 ], predominantly consist- ing of a single dendrite and one long axon, the small current flowing in the dendrite due to membrane depolarization after neurotransmitter intake at the synapse can be modeled as a current dipole with a field B ∝ 1/ r^3 , where r denotes the distance. This current dipole contributes to the magnetic field. A typical current dipole of one post- synaptic dendrite has a magnetic moment of 2×10 –15^ Am, which gives rise to a mag- netic field contribution of 3×10 –19^ T at a distance of 5 cm [ 35 ]. The summation of current dipoles from the pyramidal neurons is orthogonal to the scalp surface and results in a detectable biomagnetic signal. Considering a magnetic field resolution of typical MEG instrumentation of a few fT/√Hz, it becomes obvious that approxi- mately 50,000 synchronously firing neurons are needed to obtain a measureable signal. The so-called “action potential” (AP) contribution to the brain’s magnetic field is obtained from the depolarization current flowing along an axon, followed by a repolarization current restoring the rest state potential. Thus, the AP consists of a pair of current dipoles with opposing directions, a current quadrupole, which yield
olfactory stimuli, to form recognizable objects. With support from MEG studies, the assumption was made that synchronous oscillations in the so-called γ band at around 40 Hz play a role in the process. A review on this so-called binding process is given in [ 42 ]. Nonoscillatory magnetic brain responses have also been studied with respect to feature binding and object recognition [ 43 ]. Another important field of brain research involves the study of motor actions. Preparation, control, and execution of movements involve transient, slow, and os- cillatory magnetic activity in both primary sensory-motor and higher cognitive ar- eas. The role of the so-called magnetic μ-rhythms for the preparation and execution of movements was investigated by means of event-related spectral power changes [ 44 ]. MCG is clinically used for evaluating normal and abnormal brain functions, and for the localization of cortical sources. Currently, the localization of epileptic dis- charges and presurgical brain mapping represent the most common clinical appli- cations [ 45 , 46 ]. For presurgical evaluation in patients with intractable focal epi- lepsy, MEG is medically necessary to localize areas of epileptic activity. There is a need to measure small epileptic discharges with bigger background brain activity. Other diagnostic applications remain in the research stage. Normal spontaneous brain activity of EEG and MEG consists of various fre- quency bands. Alpha waves (8–13 Hz) are dominant when the subject is awake, beta waves (>13 Hz) are seen during wakefulness and light sleep. Theta (4–7 Hz) and delta (<4 Hz) rhythms are usually observed during sleep, but they may also appear due to brain tumors and ischemia. The source of the spontaneous activity, both normal and abnormal, spreads over the bilateral cerebral cortices, so that sep- aration of each generator is hardly possible with EEG. The higher spatial resolution of MEG may help to localize abnormally slow waves [ 47 , 48 ] due to structural brain lesions. Ischemia of the brain can also be detected with MEG. Stenotic lesions of the internal carotid artery system sometimes yields an oscillatory signal at 6–8 Hz in the temporo-parietal area [ 49 ]. It has been successfully shown that MEG is even feasible with high-T c SQUID instrumentation. By comparison with the result obtained from a commercial 248 channel whole-head MEG system, it was demonstrated that the sources of auditory evoked responses can be localized with similar precision using a single high-T (^) c dc SQUID magnetometer operating at 77 K [ 50 ]. However, a long way of development is still needed until nitrogen-cooled SQUID systems will have matured to be suita- ble for routine MEG recordings. Today, there are more than 130 MEG systems installed worldwide, which is a relatively small number as compared to the approximately 36,000 MRI machines. The application of MEG is still mainly focused on research, i.e. all the different aspects of brain activity listed above. Routine clinical applications are still scarce. In order to fully establish MEG in clinical practice, the reliability of source estima- tion needs improvement. Inversion software should be improved, in particular for multiple source estimation to overcome the nonuniqueness of the electromagnetic
inverse problem. For clinical research, a validation of the source estimation accu- racy of MEG by fusion with other imaging modalities is needed. MEG has excellent time resolution but is not perfect with respect to localization accuracy. Therefore, a combination with low field MRI as an anatomical imaging modality becomes par- ticularly promising for future work (see section 5 on hybrid biomagnetism).
