Method for generating representation of properties of substance

Abstract

A method for generating a property representation of a substance is described. The method comprises: obtaining portions of one or more nuclear magnetic resonance (NMR) data of the substance, where each NMR data portion comprises one or more NMR datasets, each NMR dataset comprising intensity data and relaxation data; processing each NMR data portion into a multi-dimensional distribution of a fitting data set by performing a plurality of fits of a function, where the function includes at least 3 fitting parameters; and generating a characteristic representation of the substance from the fitting data set. The processing includes applying at least one initial guess value. Each set of fitting data comprises data embedded with at least one subset of fitting data comprising an apparent longitudinal relaxation time (T1 *), an associated apparent transverse relaxation time (T2 *) and an associated intensity (M0), preferably said function being a function of a model representing the obtained NMR data portion.

Description

Technical Field

[0001] The present invention relates to a method and system for generating a property representation (e.g., a multidimensional spectrum, such as a T1-T2 spectrum) of a substance (e.g., a complex substance, such as a food product, a substance containing chemical contaminants, or a manure slurry). Background Technology

[0002] Substances, especially complex substances containing multiple elements, such as fats and / or water and / or proteins and / or nucleic acids and / or carbohydrates and / or vitamins and / or minerals and / or chemical contaminants and / or nutrients, where the chemical species of each element interacts with other chemical species in the substance, are typically characterized in a laboratory setting using expensive equipment and time-consuming methods such as chromatography.

[0003] In many applications, material characterization is used and is often required. One example is the food industry, where food companies are required to label the nutrients in their products and therefore must frequently characterize their products in terms of composition, or monitor changes in continuous streams. Another example is the conversion of biomass to circular energy, where it is desirable to characterize the slurry during the biomass-to-energy conversion process to identify elemental distributions and thus control the process. A third example is the chemical industry, where companies use chemical processes to convert organic and inorganic raw materials into chemicals, and where the characterization of chemicals and intermediates is crucial for process control and ensuring product quality.

[0004] Nuclear magnetic resonance (NMR) has generally proven to be an excellent non-invasive technique for studying microscopic molecular interactions, in which the Laplace transform is often used to convert NMR echo data into a relaxation time distribution.

[0005] WO2018 / 163188A1 discloses a method for characterizing the chemical and / or morphological features of a material, comprising: acquiring energy relaxation data from ¹H low-field nuclear magnetic resonance (¹H LF-NMR) measurements of the material; converting the relaxation signal into a multidimensional distribution of longitudinal and transverse relaxation times by solving an inverse problem under L1 and L2 regularization and further applying non-negative constraints; and identifying one or more properties of the material by means of the multidimensional T1-T2 distribution.

[0006] There is a substantial need for alternative or improved methods for characterizing substances, particularly highly efficient methods and / or methods that can generate valuable information about substances, even for complex substances with complex compositions. Summary of the Invention

[0007] One object of the present invention is to provide a method for characterizing substances to provide valuable information about the chemical, physical, biological and / or morphological characteristics of substances.

[0008] In one embodiment, the aim is to provide a method for generating a property representation of a substance, wherein the property representation describes the property characteristics of one or more substances, preferably relating to the contents of the substance and / or its components and / or chemical and / or physical and / or biological effects and / or interactions between components or component fragments of the substance.

[0009] In one embodiment, the aim is to provide a method for generating a property representation of a substance, wherein the property representation describes the substance in the form of a description of the substance or data representing a description of the substance.

[0010] In one embodiment, the aim is to provide a method for generating a property representation of a substance, wherein the property representation describes at least one mass characteristic of the substance, preferably wherein the property representation includes a description of the substance or data representing a mass parameter description of the substance, or a marker relating to changes in the substance during processes.

[0011] In one embodiment, the objective is to provide a method for generating a representation of the properties of a substance, wherein the representation of properties is described.

