AU2007302889A1 - NMR spectra extraction method - Google Patents

Description

WO 2008/041866 PCTINZ2007/000287 NMR SPECTRA EXTRACTION METHOD FIELD OF THE INVENTION 5 The present invention relates to a method of extracting nuclear magnetic resonance (NMR) spectra. In particular, but not exclusively, the present invention relates to an NMR spectroscopy method for obtaining a high-resolution NMR spectrum from a sample under investigation in the presence of static magnetic field gradients. 10 BACKGROUND TO THE INVENTION NMR spectroscopy is an analytical and diagnostic technique that can be used for structural and quantitative analysis of a compound in a mixture. An NMR spectrometer generally comprises one or more magnets producing a strong magnetic field within a test 15 region. NMR signals are characterised by a spectrum of frequencies that contain valuable information about the chemical environment of the nucleus under investigation. For example, the most commonly used NMR nucleus, the proton, exhibits a range of NMR frequencies, in the order of 10 ppm in separation, depending on the molecular site from which the proton signal is resolved. This effect is known as 'chemical shift'. Water and 20 oil signals, for example, are separated by a chemical shift of around 3.5 ppm. The frequency of an NMR signal is proportional to the field strength in the test region. As such, in situations where detailed spectral information is required, an ideal NMR apparatus must provide a magnetic field that is sufficiently uniform so as to ensure that 25 the spread of frequencies associated with magnetic field variation is much less than the frequency spread associated with the desired spectral information, for example, the chemical shift. In many NMR apparatuses,. it is not feasible to provide a magnetic field of sufficient 30 uniformity over the test region such that the desired spectral resolution can be achieved. One such apparatus is the one-sided-access type NMR apparatus, such as that described in the present applicant's PCT Publication WO 2004/008168. The one-sided-access apparatus projects a magnetic field outside a surface to obtain an NMR signal from a WO 2008/041866 PCT/NZ2007/000287 material in a test region exterior to the apparatus. It is common for the projected magnetic field to range from 100 ppm to 10,000 ppm over the test region. The effect of such a large field inhomogeneity is to produce a very wide NMR 'linewidth', the term given to describe the undesired spectral spread due to magnetic field inhomogeneity. 5 Inhomogeneity in other types of NMR spectrometer systems, such as bench-top systems employing permanent magnets, electromagnets and Halbach array designs, may also result in a wide NMR linewidth. One well-established method. of removing the effects of magnetic field inhomogeneity is 10 to use a spin-echo. In this technique, spin phases in the field are reversed by means of an RF pulse, so that any dephasing that has occurred in the inhomogeneous field will be subsequently refocused. However, the spin-echo also refocuses desired dephasing due to chemical shift effects and, as such, is not useful for obtaining chemical shift information. 15 To extract information about spectral properties that are much narrower in frequency spread than the NMR linewidth, it is necessary to provide other approaches. One approach is to 'shim' the magnetic field by adding adjustable magnetic field gradients that oppose, and hence cancel, the undesired static magnetic field gradient. This is done by adjusting the current flowing in a set of coils, which adds various magnetic field gradients 20 to the static field such that any inhomogeneity may be compensated. Another approach is that proposed in US 6,674,282 to Pines et al. That approach relates to a method of extracting NMR spectral information through the use of correlated, composite Z-rotation pulses in the presence of spatially matched B, and B, 25 inhomogeneous fields. This approach has proved useful for reducing the linewidth due to B,, inhomogeneity in one-sided-access type apparatus; however, the approach requires a B, coil to be specially designed to spatially match the inhomogeneities present in the NMR apparatus. As such, the approach is not applicable to all NMR apparatuses. 