4 Magnetic resonance imaging
Magnetic resonance imaging (MRI) is a powerful and probably the most versatile medical imaging modality for the human body [ 51 , 52 ]. This technique is developed from nuclear magnetic resonance (NMR) [ 53 ]. The MR signal originates from the nuclei with nonzero spin, and protons are conventionally most-commonly imaged because of widely distribution in human body fluids, proteins, lipids, glucose, etc. In a static magnetic field B 0 along z direction, an energy difference (^) ∆ E = γ B 0 oc-
curs dependent on if the spin is aligned parallel or antiparallel to the field, with γ denoting the gyromagnetic ratio (in case of proton: γ/2π = 42.58 MHz/T) and (^) being Planck’s constant h divided by 2π. The populations of parallel and antiparallel spins are nearly equal. For example, at B 0 = 1 T, the ratio between the two types of spins is approximately 1.000007. The net equilibrium magnetization of the spin sys- tem can be expressed as 𝑀𝑀 0 = 𝜌𝜌𝛾𝛾 2 ℏ^2 𝐵𝐵 0 ⁄(4^ 𝑘𝑘𝐵𝐵 𝑇𝑇), in which ρ is the spin density and T the temperature. The imaging of nuclei is realized by applying linear three-dimensional magnetic
field gradients 𝑮𝑮 ≡ 𝜕𝜕𝐵𝐵𝑧𝑧 𝜕𝜕𝜕𝜕 𝒙𝒙�^ +^
𝜕𝜕𝐵𝐵𝑧𝑧 𝜕𝜕𝜕𝜕 𝒚𝒚�^ +^
𝜕𝜕𝐵𝐵𝑧𝑧 𝜕𝜕𝜕𝜕 𝒛𝒛�. Each component is generated by special coils and can be controlled separately. The gradients G will encode the precession frequency (or phase) by the spatial position of the proton. By applying specific pulse sequences and a spatial Fourier transform, three-dimensional images can be ac- quired. Certain nuclei (commonly hydrogen) can absorb and emit rf energy. The appli- cation of an rf pulse causes the protons to precess about B 0 at their Larmor frequency f L = ( γ /2π) B 0. During the precession, M 0 undergoes two relaxation processes. The longitudinal relaxation, characterized by the relaxation time T 1, reflects the interac- tion between spin and nearby lattice with energy exchange which causes the longi- tudinal magnetization back to equilibrium. The transverse relaxation, characterized by the relaxation time T 2, describes the dephasing process of net magnetization in the transverse plane. There is no energy lost in the transverse relaxation process, and T 2 ≤ T 1. In tissues, T 2 and T 1 are both field-strength dependent, and they may range from tens of milliseconds to about one second. The T 1 (or T 2) weighted MRI images may show the diseased tissues like tumors by different image contrasts be- cause these malignant tissues usually exhibit different relaxation times from the normal tissues.
strength is reduced by four orders of magnitude, the detection sensitivity of the Far- aday coil can no longer meet the requirements of the imaging SNR. In order to increase SNR, there are two common methods. First, SQUIDs are used as detectors instead of Faraday coils. The SQUID with a sensitivity of up to 10 –15^ T/√Hz is one of the most sensitive magnetic field sensors, and its sensitivity is independent of frequency [ 69 ]. Alternatively, the introduction of pre-polarization will increase the initial macroscopic magnetization of the sample and finally improve SNR. In this technique, a strong magnetic field pulse (B p) is applied to pre-polarize the sample, and then the MRI signal is detected in the B 0 field. In addition, the entire system can be housed in a magnetically or conductively shielded room. The external magnetic field noise in the signal frequency bandwidth will be attenuated and the SNR can be improved. However, the rapid switch-off of B p—typically in 10 ms—to avoid significant decay of the magnetization before sig- nal acquisition induces transient eddy currents in nearby conducting objects, most notably the walls of the shielded room, see Fig. 5. The resultant inhomogeneous magnetic-field transient may both seriously distort the spin dynamics of the sample and exceed the dynamic range of the SQUID readout electronics, and must be greatly reduced before one can begin image encoding and acquisition.