[0012] In one embodiment, the aim is to provide an NMR system for performing methods to characterize substances, thereby providing valuable information about the chemical, physical, biological, and / or morphological characteristics of the substances.

[0013] In one embodiment, an objective is to provide a method for generating a trained computer for performing a process of processing a characteristic representation of a generated or produced substance in the method of the present invention.

[0014] These and other objectives have been achieved through the present invention or its embodiments as defined in the claims and / or as described below.

[0015] It has been found that the present invention or its embodiments have many additional advantages, which will become clear to those skilled in the art from the following description.

[0016] The method of the present invention includes a method for generating a property representation of a substance, the method comprising:

[0017] - Record nuclear magnetic resonance (NMR) data for at least one sample of the substance.

[0018] - Generate one or more portions of nuclear magnetic resonance (NMR) data from the recorded NMR data, wherein each NMR data portion is generated as comprising one or more NMR datasets.

[0019] - Each NMR data portion is processed to fit a multidimensional distribution of the dataset, the processing including selecting a function and performing multiple fits on the selected function, and

[0020] - Generate a property representation of the substance from the fitted dataset.

[0021] Each dataset advantageously includes associated intensity and relaxation data. The intensity and relaxation data can be correlated because they originate from a common isotope.

[0022] The selected function includes at least three fitting parameters.

[0023] The processing of each NMR data segment includes selecting and applying at least one initial guess value.

[0024] In the following text, the NMR data section and the NMR dataset are also referred to as input NMR data.

[0025] The function includes at least three fitting parameters, and is preferably a function representing a model of a portion of the obtained NMR data.

[0026] The function is also called the model function or model equation, for example, y = the function.

[0027] The processing includes applying at least one initial guess. Each fitted dataset includes data embedded with at least one subset of fitted data, the subset including fitted dataset parameters containing apparent longitudinal relaxation time (T1*), associated apparent lateral relaxation time (T2*), and associated intensity (M0).

[0028] Each NMR data portion advantageously includes one or more datasets for a common NMR readable isotope. In one embodiment, the one or more NMR data portions include data for the same or different isotopes. In one embodiment, the one or more NMR data portions include data for one or more readable isotopes interacting with other isotopes.

[0029] The phrases “NMR readable” and “NMR active” are used interchangeably in this article.

[0030] The inventors have discovered that even when the substance is a complex substance comprising multiple components that may interact, influence, or even react with each other, the method of the present invention provides a highly effective tool for performing chemical, physical, biological, and / or morphological analyses of the substance. Therefore, the method can provide a very detailed analysis of the substance, which can be shown, embedded in, or even directly revealed from a characteristic representation. Thus, in one embodiment, the characteristic representation can even provide unique visual characteristics, such as bioequivalents or similar features of the substance, such as fingerprints, for example, compositional fingerprints, such as mass fingerprints.

[0031] The fingerprint may advantageously include features of a combination of elements and / or compounds of the substance, which contribute to the overall identification and / or quality of the substance.

[0032] The phrase "quality fingerprint" is used herein to refer to a unique set of characteristics, parameters, and / or properties that define the quality of a substance. Therefore, the method of the present invention provides a valuable tool for examining substances, such as for comparison with reference substances having known chemical, physical, biological, and / or morphological characteristics, for determining whether quality parameters are met, for assessing compatibility, for assessing reactivity, or for assessing the response factor to the substance.

[0033] Therefore, it has been found that the method of the present invention provides a valuable tool for production (e.g., for quality control), such as at least one quality parameter and / or quality fingerprint of raw materials, precursor materials and / or production materials (such as food substances), for example as part of process control.

[0034] The method of the present invention has been found to be a useful and effective tool for the rapid and efficient detection of the potential presence of one or more impurities, for example in raw materials or final products.

[0035] In one embodiment, the method can be used in forensic medicine, for example, to analyze complex substances found at a crime scene, or the method can be applied to analyze and generate a characterization of fluid samples of biological organisms, for example, to compare with a characterization reference substance found at a crime scene.

