WO2008035068A1 - Electron spin resonance apparatus and method - Google Patents

Abstract

An electron spin resonance (ESR) detecting device having a null-detecting receiver is disclosed. An excitation signal is input to a resonator which holds a sample for ESR detection. A response signal from the resonator is coupled to a balance signal to form a detection signal received by a detector. The detector also receives a reference signal having a phase relationship with the excitation signal, whereby the detector's output is indicative of a difference between the detection signal and the reference signal. The detector's output is used in a feedback loop to generate an adjustment signal which causes the balance signal to substantially cancel the response signal. The adjustment signal can be used to measure resonance, while the input to the detector is minimised to reduce the effect of noise on measurements.

Description

ELECTRON SPIN RESONANCE APPARATUS AND METHOD

TECHNICAL FIELD

This invention relates to electron spin resonance (ESR) detection techniques, e.g. where a sample in a magnetic field is subjected to a microwave excitation signal having a frequency selected to cause paramagnetic species in the sample to resonate. In particular, the invention relates to the detection of spatially resolved ESR signals for the purpose of generating images of a sample .

BACKGROUND TO THE INVENTION

A magnetically "unpaired" (paramagnetic) electron has a magnetic moment that aligns itself parallel or antiparallel to an externally applied magnet field. The electron can absorb electromagnetic radiation having a suitable energy to resonate between the parallel and antiparallel states. For typical (laboratory) external magnetic fields, the electromagnetic radiation that achieves resonance is in the high radio frequency or low microwave region, e.g. having frequencies of 100MHz to IGHz or more.

ESR is a spectroscopic technique which detects molecular species that have an unpaired electron. In conventional devices, a sample to be measured is placed in a microwave resonator, where an external magnetic field is applied. Microwave power, e.g. an amplified output from a Gunn oscillator, is supplied as an input to the microwave resonator, e.g. such that the microwave magnetic field is perpendicular to the external (static) magnetic field at the sample, to excite resonance in the sample. An output signal, e.g. reflected from the resonator, is separated, e.g. using a circulator, from the input signal and monitored to detect resonance. The external magnetic field is swept and when the resonance condition is achieved, it can be detected in the output signal. This can be done by comparing the output signal to a reference signal derived from the microwave power input e.g. to perform homodyne detection. Thus, ESR is the resonant absorption of microwave or radio-frequency radiation by paramagnetic electrons polarised by a magnetic field. It offers both the sensitivity and, crucially, the selectivity necessary to detect paramagnetic molecular species in a living organism or other type of sample.

ESR measurements can be used to obtain images . In imaging experiments, spatially resolved ESR measurements can be obtained by the application of an inhomogeneous magnetic field to the sample (e.g. animal) . This is the same method used in the analogous (nuclear) magnetic resonance imaging (MRI) technique widely used in clinical and other contexts. ESR imaging is used primarily to provide functional images of chemical processes.

Molecular species with magnetically "unpaired" (paramagnetic) electrons occur naturally in many important physiological, patho-physiological, and pharmacological processes (oxygen, reactive oxygen species, nitric oxide and the cofactors of many enzymes, for example) . A variety of inert non-toxic "labelling" and "probe" technologies are available that use unpaired electrons . One aim of the present invention is to improve the sensitivity, speed and spatial resolution with which these paramagnetic molecular species, both endogenous and exogenous, can be imaged in a living organism. Applications in physical science and technology are also envisaged, e.g. the imaging of molecular scale defects in materials.

The potential biomedical applications of ESR imaging have already attracted widespread attention. There are several large and active groups in this area, for example a National Institute of Health centre based at the University of Chicago. The dominant ESR equipment manufacturer (Bruker, Germany) has developed a commercial instrument. The application of ESR imaging to a biomedical research has been researched in the UK by a team using the Bruker instrument. Applications in the physical sciences, such as imaging defects in artificially grown diamonds, have also been demonstrated. The microwave circuit used in current instruments is an incremental development of one originally designed for ESR spectroscopy in the 1960s. For biomedical work, the key modifications have been the use of low microwave frequencies to minimise microwave absorption in biological tissue, and the development of efficient sample/animal containing resonators for use at these frequencies. Signal recovery strategies (field modulation with lock-in detection) have also not changed since the 1950s. Images are obtained by measuring spectra in the presence of a series of magnetic field gradients, and subsequent mathematical transformation. This approach is very similar to that used in early (nuclear) MRI instruments. The main difference is the nuclear experiment is measured by pulsed/Fourier spectroscopy, whereas the electronic spectrum is almost always performed by continuous- wave/frequency-selective methods. There are very few chemical species that are amenable to pulsed ESR techniques .

However, the conventional ESR apparatus has limited measurement sensitivity. This is because the noise factor of the system that receives output signals from the resonant cavity is high. US 5389878 discloses an ESR device that aims to improve the noise factor of the receiving system by cancelling out reflections of the excitation signal at the receiver e.g. by coupling a reference signal from the excitation (input) signal, inverting its phase and adding it to the reflective (output) signal from the resonant cavity. The aim is to cancel the reflected excitation signal so that only a signal corresponding to resonance in the cavity is amplified in the detector.

The noise power of a conventional ESR imaging instrument is invariably orders of magnitude greater than the fundamental limit set by the thermal emission of the sample. The ultimate technological challenge is to reduce the instrumental noise to these fundamental levels. However, one problem is that in any practicable instrument of conventional design, microwave excitation power cannot be completely excluded from entering the signal receiver/detector. One consequence of this is well known to the ESR community: oscillator/source noise contributes to the total instrumental noise. It is this effect that US 5389878 aims to address. However, these techniques still do not provide an improvement in instrumental performance that is desirable to fulfil the potential of ESR imaging in biomedical applications.

The present invention aims to improve further the sensitivity of ESR devices and thereby provide an ESR imaging apparatus suitable for biomedical applications.

SUMMARY OF THE INVENTION

The present inventors have noticed that there is another reason for noise in the receiving systems of conventional ESR devices, which reason is largely- unrecognised in the field. The inventors have recognised that the receiver noise increases due to low frequency fluctuations in the receiver transfer function, a phenomenon known as "transposed flicker noise". Because receiver transposed flicker noise is microwave power dependent it has frequently been incorrectly assigned to an oscillator noise contribution in the ESR literature. Microwave leakage into the receiver is especially problematic in biomedical imaging. Breathing, and other animal movements, makes it impossible to maintain the tuning and matching of the microwave circuit. Conventional instruments have little or no defence against this. Imaging experiments also involve relatively high excitation levels . Although modern low noise receivers employing microwave pre-amplifiers have excellent, near thermal, white noise performance they suffer from severe transposed flicker noise. Even a few μW leaking into the receiver can dominate the noise at frequencies relevant to ESR signal modulation schemes. This is rarely mentioned in data sheets, or in microwave engineering texts, because such receivers are typically employed in (and designed for) commercial applications in which very little power enters the receiver. Superior transposed flicker characteristics are the principal reason why "old-fashioned" diode detectors/mixers are still used in ESR instruments despite their inferior white noise temperatures, susceptibility to local oscillator noise and inferior (non-transposed) flicker noise characteristics.