30 In this specification, where reference has been made to patent specifications, other external documents, or other sources of information, this is generally for the purpose of providing a context for discussing the features of the invention. Unless specifically stated otherwise, reference to such external documents or sources of information is not to be WO 2008/041866 PCT/NZ2007/000287 construed as an admission that such documents or sources of information in any jurisdiction are prior art, or form part of the common general knowledge in the art. OBJECT OF THE INVENTION 5 It is an object of the present invention to either provide an improved method of extracting an NMR spectrum from a sample under investigation in the presence of an inhomogeneous magnetic field, or at least provide the public with a useful choice. 10 SUMMARY OF THE INVENTION In one aspect, the invention comprises a method of extracting NMR spectra from a sample under investigation. The method comprises subjecting the sample under investigation to an inhomogeneous static magnetic field, one or more RF pulses, and one 15 or more applied magnetic field gradients to acquire NMR sub-sample signals from two or more regions of the sample under investigation; subjecting a reference sample to the inhomogeneous static magnetic field, the one or more RF pulses, and the one or more applied magnetic field gradients to acquire NMR sub-sample signals from two or more regions of the reference sample; deconvolving the NMR sub-sample signals acquired from 20 regions of the sample under investigation with the NMR sub-sample signals acquired from corresponding regions of the reference sample to produce a deconvolved signal for each region; and summing together the deconvolved signals for each region to obtain the NMR spectra of the sample under investigation. 25 The term 'comprising' as used in this specification and claims means 'consisting at least in part of', that is to say when interpreting statements in this specification and claims which include that term, the features, prefaced by that term in each statement, all need to be present but other features can also be present. Related terms such as 'comprise' and 'comprised' are to be interpreted in similar manner. 30 In another aspect, the method comprises subjecting the sample under investigation to an inhomogeneous static magnetic field, one or more RF pulses, and one or more applied magnetic field gradients to acquire NMR sub-sample signals from two or more regions of WO 2008/041866 PCT/NZ2007/000287 the sample under investigation; deconvolving the NMR sub-sample signals acquired from regions of the sample under investigation with NMR sub-sample signals from corresponding regions of a reference sample to produce a deconvolved signal for each region; and summing together the deconvolved signals for each region to obtain the NMR 5 spectra of the sample under investigation. In a further aspect, the method comprises receiving two or more NMR sub-sample signals from two or more regions of the sample under investigation; deconvolving the NMR sub-sample signals with NMR sub-sample signals from corresponding regions of a 10 reference sample to produce a deconvolved signal for each region; and summing together the deconvolved signals from each region to obtain the NMR spectra of the sample under investigation. Preferably, the step of subjecting the sample under investigation to an inhomogeneous 15 static magnetic field comprises placing the sample in the test region of an NMR spectrometer system where the homogeneity of the magnetic field is insufficient to obtain the desired spectral resolution. In one form, the NMR spectrometer system is a one sided-access NMR apparatus. 20 Preferably, the NMR sub-sample signals are acquired as spin-echoes. Alternatively, the NMR sub-sample signals are acquired as free induction decays (FIDs). Preferably, the one or more applied magnetic field gradients comprise one or more phase encoding gradients stepped through N discrete values. Alternatively, the one or more 25 applied magnetic field gradients comprise frequency-encoding-gradients. Preferably, subjecting the sample under investigation to the one or more applied magnetic field gradients comprises using a spin echo or gradient echo method. In one form, the gradient echo method is echo planar spectroscopic imaging (EPSI). 30 Preferably, deconvolving the NMR sub-sample signals comprises dividing time-domain NMR sub-sample signals acquired from the sample under investigation with corresponding time-domain NMR sub-sample signals acquired from the reference sample.