[0036] It should be emphasized that when the term “including / comprises” is used in this document, it should be interpreted as an open-ended term, that is, it should be regarded as indicating the presence of the specifically stated features, such as elements, units, integers, steps, components and combinations thereof, but does not exclude the presence or addition of one or more other features.

[0037] Throughout the specification or claims, unless the context otherwise indicates or requires, the singular encompasses the plural, and the plural encompasses the singular.

[0038] The method of this invention can be referred to as a computer-implemented invention because the computer system can be advantageously applied, at least, to perform multiple fittings of the selected function. The term "wt%" in this document refers to weight percentage.

[0039] "An embodiment" should be interpreted as including inventive examples that incorporate the features of the embodiment.

[0040] The term "substantially" should be considered herein to cover common product variations and tolerances. All features of this invention and its embodiments, including the scope and preferred scope, may be combined in various ways within the scope of this invention unless there is a specific reason not to combine such features.

[0041] Unless otherwise stated, any properties, ranges of properties and / or determination and / or analytical conditions are given or provided at 1 atmosphere and 37°C.

[0042] All features of the present invention and its embodiments, including the scope and preferred scope, may be combined in various ways within the scope of the present invention unless there is a specific reason not to combine such features.

[0043] Nuclear magnetic resonance (NMR) spectroscopy is a well-known analytical technique used for the structural resolution of both small and large molecules, as well as for quantitative and qualitative analysis. This invention provides an analytical tool that can produce characterizations of substances that may contain several different molecules and components, and the method can provide characterizations of the entire substance, not just individual molecules.

[0044] By performing multiple fittings on the function, at least one fitted dataset can be obtained, wherein each fitted dataset preferably includes multiple fitted data subsets embedded therein, which are collected to provide a property representation of the substance, such as data describing the property representation of the substance and / or data graphs including the property representation of the substance.

[0045] Performing multiple fittings on the function can advantageously include performing multiple regression fittings on the function. Advantageously, performing multiple fittings on the function includes performing multiple fittings on the function for each of the NMR datasets, for example, at least twice, for example, at least four times, for example, from five to fifty times, wherein the multiple fittings advantageously include multiple regression fittings on the function.

[0046] Those skilled in the art will understand that each individual fit may contain some uncertainty. However, it has been found that when multiple fits are performed, the statistical uncertainty of a single component may become very small or even practically insignificant. This allows for the generation of characteristic representations with very high accuracy, even to the point that the resulting representations constitute accurate and reproducible identification characteristics, such as material fingerprints.

[0047] Methods for generating NMR data, such as NMR intensity data, NMR frequency data, and NMR relaxation data, are known in the art and can be described, for example, as in WO2022 / 117170.

[0048] The method includes recording nuclear magnetic resonance (NMR) data of at least one sample of the substance. Recording NMR data may include performing multiple readouts, which may be the same or different from each other, as further described below.

[0049] One or more portions of nuclear magnetic resonance (NMR) data that generate the recorded NMR data may include deriving datasets from the recorded NMR data and generating one or more portions of NMR data based on the derived datasets, wherein each dataset is generated to include associated intensity data and relaxation data.

[0050] In one embodiment, each NMR dataset includes frequency data, such as at least one of resonance frequency, chemical shift, quadrupole coupling, J coupling, or dipole-dipole coupling.

[0051] The selection of desired resonance frequency data for an NMR dataset can preferably be performed based on the isotopes to which the NMR dataset is applied. Frequency data may be particularly advantageous when the applied isotopes comprise many different chemical or physical configurations of the substance.

[0052] Advantageously, each NMR dataset includes at least one of inversion recovery data or echo NMR data. Preferably, each NMR dataset includes data representing at least one inversion recovery delay (IRD) and / or data representing at least one echo delay (ED). It has been found that inversion recovery delay data and / or echo delay data and / or any combination thereof can increase the distinguishability between material composition types, such as fatty and non-fatty components.