The present invention proposes a radical change to conventional spectroscopic (and more specifically ESR) detection methods, which aims to ameliorate the problems associated with noise at the receiving system (both oscillator noise and transposed flicker noise) in order to improve the sensitivity of the device. The invention is discussed below as applied to ESR detection, but it is equally applicable to other spectroscopic techniques that require the measurement of microwave absorption or dispersion, e.g. studies of molecular rotational spectra, and dielectric spectroscopy.

At its most general, the invention proposes using a null-detecting receiver in a device for detecting the resonant absorption or dispersion of radiation. This addresses the dominant noise sources (enhanced oscillator and receiver transposed flicker noise due to excitation power leaking into the receiver) in a fundamentally different way from conventional ESR devices. It does this by substantially minimising the input to the receiving system (detector) of the device and by using a feedback arrangement to maintain this "null condition". In contrast with conventional devices, the feedback signal can be used to measure resonance. Thus, according to one aspect of the invention there may be provided a device for detecting resonant absorption or dispersion of radiation (e.g. an electron spin resonance detecting device) , the device including: a resonator (e.g. resonant cavity) for receiving a sample that is subjected to a controllable magnetic field; a power source (e.g. microwave power source) arranged to generate an excitation signal which is input to the resonator, whereby the resonator produces a response signal (e.g. which reflects the effects of resonant absorption or dispersion of the excitation signal by the sample) ; a balance signal generator arranged to produce a balance signal which is coupled to the response signal to form a detection signal; a reference signal generator arranged to produce a reference signal having a stable (e.g. fixed) phase relationship with the excitation signal; and a detector arranged to receive the detection signal and the reference signal and to generate an output based on the detection signal and reference signal (e.g. a multiplicative mixing thereof) , the output being indicative of a difference between the detection signal and the reference signal; wherein the balance signal generator includes a signal adjuster connected to receive the output from the detector and arranged to communicate an adjustment signal to the balance signal generator, and wherein the adjustment signal is generated based on the output from the detector to form a feedback loop in which the balance signal substantially cancels the response signal. The size of the adjustment signal may be indicative of the ESR absorption or dispersion and is therefore preferably monitored to provide an ESR measurement.

The ESR measurement can be used in a wide variety of experiments. The controllable magnetic field may have up to four independently controllable magnetic field variables, e.g. three orthogonal field gradients and a homogeneous field applied to the sample. In a first type of experiment, the ESR measurement may be used to obtain information relating to the ESR spectrum itself (which may be unknown) of a sample. Such experiments may be termed spectro-spatial, as they may investigate the spatial dependence of the ESR spectrum in one, two or three dimensions. Alternatively or additionally, the measurement may be used in a second type of experiment in which the ESR spectrum of a particular species is typically known beforehand. Here a set of spatially resolved ESR measurements may be used to identify a spatial concentration of known species. In this way, a image in two or three dimensions of that species in a sample may be generated.

In the null-detecting receiver arrangement of the device above, the power entering the detector can be very small e.g. of comparable magnitude to thermal noise levels, which means that both transposed flicker noise and oscillator noise are suppressed to a level where their effects are negligible, even at high sample excitation levels .

Preferably, the device comprises a microwave interferometer in which the balance signal and response signals are produced in respective arms of the interferometer. In this arrangement the sample containing resonator is in the response signal's interferometer arm. When the microwave absorption of the sample changes, the interferometer becomes unbalanced, which will cause a nonzero detection signal to be detected at the detector. The signal adjuster will therefore generate an appropriate signal to adjust the balance signal and bring the interferometer back to a null condition. The ESR measurement may be an ESR absorption signal or an ESR dispersion signal. An ESR absorption signal can be obtained from the size of the adjustment in the compensating loss required to maintain the null condition. Similarly, an ESR dispersion signal can be obtained from the size of the compensating phase shift required to maintain the null condition.

Preferably, the resonator is a conventional reflection geometry resonator, which can maximise the sensitivity of ESR measurements. The excitation signal may have any frequencies suitable for a particular sample, e.g. excitation frequencies in the range 250 MHz to 250 GHz may be used. In imaging experiments 250 MHz to 10 GHz are typical. Lower excitation frequencies (e.g. in the range 250 MHz - 1 GHz) are preferred for large samples with significant microwave losses (e.g. small animals), whereas instruments operating in at higher frequencies (e.g. 2-10 GHz) may be optimised to obtain improved resolution and sensitivity with lower loss samples. For most spectroscopic applications, much higher frequencies e.g. up to 100 GHz are preferred.

Preferably, the reference signal generator includes a forward power coupler arranged to couple the excitation signal. In other words, the excitation signal and reference signal have a common source, which can minimise errors.

The balance signal generator may be arranged in any conventional way to provide an interferometic arm opposed to the response signal. For example, the balance signal generator may be provided on an independent arm to form a Michelson interferometer. A preferred arrangement provides a Mach-Zehnder interferometer. Here the balance signal generator may include a forward power coupler arranged to couple a signal from the excitation signal produced by the microwave power source. This coupled signal is preferably provided to a signal attenuator to adjust it to the correct magnitude before it is coupled to the response signal to create the (null) detection signal. Preferably, the attenuation of signal attenuator is variable on the basis of the adjustment signal in order to effect dynamic maintenance of the null condition.

The response signal from the resonator typically comprises a reflected part of the excitation signal and a resonance signal corresponding to resonance (if any) in the sample. By dynamically adjusting the balance signal to cancel both the reflected excitation signal and the resonance, changes in resonance can be detected without power leaking into the detector degrading its noise level. Preferably, the detector comprises an amplifier (e.g. a FET low noise amplifier) arranged to amplify the detection signal before it is mixed with the reference signal. The mixing may be achieved in any conventional manner to provide an in phase (0° phase difference) output and a quadrature (90° phase difference) output. Either of or both these outputs may be used by the signal adjuster to generate the adjustment signal that is monitored as a measurement of resonance. In the Mach-Zehnder arrangement described above, the in phase output is preferably used by the signal adjuster to generate an adjustment signal for the variable attenuator. Preferably, both the in phase and quadrature outputs are arranged in feedback loops to dynamically maintain the null condition. Preferably, the in phase output is indicative of required signal magnitude adjustments, and the quadrature output is indicative of required signal phase adjustments. The signal adjuster may comprise a digital signal processor arranged to receive and process one of or both these outputs in order to generate the adjustment signal to maintain the null condition.