/A

WO 2008/041866 PCT/NZ2007/000287 In a further aspect the invention comprises a nuclear magnetic resonance spectra extraction system. The system is configured to subject a sample under investigation to an inhomogeneous static magnetic field, one or more RF pulses, and one or more applied 5 magnetic field gradients to acquire NMR sub-sample signals from two or more regions of the sample under investigation; subject a reference sample to the inhomogeneous static magnetic field, the one or more RF pulses, and the one or more applied magnetic field gradients to acquire NMR sub-sample signals from two or more regions of the reference sample; deconvolve the NMR sub-sample signals acquired from regions of the sample 10 under investigation with the NMR sub-sample signals acquired from corresponding regions of the reference sample to produce a deconvolved signal for each region; and sum together the deconvolved signals for each region to obtain the NMR spectra of the sample under investigation. 15 In a further aspect the nuclear magnetic resonance spectra extraction system comprises sample imaging means for subjecting a. sample under investigation to an inhomogeneous static magnetic field, one or more RF pulses, and one or more applied magnetic field gradients to acquire NMR sub-sample signals from two or more regions of the sample under investigation, and for subjecting a reference sample to the inhomogeneous static 20 magnetic field, the one or more RF pulses, and the one or more applied magnetic field gradients to acquire NMR sub-sample signals from two or more regions of the reference sample; deconvolution means for deconvolving the NMR sub-sample signals acquired from regions of the sample under investigation with the NMR sub-sample signals acquired from corresponding regions of the reference sample to produce a deconvolved 25 signal for each region; and summing means for summing together the deconvolved signals for each region to obtain the NMR spectra of the sample under investigation. As used herein. the term "(s)" following a noun means the plural and/or singular form of that noun. 30 As used herein the term "and/or" means "and" or "or", or where the context allows both.

WO 2008/041866 PCT/NZ2007/000287 To those skilled in the art to which the invention relates, many changes in construction and widely differing embodiments and applications of the invention will suggest themselves without departing from the scope of the invention as defined in the appended claims. The disclosures and the descriptions herein are purely illustrative and are not 5 intended to be in any sense limiting. The invention consists in the foregoing and also envisages constructions of which the following gives examples only. 10 BRIEF DESCRIPTION OF THE DRAWINGS Preferred forms of the method of the present invention will now be described with reference to the accompanying figures. 15 Figure 1 shows a schematic of the preferred form method of the present invention. Figure 2A shows an example pulse sequence employed to acquire localised spectra from sub-regions of a sample. 20 Figure 2B shows an example pulse sequence employed for the method of gradient echoes. Figure 3A shows the spectrum acquired of ethanol (the sample under 25 investigation) in the presence of significant magnetic field inhomogeneity. Figure 3B shows the spectrum acquired of water (the reference sample) in the presence of significant magnetic field inhomogeneity. 30 Figure 4 shows the resulting noise-dominated spectrum from raw deconvolution of the ethanol spectrum of Figure 3A by the water spectrum of Figure 3B.

WO 2008/041866 PCT/NZ2007/000287 Figure 5A shows the time-domain NMR signals acquired using the pulse sequence of Figure 2A on ethanol (the sample under investigation) as a function of the spatial encoding vector, k. 5 Figure 5B shows the time-domain NMR signals acquired using the pulse sequence of Figure 2A on water (the reference sample) as a function of the spatial encoding vector, k. Figure 5C shows the time-domain NMR signals acquired using the pulse sequence 10 of Figure 2A on ethanol (the sample under investigation) as a function of position, 5, in the phase-encode imaging direction. Figure 5D shows the time-domain NMR signals acquired using the pulse sequence of Figure 2A on water (the reference sample) as a function of position, , in the phase 15 encode imaging direction. Figure 6 shows the high-resolution NMR spectrum of ethanol that results from imaged deconvolution of the sub-region signals of Figure 5C by those of Figure 5D. 20 Figures 7A-7C shows comparatively: (a) the ethanol spectrum acquired of the entire target region in the presence of the inhomogeneous field (Figure 7A), (b) the noise-dominated ethanol spectrum that results from raw deconvolution (Figure 7B) and (c) the high resolution ethanol spectrum resulting from the imaged deconvolution method of the present invention (Figure 7C). 25 WO 2008/041866 PCT/NZ2007/000287 DETAILED DESCRIPTION OF PREFERRED FORMS Overview of the Preferred Form Method 5 In Figure 1, the preferred form method of the present invention is shown schematically. The method as shown is used to extract NMR spectra from a target volume of a sample under investigation, indicated generally by reference numeral 100, in the presence of an inhomogeneous magnetic field. 10 As will be explained later, the method includes an imaging step, where the sample under investigation 100 is divided into sub-regions in the- direction of an imaging gradient, and localised NMR sub-sample signals (herein 'sub-sample signals') are acquired from the sub regions. In particular, the imaging gradient comprises at least one applied magnetic field gradient such that imaging techniques can be used to spatially sub-divide the NMR signal 15 into sub-sample signals arising from different regions of the sample. The gradient may be a phase-encoding or frequency-encoding gradient, and can be applied on its own or in combination with other signals, such as a radio frequency (RF) pulse. In Figure 1, three sub-sample signals 102a, 102b and 102c are shown acquired from three sub-regions of the sample under investigation 100. The sub-sample signals are in the time domain, but can 20 be Fourier transformed to yield frequency spectra 103a, 103b and 103c corresponding to the sub-regions of the sample under investigation 100. The above sub-division process is also applied to a reference sample 104. This produces three time-domain sub-sample signals 106a, 106b and 106c, and optionally three 25 frequency-domain signals 107a, 107b and 107c, corresponding to three sub-regions of the reference sample 104. It is not essential for the present invention to obtain signals from the reference sample 104 for every iteration of the method. Instead, previously-obtained reference signals may be stored and used again for one or more subsequent iterations. 30 As will be described in detail later, the reference sub-sample signals 106a-c are used in a deconvolution step to deconvolve each of the sub-sample signals 102a-c of the sample under investigation. This deconvolution step is shown schematically as 108a, 108b and 108c for each sub-region. Once deconvolved, the signals are summed together to