[0053] In one embodiment, the echo data includes at least one of spin echo data, fast spin echo data, stimulated echo data, or gradient echo data. Such echo data can be obtained, for example, by the method described by Hahn in "Spin Echo".

[0054] Each NMR dataset can conveniently include data recorded using pulse sequences, which include at least one period in which the recorded data is affected by T1 (longitudinal) relaxation and / or at least one period in which it is affected by T2 (lateral) relaxation.

[0055] In one embodiment, each NMR dataset includes NMR data recorded using pulse sequences, the pulse sequences including at least one period in which the recorded data is affected by chemical exchange. The NMR dataset may conveniently include frequency data, the NMR dataset comprising NMR data recorded using pulse sequences, the pulse sequences including at least one period in which the recorded data is affected by chemical exchange.

[0056] In one embodiment, each NMR dataset includes NMR data recorded by at least one readout process, the readout process including a pulse sequence having at least one period in which the recorded data is affected by diffusion, such as molecular diffusion, ion diffusion, or spin diffusion.

[0057] When the dataset of input NMR data includes multi-component relaxation time or diffusion data, its analytical potential can provide valuable information for characterizing substances.

[0058] Based on the evolution of spin systems, and by applying datasets that include two or more relaxation times and / or associated diffusion data (e.g., their diffusion coefficients (D)), valuable characteristics dependent on the molecular environment and dipole interactions of matter, as well as the molecular species and specific dynamics of the matter's composition, can be provided. In addition to temporal resolution, information related to chemical exchange and molecular transfer of matter from one site to another can be enhanced, resulting in a more detailed representation of the properties and providing additional information about the structural environment of the matter's composition.

[0059] Advantageously, one or more NMR data portions include at least two NMR datasets, such as at least three, four, eight, sixteen, or more NMR datasets. On the one hand, the more datasets, the more accurate or detailed the characteristic representation obtained. On the other hand, the more datasets, the more data processing may be required to represent the characteristics of the generating substance.

[0060] At least one NMR dataset may advantageously include data from a common or identical pulse sequence. In one embodiment, one or more NMR datasets—or even all NMR datasets—may include intensity data from a common or identical pulse sequence and relaxation data from different pulse sequences.

[0061] In one embodiment, the data in at least one dataset is or includes data from a common or identical group of pulse sequences, which includes one or more sequences among inversion recovery pulse sequences, Carr-Purcell-Meiboom-Gill (CPMG) sequences, and / or spin echo sequences with one or more echoes.

[0062] Optionally, at least one NMR data portion of the dataset includes data from a common or identical pulse sequence group.

[0063] The input NMR data may include NMR data determined on any NMR-active isotope containing an NMR-active nucleus, preferably on one or more isotopes each having a relatively high abundance and / or containing an NMR-active nucleus (with a sensitivity of at least 0.01%, for example, at least 1%, relative to ¹H).

[0064] Advantageously, at least one NMR dataset is based on data from a common NMR readable isotope, and preferably, at least one NMR dataset portion is based on data from a common NMR readable isotope.

[0065] It should be noted that different NMR data segments can advantageously include data from different isotopes.

[0066] In one embodiment, processing of each of two or more NMR data portions includes performing multiple fittings of the formula separately for each NMR data portion.

[0067] In one embodiment, processing of each of two or more NMR data portions includes simultaneously and in combination performing the multiple fitting on more than one or even all NMR data portions.

[0068] In one embodiment, each NMR data fraction is processed separately from the processing of other NMR data fractions. The corresponding processing of a given NMR data fraction preferably includes fitting to the same formula. In another embodiment, the corresponding processing of a given NMR data fraction may include fitting to different formulas.