In a preferred embodiment, the microwave power source is locked to a certain frequency and the quadrature output is used to compensate for any errors, e.g. by provided a phase adjustment signal from the signal adjuster (digital signal processor) to a phase shifter (or other frequency tuning device) connected to the microwave power source. In this arrangement, the ESR absorption is exhibited more clearly in the in phase output from the detector, so the adjustment signal generated by the signal adjuster to alter the amplitude of the balance signal is monitored as a measurement of ESR absorption. An alternative arrangement is also feasible, in which dispersion (changes in the phase of the response signal) is measured e.g. by monitoring the magnitude of the phase adjustment signal needed to maintain the null-balance condition.

Preferably, the microwave power source is frequency locked to the resonant frequency of the sample containing resonator. This may significantly reduce the conversion of oscillator frequency and/or phase noise to amplitude noise. This is preferably achieved by providing a loop oscillator as the microwave power source and placing the resonator inside the oscillator circuit itself. This arrangement essentially provides a primary frequency feedback loop in which the oscillator operates close to the optimum tuning condition, without the need for any adjustment. The circuit proposed here (a loop oscillator) is similar to that found in many types of ultra-low noise oscillator. The null-detecting receiver can detect any residual errors with ultimate sensitivity and compensated for them by adjusting a phase shifter in the loop-oscillator. This provides a secondary feedback loop, which can ^λfine tune' the oscillator circuit to provide superior precision and agility when compared with the "automatic frequency control" schemes employed in conventional ESR instruments . This arrangement can ensure optimal noise performance and detection phase stability, even in the presence of large perturbations by a moving laboratory animal. Moreover, having the resonator within the loop oscillator circuit provides independent advantage that is applicable to conventional ESR instruments, i.e. it is not reliant on the null-detection technique for its benefits. As such, another aspect of the invention may provide an electron spin resonance device including a sample containing resonator and a microwave power source arranged to generate an excitation signal which is input to the resonator to cause resonance therein, whereby the resonator produces a detectable response signal indicative of resonance in the resonator, wherein the microwave power source includes a loop oscillator circuit and the resonator is connected within the loop oscillator circuit.

The detection techniques described above may offer the most sensitive and stable microwave measurements that are currently feasible. Their key technical performance characteristics can be orders of magnitude better than any conventional ESR instrument. There is also considerable scope for improvement through the application of modern digital signal processing to improve the efficiency of signal recovery. The techniques proposed herein have the potential to improve the sensitivity of ESR imaging instruments by at least two orders of magnitude, with comparable contributions being made by the innovations in microwave circuit design and signal recovery. This improvement is as large as that resulting from the introduction of Fourier method in NMR and, like that development, has the potential to stimulate many new applications . For example, the null-detection technique may enable a four or more order of magnitude suppression of flicker noise to be obtained, so that white noise dominates down to very low frequencies (a 1 kHz "flicker corner" typically) even for a very low {= 400 K) white noise temperature. The practical consequence of this is that a typical 100 kHz modulated ESR signal can be detected with at least one order of magnitude better (voltage) sensitivity than a conventional detector, and the signal can be modulated at frequencies down to about 1 kHz without invoking a significant noise penalty. The later capability is critically important because it provides more scope to use sophisticated signal modulation and recovery methods, such as those discussed in the following section. These methods add further to the instrumental sensitivity (a second order of magnitude or more) .

The increased sensitivity of the device proposed herein especially if taken in combination wit the signal recovery and analysis techniques described below, is particularly beneficial for ESR imaging, i.e. where the ESR absorption signal is adapted to produce a visual aspect of the sample.

ESR is used extensively by both physical and biological scientists to study the chemical and physical environment of unpaired electrons . Common applications include the study of transition metal and rare earth ions, organic radicals and a wide variety of defects and colour centres in materials. Transition metal ions and organic radicals are especially important in both biological and chemical catalysis. The presence of paramagnetic dopants, defects and other centres is important in determining the performance of advanced materials used in electronics and photonics. Unpaired electrons, especially nitroxide radical "spin labels", can be used to probe the structure and dynamics of proteins, and other biologically and technologically important polymers whilst chemical probes are available that allow ESR to monitor processes of physiological and pharmaceutical importance in living organisms. Essentially the same instrumentation is also used to study magnetic materials such as those used in digital data storage. Of special current importance is the recognition that ESR may be one of the most useful spectroscopic methods in "proteomics" : the characterisation of the structure, function and interactions of the new proteins that have been identified by the human genome and other large scale sequencing projects.

ESR imaging can measure a very wide range of paramagnetic species, and is therefore potentially applicable to an extraordinary diversity of important topics in biological and physical sciences, in medicine and in technology. Recognised biomedical topics where ESR imaging may be applied include: 1. Oximetry. Although oxygen partial pressure is one of the most basic of physiological parameters, there remains no satisfactory method of measuring tissue oxygen within a laboratory animal or patient. The ESR imaging technique proposes herein is capable of measuring known oxygen- sensitive paramagnetic probes to create "oxygen maps" of organs and tissue. In vivo oximetry is particularly important in cancer biology, radiotherapy and diabetes research.

2. Nitric Oxide.

This is used by biology in an extraordinarily diverse range of ways. Best known is its role as a chemical messenger in determining vascular tone, and hence blood pressure. Nitric oxide biochemistry is especially important in cardiovascular disease (including heart attack and stroke) . It is also used as a chemical weapon: white blood cells use it to attack cancer cells whilst it is present in human milk as a natural disinfectant. Nitric oxide is also very important in a variety of inflammatory diseases (including arthritis) and in septic shock. Experimentally, nitric oxide has been detected in vivo using ESR, by employing specific trapping reagents and known naturally occurring nitric oxide binding proteins . ESR imaging allows the production, distribution and removal of nitric oxide to be quantified. However, for this technique to achieve its potential in clinical applications, e.g. related to the diseases mentioned above, the enhanced sensitivity of ESR imaging spectrometers provided herein is required.

3. Reactive Oxygen Species.

These types of species, e.g. hydroxyl and superoxide radicals, are usually considered to be highly toxic. In this context, their uncontrolled production by Fenton chemistry in degenerative brain diseases such as BSE/prion disease and Alzheimer' s disease has been put forward as the primary pathological mechanism. However, there is also evidence that their tightly regulated production in blood vessels also, like nitric oxide, plays a role determining vascular tone. The production of reactive oxygen species after a heart attack, a stroke, or after reperfusion of a translated organ results in significant tissue damage. Moreover, reactive oxygen species are now known to be involved in numerous intracellular signalling cascades and the regulation of gene expression. However, the ability to detect and localise these species in vivo is virtually impossible using conventional ESR imaging. The enhanced measurement sensitivity provided by the present invention makes such detection feasible, thereby provided a pathway to clinical use.