Fig. 5. Aluminium shielded room containing the liquid-helium dewar, the B p coil and the DynaCan coil.
We developed the so-called dynamical cancellation (DynaCan) technique to sup- press adverse, pulse-induced transient eddy currents [ 70 ]. DynaCan exploits the fact that eddy currents are typically a superposition of modes that decay exponentially with their individual time constants. Different time constants correspond to different spatial eddy-current patterns. An additionally pulsed magnetic-field waveform with features at corresponding time scales thus allows selective coupling to the dynamics of the individual patterns ( Fig. 6 ). This cancellation pulse is provided by a current fed into a separate coil, spatially larger than the B p coil, during the later part and turn-off of the polarizing pulse. The computationally determined DynaCan current waveform is designed to drive the eddy currents to zero at the precise moment that the polarizing field becomes zero. With DynaCan we demonstrated a reduction of the eddy-current fields by 99% ( Fig. 7 ).
Fig. 6. Dynamical cancellation waveform I c and end of pre-polarizing pulse I p; the currents are normalized to the amplitude of I p. Arrows indicate logic switching instants for (A) opening relays in the B p and DynaCan coil circuits and (B) beginning data acquisition.
Fig. 7. Magnetic-field transients at the center of the room without (black dashed line) and with DynaCan (green solid line, also in insets) for three different B p currents I p: (a) 6.2 A, (b) 8.3 A, and (c) 9.8 A. Oscillations in the inset to (a) are residual 60 Hz interference.
(2) The enhanced intrinsic longitudinal relaxation times ( T 1) of different soft tis- sues at low field can be used to distinguish cancerous and normal tissues, e.g. pros- tate cancer, breast cancer etc., which have poor specificities in high-field MRI [ 78 , 79 , 80 ]. (3) High-field MRI images cannot be acquired if the patient has a pacemaker, screws or other metallic implants in the body, because the susceptibility difference between metal and tissue may lead to a local field inhomogeneity proportional to B 0 strength and gives rise to image distortion. By lowering the B 0 field to microtesla range, this effect becomes negligible [ 81 ]. In addition, these objects are exposed to strong magnetic forces that might hurt the patient. (4) Hybrid imaging by combing MEG with LF-MRI may provide functional and anatomical information of brain simultaneously, see the following section 5. (5) The relaxation dispersion behavior reflects the B 0 field dependence of the relaxation times (typically T 1). Because of technical constraints in commercial ma- chines, the frequency range of traditional relaxation dispersion curves obtained by fast field cycling technique usually starts from 10 kHz (234 μT) to several MHz [ 82 ]. Therefore, people developed the spin-locking technique to measure the T 1 ρ relaxation dispersion in the rotating frame at low spin-lock field to gain useful in- formation on the composition of macromolecules, like proton exchange between water and macromolecules [ 83 ]. However, the heat produced by the spin-lock pulse usually makes the T 1 ρ technique prohibitive for human study. The LF-MRI tech- nique enables direct T 1 and T 2 dispersion measurement at all frequencies below 10 kHz without heating problem. Inglis et al. showed in vivo human brain images at LF with distinguishable components, like brain tissue, scalp, blood and cerebrospi- nal fluid [ 84 ]. In order to determine whether LF-MRI has further potential ad- vantages for in vivo human brain imaging, a quantitative comparison was made be- tween relaxation dispersion in postmortem pig brain measured at ultra-low fields and spin-locking at 7 Tesla [ 85 ]. It was found that LF-MRI may offer distinct, quan- titative advantages for human brain imaging, while simultaneously avoiding the se- vere heating limitation imposed on high-field spin-locking.