[0069] In one embodiment, processing for each of the NMR data portions includes performing multiple fittings of the formula to one or more datasets of the NMR data portions during a common fitting process.

[0070] Performing multiple fittings on a portion of NMR data can advantageously include applying the at least one initial guess value to at least one of the multiple fittings. Preferably, performing multiple fittings on a portion of NMR data includes applying the at least one initial guess value to multiple of the multiple fittings, such as all fittings in the multiple fittings, wherein the at least one initial guess value may be the same or different.

[0071] The corresponding one or more initial guesses may include specific values ​​or ranges of values. The specific values ​​and / or ranges of values ​​may be selected, for example, as specific values ​​and / or ranges of values ​​with high probability, and may optionally be obtained from previous fitting of the formula.

[0072] In one embodiment, performing multiple fittings on a portion of NMR data includes applying a first initial guess to at least one fitting parameter of a first fitting in multiple fittings and applying a subsequent initial guess to the at least one fitting parameter of a subsequent fitting, wherein the subsequent initial guess of the at least one fitting parameter is in the form of the average of the fitting results of the at least one fitting parameter of the previous fittings in the multiple fittings, or in the form of a fitting result of the at least one fitting parameter of the previous fittings in the multiple fittings.

[0073] Advantageously, the initial guess value includes an initial guess value for at least one of the at least three fitting parameters for at least one of the multiple fits. Preferably, the initial guess value includes an initial guess value for at least one of the at least three fitting parameters for multiple of the multiple fits (e.g., for all fits in the multiple fits), wherein the initial guess value may be the same or different from one fit to subsequent fits depending on the selected pattern.

[0074] Optionally, the initial guesses include initial guesses for two or more of the at least three fitting parameters for at least one of the multiple fits. This is particularly relevant when the initial guesses for the at least one fitting parameter comprise an interval.

[0075] In one embodiment, the at least one initial guess includes at least one of a random initial guess or an initial guess generated by applying a bootstrap method. The bootstrap method may, for example, involve creating simulated data based on an original dataset. The original dataset may, for example, be a plurality of previously determined fitting parameters, or fitting parameters derived from an estimated range of possible or feasible fitting parameters. Examples of bootstrap techniques include Bayesian bootstrap and parametric bootstrap.

[0076] The at least one initial guess can be applied to the fitting parameters in multiple fittings, for example, to each fitting in multiple fittings of the function. In one embodiment, the initial guess applied to the at least one fitting parameter is the same initial guess in at least two or more fittings in multiple fittings of the function, for example, the same initial guess in each fitting in multiple fittings of the function. This embodiment may be preferred when there are many fitting parameters (e.g., four or more, such as five or more fitting parameters) to optimize the uniqueness of the information content / relative to the required computational resources.

[0077] In one embodiment, the input NMR data is generated as comprising two or more NMR data portions, and the processing of a corresponding NMR data portion among the plurality of NMR data portions may advantageously include applying a randomly selected initial guess value to at least one fitting parameter for each of the NMR data portions.

[0078] In one embodiment, the at least one initial guess value is characterized in that it is a Monte Carlo type random initial guess value.

[0079] In one embodiment, the at least one initial guess value can be selected as a starting range. For example, the initial guess value can be a starting range represented by a span of selected values, wherein the span of selected values ​​includes typical expected values ​​and / or estimated values ​​or consists of typical expected values ​​and / or estimated values. The at least one initial guess value can include an initial guess value for each of one or more fitting parameters.

[0080] The at least one initial guess can be conveniently selected as an initial guess for a fitting parameter associated with at least one desired fitting parameter or at least one fitting parameter from a previously fitted dataset. In one embodiment, the at least one initial guess can be selected as an initial guess for a fitting parameter associated with a subset of desired fitted data or a subset of previously fitted data.

[0081] Advantageously, the initial guess value is based on or includes a starting range within a selected value span, wherein the selected value span includes or consists of typical expected values, wherein the typical expected values ​​are determined based on the typical values ​​of the at least one expected or previously fitted parameter.