4. Nitroxide spin-labels.

These are stable and non-toxic radicals that can be used to label macromolecules, and probe their local dynamical environment. In vivo imaging of spin-labels may allow sophisticated studies of drug delivery mechanisms to be performed (diffusion of drugs across the skin, the distribution of particles in the lung, for example) . Spin- labelled antibodies may also allow tumours and other pathological tissues to be identified and imaged.

5. Organic radicals and transition metal ions. These occur naturally as the cofactors of many important enzymes and other proteins. Visualising their three dimensional distribution in tissues would allow a wide variety of physiological processes to be monitored. Transition metals ions can exhibit toxic effect, which has been associated with many forms of dementia. Organic radicals can be formed during irradiation of the body by X- rays and ultrasound. Examples of specific areas that may benefit from applying sensitive ESR imaging to the above topics include: in vivo investigation of sepsis processes; monitoring cardiovascular disease; imaging irradiation patterns in radiotherapy; measuring production of toxic radicals in high intensity ultrasound treatment; measuring the mechanism of bacterial infection in higher organisms; measuring the degradation of orthopaedic implants; monitoring the release of drugs in various treatment methods; imaging arthritic joints; and assessing the effect of reperfusion suppressant drugs.

In addition to impacting significantly on scientific quality in these, and other, areas of biomedical research, a substantially improved ESR imaging instrument will bring important improvements in the welfare of laboratory animals. Improved performance will make it possible to take measurements under less extreme physiological conditions. Studying ""healthier" animals can made the experiment a closer model of the human case. Furthermore, the ESR imaging technique proposed herein may be applied in physical sciences and technology outside the biomedical field. The important components in many- electronic, photonic and data storage devices are in essence engineered electron controlling environments. The ability to study closely the movement and distribution of electrons is therefore beneficial in the design of new devices and the identification of defects or impurities in existing objects. For example, ESR imaging may be used in the following technical areas: 1. Electrochemistry Intermediates produced in chemical reactions can have very short life times. An electrochemical cell can be employed to generate these intermediates (radicals) and ESR imaging can be used to generate information about their number, location and molecular structure. By using an electrochemical flow cell it may be possible obtain detailed quantitative data on the interconversion rates of these species by converting their kinetic behaviour into a spatial concentration distribution. 2. Polymer chemistry

Many polymerisation reactions proceed by a radical mechanism. The degradation of polymers frequently occurs by the photochemical generation of radicals and commercial materials include radical trapping dopant chemicals to mitigate their effect. ESR imaging may allow diffusion measurements of these species. It may be especially powerful in understanding catalysis of polymerisation reactions by ultra-sound. Spin labelling methods can be used to report on the structure and dynamics of polymers. 3. Amorphous materials

The most basic of structure-function relationships are still poorly understood in many technologically important amorphous materials such as inorganic and colloidal glasses. ESR imaging may allow structural investigations of such materials to be performed on length scales that are complementary too more established techniques. The method may also, through measurement of magnetic relaxation characteristics, provide information about the dynamical properties of these materials, something of crucial importance in the study of, for example, phase transitions. 4. Catalysis

Transition metal ions form the functionally active sites of many important catalysts. Metallic nanoparticles supported on surfaces or within porous materials are an area of active research. ESR imaging may allow the distribution of these paramagnetic species to be visualised, and also information such as spatial distributions in nanoparticle size to be measured. The diffusion of reactant and product molecules within porous materials may be accessible to kinetic isotope exchange measurements. The occurrence of clustering effects in the corruption and degradation of catalysts may be observable. Improved sensitivity will be critically important in the study of paramagnetic species on surfaces. 5. Semiconductor physics

In many semiconductor devices the spatial distribution of electrons and their environments can be critical in determining performance. The proposed ESR imaging technique may permit visualisation of distributions in dopant and defects in semiconductor devices that are useful in a variety of photonic and spintronic applications .

Typically, ESR measurements are taken by locating the sample containing resonator in a polarising magnetic field which is varied e.g. by sinusoidal field modulation while the ESR absorption is monitored.

Current ESR imaging instruments employ a signal recovery method (sinusoidal field modulation with lock-in detection) that was developed in the 1950s. Images are obtained by "bolting on" a method developed in (nuclear) MRI in the 1970s the measurement of spectra in a series of magnetic field gradients. Historically, these methods arose because of technological limitations at the time. For example, a 100 KHz lock-in amplifier is a signal processor/correlator capable of 10⁵ multiplications/correlations per second. In contrast, modern digital signal processors offer in excess of 10¹⁰ multiplications/correlations per second - a hundred thousand times more powerful. Moreover, digital signal processors employing field programmable gate array (FPGA) technology can be reconfigured (re-wired) under computer control, allowing signal recovery to be optimised "on the fly" in response to previous measurements.

A further aspect of the present invention therefore aims to redesign fundamentally the signal recovery strategies for ESR imaging devices to exploit modern technology. While the methods discussed below are particular suitable for use with null-detection apparatus discussed above, they are also applicable to, and can exhibit improvements with, existing instruments. In a complete four-dimensional spectro-spatial ESR imaging experiment a signal variable (e.g. the microwave absorption) is measured at one or more spatially resolvable volume elements within a sample as a function of four magnetic field variables, which are conventionally taken to be the three orthogonal field gradients, and a homogeneous field applied to the sample by the polarising magnet mentioned above. In a conventional ESR imaging experiment, the homogeneous field component is arbitrarily considered to be "special" and absorption measurements are made while sweeping this variable, keeping the other three constant during a sweep. The fundamental weakness of this approach can be seen by considering the analogous two-dimensional problem a surveyor mapping a valley for example. The conventional ESR imaging approach to mapping the valley would be to walk in a straight line across the valley, measuring the localised slope of the path taken by the surveyor at periodic intervals, and then repeating for many different parallel straight lines, i.e. many different transects. This approach is poor for three reasons: (i) measuring the slope of the surveyor' s path over a short distance provides individual measurements with a poor signal-to-noise ratio; (ii) the method provides little or no information about the relative heights of two regions on different transects even if those regions are physically very near each other. An error in the measurement of the slope anywhere on a transect, or drift in the measuring device results in gross errors in the surveyor' s map and therefore gross errors in a corresponding ESR image; and (iii) the surveyor spends more time than is necessary measuring broad featureless parts of the valley.