5 Hybrid biomagnetism and magnetic resonance imaging
The temporal resolution of MEG, typically 1 ms, is much better than fMRI, but a drawback of magnetic neuroimaging is the fact that the three-dimensional inverse problem is ill-posed. Helmholtz showed more than 150 years ago that it is impossi- ble to uniquely determine the current distribution inside a conductor from a meas- urement of the magnetic field in its surroundings. Therefore, MEG is not perfect with respect to localization accuracy of the source. If, however, a priori information on the shape and on the conductivity of the conductor is available, the ill-posedness of the problem can be overcome. If MRI and MEG measurements are performed in different systems, data migration from one system to the other is a big challenge. In
addition, geometrical positioning errors cannot be avoided. Typically, MEG/MRI co-registration errors may reach the order of 5–10 mm in the cortex. The spatial resolution is less precise for sources in the deep brain region. Since MRI and MEG signals can be distinctly separated in the frequency domain, co-registration of both modalities is feasible. Because of simultaneous measurement, positioning errors are completely avoided. Furthermore, the combination of MEG with LF-MRI no longer requires moving the patient and reduces the total system cost. Several groups com- bined LF-MRI with MEG. The Los Alamos group first demonstrated the possibility of simultaneous meas- urement of MEG and MRI using their homemade seven-channel low-Tc SQUID system [ 86 ]. It allows three-dimensional matching of LF-MRI images and MEG data with better accuracy than that of traditional subsequent MEG and MRI regis- tration. They also suggested that parallel imaging of LF-MRI with hundreds of SQUID channels would significantly reduce the system noise. The total imaging time would be accelerated, which finally would make the hybrid MEG/MRI system more reliable and efficient for clinical diagnosis [ 87 ]. Subsequently, the Finnish group developed a combined MEG/MRI using a commercial whole-head 306-chan- nel MEG machine [ 88 ]. Great efforts are made to optimize the sensors, the pulse sequences and the reconstruction methods.
6 Magnetic resonance imaging of neural activity
The magnetic field generated by neuronal activity adds to the magnetic field im- posed on the human body for MR imaging. In case of LF-MRI, this local distortion of the imaging field on the order of hundreds of picotesla may affect spin dynamics because it slightly changes the imaging field in the microtesla range. This local field change may lead to a detectable change in the NMR signal. This modality is called direct neuronal imaging (DNI) [ 89 ] or neuronal current imaging (NCI) [ 90 ]. Sus- tained neuronal activity characterized by a local quasi-static magnetic field change can be observed as a change of the local spin-precession frequency. Fast neuronal activity may act as a tipping pulse, the so-called AC or resonant effect in NCI [ 91 ]. Using a priori information on anatomical structure and on the electrical conduc- tivity of the different tissues in the brain may considerably improve the accuracy of source localization. MRI allows to measure the electric current density in an object by observing how the associated magnetic field affects the spin precession. This so- called current-density imaging (CDI) modality has been shown to be feasible with externally impressed current [ 92 ]. It has been shown that conductivity imaging can be done in a standard MRI system without applying a current just by post-processing analysis of the phase distribution of the imaging rf pulse [ 93 ]. Evaluation of the magnitude allows to perform permittivity imaging [ 94 ]. However, conductivity im- aging is usually done in standard high-field MRI systems, leading to a measurement
can be determined from the measured complex magnetic susceptibility [ 103 ]. Therefore, the size enhancement due to biocompatible surface coating and subse- quent functionalization and analyte binding can be measured [ 104 ]. A key disad- vantage of the susceptometry technique is its lack of selectivity. In case of low con- centrations of biomolecules and consequently low concentrations of magnetic marker particles, the resultant susceptibility of the solution is hard to discern from the parasitic susceptibility of the sample container, of the reagents and of the labor- atory environment. (2) Relaxometry [ 105 ] is based on recording the time transient of the magnetic response of the particles during the off-time of a pulsed excitation field. By analyz- ing the relaxation time of the particle’s magnetization, a distinction