O

WO 2008/041866 PCT/NZ2007/000287 produce a signal 110 representing the time-domain NMR response of the sample under investigation 100 as a whole. For spectroscopy purposes, the signal 110 is Fourier transformed to produce a spectrum 112 representing the frequency-domain NMR response of the sample under investigation 100 as a whole. 5 The Imaging Step The imaging step comprises subjecting the sample under investigation, and where necessary the reference sample, to an inhomogeneous magnetic field, one or more Bj (RF) 10 pulses and one or more applied magnetic field gradients, and acquiring NMR sub-sample signals. Referring to the example sequence of Figure 2A, the magnitude of the inhomogeneous magnetic field can be labelled as B,, and its gradient as G 1 ,, the one or more applied magnetic field gradients are shown as a gradient pulse represented as G, and the excitation and detection of the NMR signal is represented as B,. 15 It can be seen that the B field includes pulses of 900 and of 1800, where this refers to the nutation angle of the magnetisation under the action of the B, pulse. It is noted that other turn angles are possible as variants on the preferred form. Following the example in Figure 2A, the 90' pulse excites the system and the subsequent NMR signal, which decays 20 in time due to the combined effects of spin-spin relaxation and magnetic field inhomogeneity, is a free-induction decay (FID). In one variant of the method the sampled signals are FID signals. In the variant of the method presented in Figure 2A, a 1800 pulse is applied following the gradient pulse such that the effects of field inhomogeneity are reversed and the NMR signal forms a so-called spin echo. Therefore 25 in this variant of the method the sampled signals are spin-echo type signals. The G, pulse sequence, which comprises one or more applied magnetic field gradients in the preferred form, is used to impart additional information to the NMR signal that can be of use in extracting a superior NMR spectrum. The magnetic field gradient of the 30 applied gradient pulse is defined as the derivative, along some axis (or axes), of the component of the magnetic field associated with the pulse which is parallel to the prevailing field direction of the static inhomogeneous magnetic field. Preferably, the axis (or axes) is that along which the inhomogeneous magnetic field is most inhomogeneous.