At its most general the signal recovery aspect of the invention proposes measuring in sequence an ESR signal (e.g. indicative of absorption and/or dispersion due to ESR) at each of a plurality of predetermined points in the four dimensional magnetic space and cross-referencing one or more of the signals measured later in the sequence with a signal measured earlier in the sequence by returning to the point in four dimensional magnetic space to repeat the earlier signal measurement. This technique permits the relationship between individual measurements to be monitored, which improves the overall measurement sensitivity. Thus, in this further aspect, the present invention may provide a method of obtaining an ESR signal, the ESR signal being detectable from a sample subject to an adjustable magnetic field capable of moving between points in one or more (preferably two or more) dimensions of a four dimensional magnetic space defined by a homogeneous magnetic field component and three orthogonal magnetic field gradient components, the method including: measuring in sequence the ESR signal at each of a plurality of predetermined points in the four dimensional magnetic space; and cross-referencing one or more of the signals measured later in the sequence with a signal measured earlier in the sequence by taking another measurement at the point in four dimensional space of the earlier signal. Preferably, a plurality of cross-references are taken so that the method provides a way of comparing signals not only of adjacent points in the sequence but also from earlier in the sequence. This comparison may permit further analysis of the measured results to generate a set of ESR absorption or dispersion values which best fit the cross-referenced information. In this way, errors and instrumental drift etc can be reduced. This method can therefore give a more accurate and complete picture of the ESR signal in the four dimensional magnetic space.

Preferably, the predetermined points in the sequence are more widely spaced than measurements taken in conventional ESR imaging devices. Conventional ESR instruments needed to apply small modulation amplitudes to maintain spectral resolution. However, modern digital processors are powerful to reconstruct a spectrum without loss of resolution. Large step-like digital modulations of the applied magnetic field are therefore feasible. This can maximise the signal-to-noise ratio of the individual measurements by allowing much more of the available signal power to be recovered. Thus, measurements are no longer constrained e.g. to measure only a small portion of the absorption (or dispersion) peak gradient as in the conventional modulated sweep technique. The present method may permit adjacent measurements to include the maximum and base level of the signal peak. Data points spaced by between 30 and 100 Gauss in the field space may be sufficient to achieve this in typical sample types.

The method above can be understood with reference to the 2D surveying analogy. A surveyor would not use the conventional ESR imaging algorithm to map a valley. Instead (i) a grid of widely space measurement points would be setup; (ii) the relative heights of these points would be measured, and cross-referenced to generate a set of numbers (altitudes) most consistent with the measurements (thereby minimising susceptibility to measuring device noise and drift) . Moreover, areas of the valley where the height was changing rapidly would be identified and additional, more closely spaced, measurements would be made in these areas (which means that the surveyor spends more time making measurements where the largest errors in the initial map are likely to exist and hence has a higher probability of reducing those errors in a given measurement period - the surveyor is making use of Bayesian inference principles in choosing the next measurements to make) . This targeting approach may be applied to the present invention. An additional sequence of more closely spaced points in the four dimensional magnetic space may be generated around measured point or points that exhibit e.g. the greatest difference in absorption signal to surrounding measurement points. The criteria used to determined this targeting approach can vary depending on the aim of the user. In general it is preferable to target areas where there is likely to be the most error. Examples of the criteria used to determined this may include: Does a particular data point differ significantly from the value expected from an interpolation or extrapolation of surrounding data points? Are several such anomalous data points localised in a specific region? In a related aspect, the present invention may provide a computer product e.g. programmed with an algorithm arranged to execute the above described method on a computer which is preferably arranged to control the polarising magnet to navigate between points in the four dimensional magnetic space.

One technical feature that facilitates the implementation of the new method is the ability to move rapidly between two arbitrary points in the four dimensional magnetic field space so that the difference in microwave absorption (or dispersion) can be measured.

Preferably, a system will allow a maximum step of around 100 G to be made in about 1 ms and smaller steps made in proportionately smaller times. The technical requirements of the coils and drivers (physical size, inductance, homogeneity, maximum field strength, amplifier current and slew rate) are very similar to those typically employed in nuclear magnetic resonance microscopy and are therefore known to be feasible with standard methods. The magnetic field sweep/slew rate is preferably smaller than that typically employed in conventional field modulation; the present approach may therefore be less susceptible to transient "rapid passage" effects and magnetic induction induced mechanical vibrations. Heat dissipation is comparable to that of conventional ESR imaging field gradient coils, and can be removed with modest water cooling. The maximum acquisition rate of 100 kHz and the average sustained acquisition rate of 10 kHz are sufficiently low to allow the signal recovery algorithms to be performed in real-time by a digital signal processor. The increase in sensitivity of the new modulation scheme arises, fundamentally, from (i) that we can employ magnetic field steps that are on average larger than the sinusoidal modulation employed in conventional detection (which is limited by resolution considerations); (ii) that the instrument can search intelligently for regions of the field modulation space where large errors are expected to have occurred; and (iii) it has much higher immunity to instrumental drift (which is especially important in the suppression of animal movement artefacts in biomedical applications). A basic sensitivity improvement over conventional ESR imaging techniques well in excess of one order of magnitude may be achieved. In addition to this, a significant improvement in image quality can be expected (i.e. better resolution and fewer artefacts). Where fundamental sensitivity considerations permit, the new algorithm should allow two orders of magnitude faster acquisition than conventional lock-in techniques.

Improvements in basic instrumental sensitivity will impact in a variety of ways. Being able to perform experiments with a lower concentration of paramagnetic electrons, critically important though that is, is just one aspect. Currently ESR imaging, like (nuclear) MRI, has a spatial resolution that is sensitivity limited in common applications. The number of volume elements resolvable in a given period scales as the square of the sensitivity.

Thus a projected 100 times improvement in sensitivity will allow 10,000 times more volume elements to be resolved. In a 3D experiment each element will have 20 times smaller linear dimensions than previously. The practical consequence of such an improvement on a mouse experiment, for example, will be the ability to resolve the distribution of paramagnetic species within organs (about 100 μm resolution) , rather than between organs (about 2 mm resolution) , as is currently the case in typical experiments. Still larger improvements in linear resolution will be possible in the acquisition of 2D images (a factor of 100) and ID concentration gradients (a factor of 10,000) . Measurements of the former type will be important in the imaging of pharmaceutical delivery by diffusion across the skin.

Another important consequence of improved sensitivity is the ability to perform experiments more quickly. In typical sensitivity limited experiments, the time required to achieve an acceptable signal to noise ratio scales as the inverse square of the sensitivity. Thus a projected 100 times improvement in sensitivity would allow experiments to be performed 10,000 times faster. Currently- inaccessible experiments (that would take months or years to measure in principle) will be measurable in minutes or hours. Crucial in the full exploitation of this opportunity will be rapid data acquisition techniques. Preferably, the device of the present invention has a data- point acquisition rate 10 kHz or more, which is at least two orders of magnitude faster than current methods permit. It would allow, for example, a 10⁸ (100⁴) point four- dimensional spectro-spatial image to be acquired in 3 hours. Such an experiment is not currently feasible. In experiments where the spectrum of the species of interest is known (a common occurrence in practice) , very rapid image acquisition becomes possible (for example a 100 x 100 image in 1 second) . This will allow more satisfactory- kinetic measurements to be performed. There are many important physiological and pharmacological processes that occur on the second timescale. Being able to measure such processes, even with modest spatial resolution, would be very important. Fast measurements will also reduce problems due to animal movement, and may allow reduced used of anaesthetics that can invalidate some types of measurement. Short, unanaesthetised measurements also provide superior animal welfare.