between the Néel relaxation of bound particles and the Brownian relaxation of unbound carriers is feasible [ 106 ], since the reorientation of the magnetization vector inside the mag- netic core is significantly slower than the Brownian relaxation of particles in solu- tion. It is possible to obtain information on the size distribution of the magnetic cores of nanoparticles, especially on the mean value and the standard deviation of the core diameter of the magnetic crystallites. The technique allows to monitor bind- ing kinetics [ 107 ]. Since the relaxometric magnetic field signals are typically very small, the technique usually requires the use of ultra-sensitive SQUIDs as magnetic field sensors. On samples with higher particle concentration, relaxometry can also be measured with fluxgate sensors [ 108 ]. (3) The frequency mixing technique [ 109 ] probes the nonlinear magnetization curve of superparamagnets. Upon magnetic excitation at two distinct frequencies f 1 and f 2 incident on the sample, the response signal generated at a frequency repre- senting a linear combination m. f 1 + n. f 2 is detected. The appearance of these compo- nents is highly specific to the nonlinearity of the magnetization curve of the parti- cles. With this magnetic measurement technique, a magnetic immunoassay for detection of tetanus toxoid was developed. Coaxial coils provided magnetic excita- tion fields at two distinct frequencies f 1 = 49.38 kHz and f 2 = 61 Hz incident on the sample. By means of a differential pickup coil, the response signal of the sample inside the coil at a frequency f 1 + 2. f 2 was detected. This mixing component was chosen since it is maximum for vanishing static offset field. Prior to the measure- ment, primary antibodies were immobilized on a polyethylene filter (Abicap from Senova, Weimar). Then, 500 μl sample was added. When the sample passed the filter, 500 μl of secondary antibody solution (anti-h-IgG biotinylated in PBS) and 500 μl magnetic bead solution (fluidMAG-Streptavidin 200 nm from chemicell, Berlin) were added and rinsed with 750 μl PBS. Fig. 9 shows the measured signals of different tetanus immunoassay samples as a function of the concentration of the analyte. At low concentrations of the analyte, unspecific binding of MNP and the thermal noise of the detection coil determines the detection limit. At high concen- trations, saturation occurs because nearly all available binding sites in the filter are occupied. Numerous magnetic immunoassays have been demonstrated [ 110 , 111 , 112 ] which usually yielded a better sensitivity than conventional immu- noassays.
0.01 1 10 100 1000
1
10
Signal [V]
Concentration [ng/ml]
Fig. 9. Calibration curve of a magnetic tetanus immunoassay.
8 Conclusion and perspectives
Biomagnetic measurements of the magnetic field surrounding the human body, gen- erated by the electrophysiological processes of life in the body, have developed continuously over the past decades. The evolution of supersensitive SQUID tech- nology has led to the development of multichannel systems with hundreds of sensor channels for recording MCG and MEG. They provide unique images of heart and brain activity on a millisecond time scale. Increasing computational power and the fusion of data obtained with different imaging modalities has led to unique solutions of the inverse problem of electromagnetism and thus opened up a new window into the human body. Numerous clinical diagnostic applications of MCG, MEG and LF- MRI have been already demonstrated. LF-MRI allows more broadly applicable, less dangerous and eventually cheaper instrumentation than its well-established high field counterpart. In addition, promising first results with respect to distinction be- tween malignant and healthy tissue have been obtained. The fusion of these tech- niques, in conjunction with the novel imaging modalities NCI and CDI, is expected to lead to added diagnostic value as compared to the sum of the individual tech- niques. Major obstacles are the need for heavy magnetic shielding and for liquid helium coolant, both of which account for a major fraction of the cost of procure- ment and operation. Recently, flexible shielding solutions adapted to the specific requirements of the location are being offered at affordable cost. As high-T (^) c SQUIDs are becoming increasingly sensitive and reliable, they might establish as an alternative in MCG instrumentation where sensitivity is not as critical as for MEG. Due to its noninvasive and almost passive nature, biomagnetic sensing is expected to become increasingly important in the near future, both in scientific re- search and in clinical diagnostics.