WO 2008/041866 PCT/NZ2007/000287 It is possible for the present invention to use only a single uniform imaging gradient, despite a prevailing non-uniform inhomogeneous field. It is preferable, however, that the single gradient has its field varying along the direction in which the background 5 inhomogeneous field varies most strongly. In one-sided-access NMR, this will typically be the direction normal to the surface of the magnet array. It should be noted that all that is necessary is to break the sample into sufficiently small regions of space so that the resolution in each such region is improved to such an extent that deconvolution can be successfully applied to each sub-sample. 10 The one or more applied magnetic field gradients are preferably pulsed and stepped in N separate equally-spaced intervals between negative and positive limits, with at least one separate time-domain NMR signal being acquired and stored for each of the N gradient pulse values. Fourier transformation of the data with respect to the variable gradient 15 value results in N separate NMR sub-sample signals corresponding to N separate spatial elements of the sample. It is, however, not essential to acquire N sub-sample signals corresponding to the N spatial elements. It is possible to acquire less than N sub-sample signals, provided at least two sub-sample signals are acquired. Also, it should be noted that the spatial elements need not and normally will not be of equal size or shape. 20 Alternate imaging methods use a train of spin echoes or gradient echoes along with applied magnetic field gradients in order to simultaneously spatially encode the signal and monitor the evolution of the signal due to the desired frequency spectrum, for example due to differences in chemical shift. Referring to Figure 2B, an example pulse sequence 25 for a method employing gradient echoes is shown. In this method, a sample is subjected to a gradient pulse, -G, for a fixed period of time, t,, following which the polarity of the gradient is switched to + G. The NMR signal is acquired as function of time, in the presence of G,. for a time period of 2*t,. During this time period, the spatially dependent de-phasing caused by the first negative gradient pulse is 'unwound' and an echo is formed. 30 Alternatively the negative and positive parts of the gradient may be of unequal magnitude, in which case the echo will be formed at a time such that the time integral of the gradient is zero. The Fourier transform of the echo signal is an 'image' of the sample in the direction of the gradient G. One example gradient echo method is echo planar WO 2008/041866 PCT/NZ2007/000287 spectroscopic imaging (EPSI). Note that the gradient echo method does not refocus the effects of either the inhomogeneity or the desired chemical shift information, therefore in gradient-echo based chemical shift imaging techniques, gradient echo signals (for a fixed value of G) are acquired with different offset times (with respect to the original RF 5 excitation pulse) so as to encode spectral information into each of the spatially-encoded echoes. In echo planar spectroscopic imaging (EPSI), the gradient echo step is done very fast so that following a single RF pulse (the 900 B, pulse in Figure 2B), all of the required gradient echoes are sampled one after another by switching gradients back and forth from positive to negative and acquiring during the positive and/or negative gradient pulses. 10 The one or more applied magnetic field gradients used in the present invention are preferably phase-encoding gradient pulses. Phase-encoding occurs where nuclear spins are permitted to evolve for a period of time in the presence of a gradient pulse, but where the NMR signal is acquired in the absence of the pulse. This can be seen in Figure 2A, 15 where the acquisition stage occurs in the absence of the gradient pulse. As an alternative to phase-encoding, frequency-encoding can be used; this requires the NMR signal to be acquired in the presence of the gradient pulse. The general method of encoding spatial as well as spectral .information into a 20 multidimensional NMR data set, with one or more spatial dimensions and one spectral dimension, is called chemical shift imaging. Such imaging is known from prior art such as US 6,320,381 to Hennig. It should be noted that such prior art chemical shift imaging methods are not the same as the method used in the present invention. The purpose of chemical shift imaging is to acquire spatially-resolved spectroscopic information, whereby 25 a pluraigo of spectra is acquired, each arising from one of the plurality of regions defined by the imaging dimensions. This is in contrast to the preferred form method of the present invention, which uses spatial encoding of spectral information as a step in acquiring a high-resolution spectrum from a single target region in the presence of magnetic field inhomogeneity. 