[0082] In one embodiment, the at least one initial guess is selected as an initial guess for a fitting parameter associated with at least one parameter of a subset of fitted data, wherein the at least one fitted dataset parameter is selected from apparent intensity (M0), T1* and T2*, chemical exchange, frequency, or diffusion parameters of species in the sample.

[0083] In one embodiment, the at least one initial guess is selected as an initial guess for a fitting parameter, which is related to a derivable parameter that can be derived from at least one parameter of the one or more fitting datasets.

[0084] In one embodiment, the at least one initial guess value is related to the T1 or T2 of the substance.

[0085] The function can be selected based on the complexity of the substance and / or the required level of detail in representing its properties. The complexity of the substance can be determined based on the number of different components and / or the interactions and / or influences between components or parts thereof (e.g., fragments or molecules). Advantageously, the function is a nonlinear function.

[0086] In one embodiment, the function includes one or more exponential terms. In one embodiment, the function includes a multilinear function with two or more exponential terms (e.g., up to five exponential terms, such as two or three exponential terms). In one embodiment, the function includes one or more hyperbolic terms.

[0087] In one embodiment, the function includes one or more triangular terms.

[0088] In one embodiment, the function includes one or more terms selected from logarithmic terms, sine terms, polynomial terms, or differential terms.

[0089] In one embodiment, the function includes a combination of at least one exponential term and one or more other terms, such as one or more terms selected from logarithmic, sine, polynomial, or differential terms.

[0090] The function can, in principle, have any number of fitting parameters. To obtain a characteristic representation with the desired level of detail, the number of fitting parameters should not be too low. To ensure that the required computational resources are not excessive and / or the processing time is not excessive, the number of fitting parameters should not be too high. Advantageously, for each spin type, the function includes 3-10 fitting parameters, for example, 4 fitting parameters, 5 fitting parameters, 6 fitting parameters, 7 fitting parameters, 8 fitting parameters, or more.

[0091] In one embodiment, the function includes function FI

[0092]

[0093] Where A i B i C i D i and E i These are the fitted parameters, and IRD j TE k,l It is input data from the corresponding NMR dataset;

[0094] i is one of the n different spin types;

[0095] IRD j It is the inversion recovery delay applied when obtaining j corresponding NMR datasets from a total of m different inversion recovery delays;

[0096] TE k,l is the echo time, where k represents the k-th length of each echo from 1 to p in the echo string, and l is the number of echoes from 1 to q in the corresponding echo string;

[0097] Where n, m, p, and q are each an integer of 1 or higher, for example, at most 100, for example, from 1 to 50, for example, from 4 to 25.

[0098] The term "spin type" or spin type is used herein to refer to a spin that differs from other spins in at least one parameter, such as a different spin in at least one of nuclear spin angular momentum, spin angular momentum magnitude, chemical shift, spin-spin coupling, spin-spin splitting, spin-lattice relaxation, spin-spin relaxation, or spin-spin relaxation, such as a different spin in at least one of T1*, T2*, intensity M0, frequency (e.g., resonance frequency, chemical shift, quadrupole coupling, J coupling, or dipole-dipole coupling), chemical exchange, diffusion, or other spin parameters, such as spin parameters that can be determined by the methods of one or more embodiments of the present invention.

[0099] In one embodiment, the function is function FI:

[0100] .

[0101] As can be seen, the function FI has two exponential terms.