In summary, the aspects of the invention above relate to apparatus and method for electron spin resonance imaging, which allows molecular species of importance in basic physiological, patho-physiological and pharmacological processes to be mapped inside a living organism. The invention may provide

(i) A spectroscopic or imaging device which incorporates advanced microwave metrology methods which offer dramatically more sensitive detection than the approach currently employed. Their integration into an ESR imaging instrument may improve basic sensitivity by at least one order of magnitude.

(ii) A signal modulation and recovery technique that fully exploit the power and flexibility of modern digital signal processors. A significant improvement in signal recovery efficiency may contribute an additional one or more orders of magnitude to the basic sensitivity. Improved suppression of artefacts due to instrumental instability may significantly improve image accuracy. Data collection speeds two orders of magnitude faster than current methods may also be possible.

An overall fundamental sensitivity improvement of two or more orders of magnitude may therefore be achieved. In typical sensitivity limited experiments this translates into a four order of magnitude improvement in spatial information or measurement speed. Rapid data acquisition strategies may allow this improved spatial and temporal resolution to be practically exploited. Advances of these magnitudes may make a very significant impact on the usefulness of ESR imaging in biomedical research. They may allow chemical species of relevance to major human diseases to be imaged for the first time, and may allow existing applications to be performed with greatly improved spatial and temporal resolution. This improved capability may be very important in the quantitative evaluation of new therapies in animal models of human disease. It may also allow such experiments to be performed under conditions of significantly improved animal welfare. The advance will also impact on non-biomedical applications such as the analysis of technologically important materials and devices .

BRIEF DESCRIPTION OF THE DRAWINGS

Examples of the present invention will now be described with reference to the accompanying drawings, in which :

Fig. 1 shows a schematic system diagram for a null detecting ESR detection device that is an embodiment of the invention;

Fig. 2 shows an implementation of a loop oscillator ESR detection instrument; and

Fig. 3 shows a flow diagram of a signal recovery method that is an embodiment of the invention.

DETAILED DESCRIPTION; FURTHER OPTIONS AND PREFERENCES

Fig. 1 shows a schematic arrangement of an ESR imaging apparatus that is to illustrate the null detecting idea of the present invention. A microwave power source 10 generates an excitation signal 12 that is input via circulator 14 to resonator 16 which contains a sample 18 to be imaged. The resonator 16 is surrounded by an adjustable polarising magnet 20 which is arranged to controllably vary the magnetic field conditions in the resonator 16 while the excitation signal 12 is input in order to monitor how ESR absorption varies with magnetic field conditions. The resonator 16 generates a response signal 22 in response to the excitation signal 12. Typically, the response signal

22 is made up of reflection of the excitation signal 12 and resonance in the sample. It may also include some of the excitation signal itself that may leak through the circulator 14. The response signal 22 is directed by the circulator 14 towards a detector 24. Before it reaches the detector 24, the response signal 22 is coupled to a balance signal 26 generated by a balance signal generator 28 to form a detection signal 30 which is input to the detector 24. A phase reference signal 32 which has a fixed (i.e. substantially constant with time) phase relationship with the excitation signal 12 is generated by a reference signal generator 34. The phase reference signal 32 is provided to the detector 24 and mixed (multiplicatively) with the detection signal 30 both in phase to produce an in phase output 36 and 90° out of phase to produce a quadrature output 38. The quadrature output 38 is representative of phase differences between the detection signal 30 and the phase reference signal 32. It is desirable for the excitation signal 12 to be at the resonant frequency, so the quadrature output 38 is provided to the microwave power source 12 to enable compensation frequency adjustments to be made. The in phase output 36 is representative of the amplitude of the detection signal 36. The key feature of the present invention is to dynamically maintain the detection signal 30 at or as close as possible to zero. This is done by arranged for the balance signal 26 to substantially cancel the response signal 22. When the detection signal 30 is non-zero, a feedback amplifier/generator 40 (e.g. embodied by a digital signal processor) , which receives the in phase output 36, detects it and sends an appropriate adjustment signal 42 to the balance signal generator 28 to alter the balance signal 26 to restore the null condition. A dynamic feedback loop is therefore formed by the in phase output 36, adjustment signal 42 and balance signal 26. The adjustment signal 42 is monitored by measurement apparatus (not shown) because it represents a measure of ESR absorption in the resonator 16.

There are many ways of implementing the general concepts illustrated in Fig. 1 is a practical system.

Supplying an excitation signal and reference signal and detecting an in phase and quadrature output are all conventional steps. Balance signals to try to counteract the reflected excitation signal are also known. The present invention goes further proposes that the balance signal should cancel substantially all of the response signal and be dynamically adjustable on the basis of detected output to maintain this cancellation. Moreover, in the embodiment above the ESR measurement is obtained from the adjustment signal generated to maintain the null condition. Obtaining an ESR measurement from the magnitude of the adjustments in amplitude and/or phase required to maintain the null condition has not been done before.

Fig. 2 shows a more detailed implementation of the present invention. The microwave power source in this arrangement is a loop oscillator 44 which includes a bipolar low noise amplifier 46 arranged as a preamplifier to provide a low noise feed to a variable gain amplifier 48 for generating the excitation signal 12. In this arrangement, the resonator 16 is located in the loop oscillator circuit 44 itself. This arrangement means the loop oscillator 44 is automatically frequency locked to the resonator 16 because any frequency error will result in a phase shift in the feedback reaching the loop amplifiers. The loop oscillator 44 also includes a phase shifter 50 that is arranged to receive the quadrature output 38 via a servo circuit 52. This represents a secondary feedback loop to further enhance frequency locking. This design is particularly advantageous because the primary loop provides an agile (fast responding) ^coarse' adjustment while the secondary loop provides accurate ^λfine' adjustment. Because the secondary loop only needs to make minor adjustments, it does not introduce significant excess noise. This can be a problem in conventional devices because the phase shifter must make large adjustments e.g. to compensate for changes in resonant frequency in the resonator.

The loop oscillator 44 also includes a variable attenuator 54, which allows adjustment of the excitation signal 12 to a suitable level.