30 The extraction of a plurality of spatially resolved sub-samples of an NMR signal from a single target region, which can be likened to a chemical shift imaging experiment, is in the present invention followed by the deconvolution of each sub-sample by previously WO 2008/041866 PCT/NZ2007/000287 acquired sub-samples of a reference signal (preferably having a high signal-to-noise ratio), and the summation of the resultant deconvolved sub-samples to yield a single, high resolution spectrum from the entire target region. The deconvolution step will now be described in greater detail below. 5 The Deconvolution Step The deconvolution step of the present invention is an improvement of a technique called reference deconvolution. Using reference deconvolution, one can gain information about 10 the magnetic field inhomogeneity based on a reference NMR signal obtained from a sample whose spectrum is known precisely. In the case of proton NMR, which is most relevant to one-sided-access NMR, the reference sample would normally be water. Having obtained the reference NMR signal, the signal obtained from an unknown sample 15 can be deconvolved so as to remove that .part of the spectral line-shape due to magnetic field inhomogeneity; this reveals the desired spectrum arising from the chemical shifts of the unknown sample. Deconvolution in the spectral or frequency domain can be performed in the Fourier or 20 time-domain by dividing the unknown sample time-domain signal by the reference sample time-domain signal. Such signal division makes for signal-to-noise ratio difficulties at longer times when the reference sample signal has decayed to the level of the noise. However, the reference signal may be acquired once using a large number of acquisitions, optimising the available signal-to-noise ratio for this reference. 25 Reference deconvolution in the prior art (herein 'raw deconvolution') is carried out by deconvoluting an NMR signal representing the entirety of the sample under investigation with an NMR .signal representing the entirety of the reference sample. In contrast, the deconvolution used in the present invention (herein 'imaged deconvolution') involves 30 deconvoluting two or more sub-sample signals acquired from sub-regions of the sample under investigation and the reference sample, and summing together the deconvolved signals. As noted earlier, the process of obtaining NMR sub-sample signals from two or WO 2008/041866 PCT/NZ2007/000287 more sub-regions is done in the imaging step. Where there are N sub-regions, N separate NMR sub-sample signals are typically obtained. The imaged deconvolution technique of the preferred form method is preferably used 5 where it is necessary to reduce an inhomogeneous line-width by more than a factor of 7. This will almost always be the case in practical efforts to improve spectral resolution of NMR performed with an inhomogeneous field. Typical line-widths will at best be 100 ppm, and the typical desired resolution will be in the order of 1 ppm, sufficient to resolve water and oil (3.5 ppm) in the proton NMR spectrum. 10 The improvements offered by imaged deconvolution over raw deconvolution are best conveyed by way of experimental results. One non-limiting example experiment and its results are noted below. 15 Example An experiment was conducted using a laboratory spectrometer in which a highly inhomogeneous magnetic field was deliberately added by placing in the vicinity of the radio frequency (RF) coil (at a distance of 57 mm) a coil of 38 mm diameter and 220 turns 20 through which a current of 20 amperes was passed. This was sufficient to broaden the NMR spectrum to approximately 70 ppm on a 400 MHz spectrometer. Raw deconvolution was tested by first obtaining an NMR signal from a sample of water and a sample of ethanol, where water is used as the reference sample and ethanol as the 25 sample under investigation. The NMR signals from the ethanol and water samples were then Fourier transformed to obtain the severely line-broadened NMR spectra shown in Figures 3A and 3B respectively. A deconvolved signal was obtained by dividing the NMR time-domain signal from ethanol by the equivalent time-domain signal from water. Due to the severity of the line broadening, the deconvolved signal yields a spectrum dominated 30 by noise, as shown in Figure 4. As for the imaged deconvolution technique of the present invention, NMR signals were obtained from both ethanol and water using the pulse sequence of Figure 2A. A total of WO 2008/041866 PCT/NZ2007/000287 128 phase-encoding steps were used with a maximum phase gradient of 58.2 mT/m. The same total acquisition times were used in the imaged deconvolution experiments as in the raw deconvolution experiments. The water signal was acquired in 3.5 hours (to improve its signal-to-noise ratio) and the ethanol signal was acquired in 25 minutes. 