[0102] In one embodiment, the function is function FII:

[0103]

[0104] Where A i B i C i D i and E i These are the fitted parameters, and IRD j TE k,l and RD h It is input data from the corresponding NMR dataset;

[0105] i is one of the n different spin types;

[0106] IRD j It is the inversion recovery delay applied when obtaining j corresponding NMR datasets from a total of m different inversion recovery delays;

[0107] TE k,lis the echo time, where k represents the k-th length of each echo from 1 to p in the echo string, and 1 is the number of echoes from 1 to q in the corresponding echo string;

[0108] RD h It is a cyclic delay during which the magnetization recovers or partially recovers from the h-th cyclic delay of 1 to r cyclic delays from the run of one NMR recording to the run of a subsequent NMR recording;

[0109] Where n, m, p, q and r are each an integer of 1 or higher, for example, at most 100, for example, from 1 to 50, for example, from 4 to 25.

[0110] The phrase “partial recovery” advantageously means that the magnetization is restored by at least 5%, for example at least 20%, or for example at least 50%.

[0111] A formula that has been proven suitable for generating desired high-quality characteristic representations is the function FII specified below:

[0112]

[0113] As can be seen, the function FII has three exponential terms.

[0114] The intensity data of the corresponding NMR dataset advantageously includes echo time, and the relaxation data of the corresponding NMR dataset includes inversion recovery delay, preferably the corresponding NMR dataset also includes cyclic delay.

[0115] In one embodiment, parameter C i The T1 or T1* subset of fitted data is associated with the echo data representing spin type i among n spin types.

[0116] Parameter D i It can be correlated with T2 or T2* of the fitted data subset representing the echo data of spin type i out of n spin types. Parameter A i and B i One or both of them can be related to the intensity M0 and / or chemical exchange value of the fitted data subset associated with the echo data of spin type i in n spin types.

[0117] Advantageously, each subset of fitted data includes the factor Beta. i Beta i = -1 / A i A factor, i, is used for spin type i among the n spin types in each fitted dataset, representing chemical exchange.

[0118] In one embodiment, each subset of fitted data includes an M0 factor, where M0 i=-A i *B i The factor, used for spin type i among the n spin types in each fitted dataset, represents the strength of the spin type.

[0119] The M0 factor can be intensity.

[0120] In one embodiment, T1* is based on T1* i =-1 / C i Factor-determined spin type i from n spin types for each fitted dataset, where T1* i T1* represents spin type i.

[0121] In one embodiment, T2* is based on T2* i =-1 / D i Factor-determined spin type i from n spin types for each fitted dataset, where T2* i T2* represents spin type i.

[0122] E for spin type i among n spin types for each fitted dataset i Factors can advantageously represent diffusion parameters of molecules, fragments, or ions containing spin type i.

[0123] The number of spin types n is advantageously 2 or greater, for example, up to 1000. Advantageously, the number of spin types n is from 3 to 100.

[0124] Advantageously, each subset of fitted data represents a spin-type property, preferably including at least T1*, associated T2*, and associated intensity (M0). Preferably, each subset of fitted data represents a spin-type property, which also includes one or more of diffusion parameters, frequency shift, chemical exchange, J-coupling, quadrupole coupling, or dipole-dipole coupling.

[0125] In one embodiment, generating a property representation of a substance from the fitted dataset includes: deriving representation components from each fitted dataset to obtain a plurality of representation components, and combining the plurality of representation components to generate a property representation of the substance, the representation components including the at least one subset of fitted data containing T1*, associated T2*, and associated intensity (M0).

[0126] Feature representations can be continuously improved by adding additional representation components to them.

[0127] Therefore, the characteristic representation can be generated over time as the corresponding NMR data portion or NMR dataset is processed.

[0128] Therefore, when a user aims to identify a specific sub-property of a substance, and that sub-property is realized in the construction of the property representation, the construction of the property representation can be terminated, and the generation of the property representation is considered complete. This saves time and improves the efficiency of the method.

[0129] In one embodiment, the T1*, associated T2*, and associated intensity (M0) of the at least one subset of fitted data may include data representing the spin diffusion characteristics of molecular or ionic species in a non-uniform static field or radio frequency field.