While the loop oscillator 44 is shown in Fig. 2 in conjunction with the null detecting arrangement of Fig. 1, it is equally applicable to conventional devices, where the primary and secondary feedback loops mentioned above in particular may provide beneficial effects. The phase reference signal 32 is generated in the device shown in Fig. 2 from a forward power coupler 56 on the loop oscillator 44. This ensures it has a fixed phase relationship with the excitation signal 12. A variable delay line 58 is provided to ensure the phase reference signal 32 is properly received at a mixer 60 for mixing with the detection signal amplified by low noise amplifier 62 (which may be a GaAs field-effect amplifier) . Modern low noise amplifiers can be used because the null detecting technique of the present invention minimises the effect of transposed flicker noise, which is particularly prevalent in such devices .

The balance signal 26 is generated in the device shown in Fig. 2 from a forward power coupler 64, and it travels through a variable delay line 66 and a variable attenuator 68 before being coupled to the response signal 22 at a coupler 70. The delay line 66 ensures that the balance signal 26 is in anti-phase with the response signal 22 at the coupler 70 and the variable attenuator 68 is arranged to ensure that the balance signal amplitude is sufficient to cancel the response signal 22. The attenuation is selected on the basis of an adjustment signal 42 generated by a servo circuit 72 on the basis of the in phase output 36. The adjustment signal 42 is also provided to a measurement apparatus (not shown) to generate an ESR image.

The device in Fig. 2 shows a Mach-Zehnder interferometric arrangement to achieve the null condition. Other interferometric arrangements can also be used.

To create an ESR image of a sample, it is desirable to measure the way the ESR absorption changes as the magnetic field condition applied to the sample changes. The new signal recovery technique proposed above requires the polarising magnets 20 to be navigable through four dimensional magnetic field space (three orthogonal field gradients and a homogeneous field along one of the gradients) . This can be achieved with four independent coils each having its own controllable (e.g. programmable) power supply e.g. arranged as the known manner for NMR microscopy explained in Callaghan' s "Principles of Nuclear Magnetic Resonance Microscopy", 1991, OUP.

A volume in the four dimensional magnetic space in which the ESR absorption is to be investigated is defined. A ^Λgrid' of individual points within this volume e.g. equally spaced from one another is then defined together with a sequence in which ESR measurements at those points will be taken. The method of the present invention then includes a series of cross-references by inserting into the sequence return visits to previously measured points so that data relating to the relationship between a plurality of the points in the sequence can be obtained and analysed to generate a more accurate overall assessment of the ESR absorption results.

Fig. 3 shows a flow diagram which outlines a basic signal recovery method 100 according to this technique. After the apparatus (e.g. of the embodiments discussed above) is prepared, at step 102 a magnetic field is applied to the sample corresponding to a first predetermined point in the four dimensional magnetic volume. At step 104, a ESR response of the sample is measured. At step 106, the system queries if n measurements have been taken. If no, the method moves to step 108, where the magnetic field is adjusted to the next predetermined point in the sequence. The method then returns to step 104, where a ESR response for that volume is detected, and again to step 106. When n measurements have been taken, the answer at step 106 is yes, and the method moves to step 110. At step 110 a cross-reference is made by repeating a measurement already- taken, i.e. navigating the magnetic field back to a previously measured point in magnetic space and measuring the ESR response again. Following this, the method may reset n to zero and return to the loop sequence defined by steps 104, 106 and 108.

The defined volume may itself be four-dimensional, but the present method is equally applicable to sub-volumes in two or three dimensions (e.g. where one or more of the magnetic field space variables is kept constant) .

For example, the volume may have a regular geometric shape. The grid of points may be located on the surface of the shape. One example is a 4D hypercube whose 16 corners in four-dimensional space comprise a grid of points making up a measurement cell. Preferably, the measurement sequence is defined so that the differences between all the points is measured - 120 differences in all. To measure all the differences one may be forced to measure some of them twice because there are topological constraints on navigating around the cell. To measure all differences, each point must be measured eight times (i.e. cross referenced seven times). Of course, only a sample e.g. subset of the 120 possible differences may be taken. The size and nature (e.g. geometry, dimensional extent) of a measurement cell may depend on the type of sample and what exactly is being measured. A benefit of the method described herein is the broad adjustability of magnetic field parameters, e.g. to allow a wide range of measurement cell shapes and sizes to be investigated. For example, each of the four independent field generators may^¬ be arranged to adjust the field at a specific point in the sample to any value between 0 and about 300 Gauss. In a preferred embodiment, the method comprises selecting a shape for the measurement cell and defining a ^coarse' grid of measurement points which lie in or on of the cell to take a ^λfirst pass' measurement. After the first grid of points is measured and cross-referenced, the results are analysed to determine where in the volume it is necessary to measure in more detail (e.g. a more closely spaced grid) to refine the results e.g. obtain better resolution and reduce errors. A sub-volume of the initial volume may therefore be determined and a second ^λgrid' of points established within the sub-volume on which the method described above is repeated.

For example, a four dimensional data space within which measurements may be taken may comprise 100 x 100 * 100 x 100 = 10⁸ regularly spaced measurement points. Rather than attempt to measure between all of these points, the present invention aims to focus detailed measurement at points located around areas of interest (e.g. rapid variation in ESR signal) . The first pass measurement may cover the whole data space, but only comprise measurements between less than 100 points. The first pass points are also preferably distributed substantially regularly over the measurement cell. For a 4D cube, the first pass may measure an array of 3 * 3 ^χ 3 χ 3 = 81 points e.g. taken along the edges and through the centre of the 4D cube. There are 3321 differences to be measured in this case, so the first pass measurement can be achieved relatively quickly.

The decision about which area of the data space (measurement cell) to measure next can be based on two criteria; (i) what is the probability of any missed signal being in that region and (ii) how much effort (time) is required to reach that region.

Given an existing grid of measured data points, each (4D) sub-volume with adjacent points from that grid at its corners may be considered a "sub-cell". A simple starting point for assessing how likely it is that missing signal is present within a sub-cell might be: (the square of the difference in signal voltage across the cell) x (the volume of the cell) . The first factor is the proportional to the change in signal power here assuming that errors in the "digitisation" of the signal are most likely to occur in regions where it is changing rapidly, or where the points are widely spaced.

More sophisticated criteria may be used, e.g. criteria that take into account what is going on in the surrounding cells, e.g. second derivatives of the signal, or deviations from extrapolations taken from surrounding points etc.

The method may also include a running estimate of the signal to noise ratio, so that the instrument can be prevented from measuring fine grids of points when really what is needed is additional measurements of the first grid to improve their statistics. Simply dividing the estimating of missing signal power factor by a factor reflecting how many times the cell has been measured is a simple example of this.