5 The signal acquired following the application of the pulse sequence is shown in Figures 5A-D. Figures 5A and 5B show the time-domain signals for ethanol and water respectively at each different phase gradient (k-space) value. Figures 5C and 5D show the result of Fourier transforming with respect to k in order to obtain separate time-domain 10 signals for each separate sub-region of space for ethanol and water respectively. Under imaged deconvolution, the ethanol signals from each spatial element or sub-region are divided by the corresponding water signals, and the resulting time-domain signals are subsequently summed to yield a deconvolved time-domain signal for the ethanol sample as a whole. As seen in Figure 6, a Fourier transform of this time domain signal yields a 15 high-resolution spectrum in which the three chemical-shifted NMR resonances of ethanol are well resolved (to better than 1 ppm). Figures 7A to 7C show a direct comparison between the ethanol spectrum obtained through the imaged deconvolution method of the preferred form of the present invention 20 (Figure 7C) with the original ethanol spectrum acquired in the inhomogeneous field to which no deconvolution has been applied, (Figure 7A) and the result of the raw deconvolution of the water and ethanol signals (Figure 7B). It is clear that the spectrum obtained through imaged deconvolution is superior to the spectrum obtained through prior art techniques. 25 Advantages of the Preferred Form Method Some of the specific advantages of the preferred form method are noted below. The advantages are advantages of certain forms of the method of the present invention; it is 30 not necessary for all forms of the method to provide the stated advantages. 1 A WO 2008/041866 PCT/NZ2007/000287 One-Sided-Access NMR The preferred form method provides a particular advantage when used with one-sided access NMR apparatuses, such as that described in the present applicant's PCT 5 Publication WO 2004/008168. The advantage, which makes deconvolution particularly effective, is based on the characteristics of the inhomogeneous field region that are the same for all samples of size larger than the sensitive volume of the apparatus, provided that the same magnet system is used and the same RF pulse excitation conditions are used, thus ensuring that the same region of sample (sensitive volume), is excited. This 10 makes it possible to obtain the reference signal with arbitrarily high signal-to-noise ratio, by simply acquiring the reference signal data for a very long time. This reference signal, once acquired, can be stored and used for all subsequent deconvolutions. Avoiding the Needfor Shimming 15 As noted in the Background, one approach to achieving magnetic field homogeneity is by shining the magnetic field, which involves adding adjustable magnetic field gradients that oppose and cancel any undesired static magnetic field gradient giving rise to inhomogeneity. In order to be effective, shimming requires gradients that are perfectly 20 matched to the field inhomogeneity profile. A field with predominantly non-linear inhomogeneities cannot be shimmed with any superposition of linear magnetic field gradients. The preferred form method proposed here does not require any correlation between the 25 inhomogeneities present and the applied. magnetic field gradients. It only requires that the sample be 'broken' into smaller pieces. The particular characteristics of the inhomogeneous field are only relevant in that it is preferable to apply the magnetic field gradients in the direction of the greatest degree of inhomogeneity across the sample. 30 The foregoing describes the invention including preferred forms thereof. Alterations and modifications as will be obvious to those skilled in the art are intended to be incorporated within the scope hereof, as defined by the accompanying claims. 1l''

Claims (15)

1. A method of extracting NMR spectra from a sample under investigation, the method comprising: 5 subjecting the sample under investigation to an inhomogeneous static magnetic field, one or more RF pulses, and one or more applied magnetic field gradients to acquire NMR sub-sample signals from two or more regions of the sample under investigation; subjecting a reference sample to the inhomogeneous static magnetic field, the one or more RF pulses, and the one or more applied magnetic field gradients to acquire.NMR 10 sub-sample signals from two or more regions of the reference sample; deconvolving the NMR sub-sample signals acquired from regions- of the sample under investigation with the NMR sub-sample signals acquired from corresponding regions of the reference sample to produce a deconvolved signal for each region; and summing together the deconvolved signals for each region to obtain the NMR 15 spectra of the sample under investigation.

2. The method of claim 1 where subjecting the sample under investigation to an inhomogeneous static magnetic field comprises placing the sample in the test region of an NMR spectrometer system where the homogeneity of the magnetic field is insufficient to 20 obtain the desired spectral resolution.

3. The method of claim 2 wherein the NMR spectrometer system is a one-sided access NMR apparatus. 25

4. The method of any one of the preceding claims wherein the NMR sub-sample signals are acquired as spin-echoes.

5. The method of any one of claims 1 to 3 wherein the NMR sub-sample signals are acquired as free induction decays (FIDs). 30

6. The method of any one of the preceding claims wherein the one or more applied magnetic field gradients comprise one or more phase-encoding gradients stepped through N discrete values. WO 2008/041866 PCT/NZ2007/000287

7. The method of any one of claims 1 to 5 wherein the one or more applied magnetic field gradients comprise frequency-encoding gradients. 5

8. The method of any one of the preceding claims wherein subjecting the sample under investigation to the one or more applied magnetic field gradients comprises using a spin echo method.

9. The method of any one of claims 1 to 7 wherein subjecting the sample under 10 investigation to the one or more applied magnetic field gradients comprises using a gradient echo method.

10. The method of claim 9 wherein the gradient echo method is echo planar spectroscopic imaging (EPSI). 15

11. The method of any one of the preceding claims wherein deconvolving the NMR sub-sample signals comprises dividing time-domain NMR sub-sample signals acquired from the sample under investigation by corresponding time-domain NMR sub-sample signals acquired from the reference sample. 20

12. A method of extracting NMR spectra from a sample under investigation, the method comprising: subjecting the sample under investigation to an inhomogeneous static magnetic field, one or more RF pulses, and one or more applied magnetic field gradients to acquire 25 NMR sub-sample signals from two or more regions of the sample under investigation; deconvolving the NMR sub-sample signals acquired from regions of the sample under investigation with NMR sub-sample signals from corresponding regions of a reference sample to produce a deconvolved signal for each region; and summing together the deconvolved signals for each region to obtain the NMR 30 spectra of the sample under investigation.

13. A method of extracting NMR spectra from a sample under investigation, the method comprising: WO 2008/041866 PCT/NZ2007/000287 receiving two or more NMR sub-sample signals from two or more regions of the sample under investigation; deconvolving the NMR sub-sample signals with NMR sub-sample signals from corresponding regions of a reference sample to produce a deconvolved signal for each 5 region; and summing together the deconvolved signals from each region to obtain the NMR spectra of the sample under investigation.

14. A nuclear magnetic resonance spectra extraction system, the system configured to: 10 subject a sample under investigation to an inhomogeneous static magnetic field, one or more RF pulses, and one or more applied magnetic field gradients to acquire NMR sub-sample signals from two or more regions of the sample under investigation; subject a reference sample to the inhomogeneous static magnetic field, the one or more RF pulses, and the one or more applied magnetic field gradients to acquire NMR 15 sub-sample signals from two or more regions of the reference sample; deconvolve the NMR sub-sample signals acquired from regions of the sample under investigation with the NMR sub-sample signals acquired from corresponding regions of the reference sample to produce a deconvolved signal for each region; and sum together the deconvolved signals for each region to obtain the NMR spectra 20 of the sample under investigation.

15. A nuclear magnetic resonance spectra extraction system comprising: sample imaging means for subjecting a sample under investigation to an inhomogeneous static magnetic field, one or more RF pulses, and one or more applied 25 magnetic field gradients to acquire NMR sub-sample signals from two or more regions of the sample under investigation, and for subjecting a reference sample to the inhomogeneous static magnetic field, the one or more RF pulses, and the one or more applied magnetic field gradients to acquire NMR sub-sample signals from two or more regions of the reference sample; 30 deconvolution means for deconvolving the NMR sub-sample signals acquired from regions of the sample under investigation with the NMR sub-sample signals acquired from corresponding regions of the reference sample to produce a deconvolved signal for each region; and 10O WO 2008/041866 PCT/NZ2007/000287 summing means for summing together the deconvolved signals for each region to obtain the NMR spectra of the sample under investigation.