After the raw data is collected using the techniques described above, it needs to be further converted to create an image. The conversion of the raw data into an image (or a 4D spectro-spatial image) is a mathematical transformation known as an "inversion". There are many methods of performing this transformation, although only one is commonly employed in ESR imaging - filtered back- projection. This method gained popularity when computer power was severely limited. The key advantage of alternative, more sophisticated, methods is that they allow information that we know about the sample/animal to be introduced prior to image reconstruction. Some methods (for example the projected Landweber method) allow conventional "deterministic" constraints to be incorporated. Still more powerful are statistical algorithms employing Bayesian inference techniques. These allow "probabilistic" constraints to be incorporated in an extremely flexible way. In the case of a mouse experiment, for example, the number, size, shape and relative position of the key organs are known with reasonably high certainty prior to image reconstruction. It is possible to use this information to construct a more accurate image. Exactly this approach has already been employed with considerable success in the context of positron-emission tomography by using anatomical information from MRI. Such techniques can be applied to the present invention, i.e. of what an ESR spectrum looks like can be incorporated at the time an image is being reconstructed.

An image is not in itself the answer to a biological question. Real biological questions include: is the data consistent with the molecular species of interest being localised in one, four or seventy principle regions? With what accuracy can we estimate the total number of molecules in each region? With what accuracy can we determine the volume of the localised region and hence the average molecular concentration? Do we know enough about the shape and spatial distribution of areas of localisation to associate them with known anatomical structures? Again powerful statistical analysis, employing Bayesian and related methods, can provide answers to such questions from available data. However, it is possible to go further than the use of such methods to analyse a single image. Statistical approaches to image reconstruction generate an array of possible images with each image having a different probability. Conventionally, one focuses on a single "most probable" image. However, when the goal of the analysis is to answer a biological question, rather than generate a single image, it is more correct to include lower probability images in one's analysis. In biomedical applications, it may therefore be preferable for the present method to focus on the task of answering the key biological questions, rather than the generating of a single recognisable image. Thus, the method may include producing a plurality of images using an ESR signal, wherein each image has a probability associated with it. Such a tool may be advantageous in image analysis by offering the possibility of obtaining statistically significant answers to the biologically relevant questions with far less information than required to generate a conventional image.

Claims

1. A device for detecting resonant absorption or dispersion of radiation, the device including: a resonator for receiving a sample that is subjected to a controllable magnetic field; a power source arranged to generate an excitation signal which is input to the resonator, whereby the resonator produces a response signal; a balance signal generator arranged to produce a balance signal which is coupled to the response signal to form a detection signal; a reference signal generator arranged to produce a reference signal having a stable phase relationship with the excitation signal; and a detector arranged to receive the detection signal and the reference signal and to generate an output based on a multiplicative mixing of the detection signal and reference signal; wherein the device includes a signal adjuster connected to receive the output from the detector and arranged to communicate an adjustment signal to the balance signal generator, and wherein the adjustment signal is generated based on the output from the detector to form a feedback loop in which the balance signal substantially cancels the response signal.

2. A device according to claim 1, wherein the adjustment signal is monitored to generate a resonance measurement.

3. A device according to claim 2 arranged to obtain a resonance measurement for each of a plurality of spatially resolvable volumes in the sample.

4. A device according to claim 3, wherein the resonance measurements are for assessing the spatial dependence of the ESR spectrum of the sample in one, two or three dimensions.

5. A device according to claim 3, wherein the resonance measurements are compared with the ESR spectrum of a known species to determine the spatial concentration of that species in the sample in one, two or three dimensions.

6. A device according to claim 5 arranged to display an image of the spatial concentration of the species in the sample .

7. A device according to any preceding claim comprising a interferometer in which the balance signal and response signals are produced in respective arms of the interferometer .

8. A device according to any preceding claim, wherein the reference signal generator includes a forward power coupler arranged to couple the excitation signal.

9. A device according to any preceding claim, wherein the balance signal generator includes a forward power coupler arranged to couple the excitation signal.

10. A device according to any preceding claim, wherein the output generated based on the multiplicative mixing comprises an in phase output and a quadrature output, either one or both of which outputs are provided to the signal adjuster.

11. A device according to claim 10, wherein the adjustment signal is arranged to vary the amplitude of the balance signal based on the in phase output to dynamically maintain cancellation of the response signal.

12. A device according to claim 10 or 11 including frequency tuning means arranged to vary the phase of the balance signal based on the quadrature output to dynamically maintain cancellation of the response signal.

13. A device according to claim 12, wherein the balance signal is phase locked to the excitation signal, and the frequency tuning means is arranged to adjust the phase of the excitation signal.

14. A device according to any preceding claim, wherein the power source is frequency locked to the resonant frequency of the resonator.

15. A device according to claim 14, wherein the microwave power source comprises a loop oscillator circuit and the resonator is located inside the loop oscillator circuit .

16. An electron spin resonance device including: a resonator for containing a sample that is subjected to a controllable magnetic field, and a power source arranged to generate an excitation signal which is input to the resonator to cause resonance therein, whereby the resonator produces a detectable response signal indicative of resonance in the resonator, wherein the power source includes a loop oscillator circuit and the resonator is connected within the loop oscillator circuit.

17. A computer-implemented method of obtaining an electron spin resonance (ESR) measurement, the ESR measurement being detectable from a sample subject to a controllably adjustable magnetic field space, the method comprising computer operations including: adjusting the magnetic field experienced by the sample to navigate in sequence between a plurality of predetermined points in the magnetic field space; measuring in sequence the ESR response at each of the plurality of predetermined points in the magnetic field space; and cross-referencing one or more of the responses measured later in the sequence with a response measured earlier in the sequence by adjusting the magnetic field to return to the point in the magnetic field space of the earlier response and taking another measurement.

18. A method according to claim 17, wherein the magnetic field space comprises two or more dimensions of a four dimensional magnetic space defined by a homogeneous magnetic field component and three orthogonal magnetic field gradient components.

19. A method according to claim 17 or 18, wherein the plurality of predetermined points lie in or on a predefined volume in the magnetic field space, and the method includes calculating the difference between measured ESR responses by comparing the response measured at each point with one or more or all of cross-reference responses measured at the other points .

20. A method according to claim 19, wherein each cross-reference response is measured directly after the response to which it is to be cross-referenced.

21. A method according to claim 19 or 20, wherein adjacent points in the predefined volume form measurement cells, and the method includes using the measured responses to select a measurement cell for further investigation.

22. A method according to claim 21, including defining a further plurality of measurement points in the selected measurement cell and repeating the adjusting, measuring and cross-referencing steps for the points in the measurement cell.

23. A computer-readable medium having computer executable instructions thereon arranged to cause a computer to perform a method according to any of claims 17 to 22.

24. A device according to any of claims 1 to 15 arranged to perform the method of any of claims 17 to 22.