[0130] In one embodiment, the T1*, associated T2*, and associated intensity (M0) of the at least one subset of fitted data may include data representing chemical exchange properties involving one, two, or preferably multiple spin types in the material, such as 20 or more, 50 or more, 5 to 100 or even more spin types.

[0131] The characteristic representation can advantageously include probability diagrams of the corresponding spin types T1* and T2*. Such probability diagrams can be provided in the form of visual probability diagrams or in the form of data representing such probability diagrams.

[0132] In one embodiment, the characteristic representation is in the form of a depiction or includes a depiction, preferably showing some or all of the derived characteristics of n spin types. The depiction can be selected according to the intended use of the method. For example, when the method is used to determine whether a substance satisfies one or more mass criteria, the depiction is selected to show or even demonstrate the characteristics associated with these mass criteria.

[0133] It has been found that when one or more subsets of the fitted data have been solved (Beta), i Factor (-1 / A) i ( ) and the expected Beta factor x2 (i.e., (expected Beta factor) 2 At the same time, this may be for events occurring with at least one IRD. j Indicators of chemical exchange or diffusion over similar timescales.

[0134] The expected Beta factor can be advantageously determined empirically, and optionally chosen as a value between 1 and 2.

[0135] In one embodiment, when one or more subsets of fitted data have been solved (Beta) i Factor (-1 / A) i When the expected Beta factor x2 differs from the expected Beta factor, it is for at least one IRD. j Indication of chemical exchange or diffusion on a timescale ranging from 25% to +200%.

[0136] The T1* and associated T2* of the fitted subset of data may not represent the true T1 and T2 values ​​of the spin in the problem in one embodiment, for example, because the corresponding spin of the corresponding nucleus of the component in the matter is subject to temporary or permanent influence from adjacent and / or neighboring elements and / or component fragments. However, in one embodiment, the true T1 and T2 of the spin in the problem can be derived from the T1* and T2* values ​​of the spin, taking into account chemical exchange and / or spin diffusion, for example, during the fitting process.

[0137] In one embodiment, at least one of the fitted datasets comprises data embedding two or more subsets of fitted data, each subset including T1*, associated T2*, and associated intensity (M0). Advantageously, generating a property representation of the substance from at least one fitted dataset including the two or more subsets of fitted data comprises: deriving representation components, said representation components including the two or more subsets of fitted data for T1*, associated T2*, and associated intensity (M0).

[0138] Advantageously, at least one subset of fitted data each includes properties of a pair or set of spin types (including at least two spin types), and the method includes deriving chemical exchange properties from the properties of the pair or set of spin types, and determining T1 and T2 (true T1 and T2) based on the properties of T1* and T2* and the associated chemical exchange processes or processes. This allows for the determination of additional information about the substance.

[0139] Advantageously, each subset of fitted data includes at least one Beta factor and / or at least one diffusion parameter, wherein the at least one Beta factor and / or the at least one diffusion parameter comprises a characteristic of one, a pair, or a set of spin types. It has been shown that chemical exchange characteristics can be determined from these parameters, and advantageously, the method comprises deriving such chemical exchange characteristics from the characteristics of the pair or set of spin types and / or the diffusion characteristics of at least one spin type.

[0140] Input NMR data can include NMR data from any NMR-active isotope.

[0141] In one embodiment, the NMR data portion includes data from at least one NMR-readable isotope. Preferably, the NMR data of each of the one or more data portions is monoisotope NMR data including NMR data from one isotope.

[0142] In one embodiment, the one or more NMR data portions include data from at least two isotopes. Preferably, the method includes obtaining at least a plurality of NMR data portions, including at least a first data portion containing NMR data from a first isotope and a second NMR data portion containing data from a second isotope. The plurality of NMR data portions advantageously include at least a certain number of NMR data portions comprising NMR data from corresponding isotopes. The number of NMR data portions can conveniently be up to 50, for example up to 10, for example from 2 to 5.

[0143] To obtain a high-quality representation of the characteristics, it is desirable to fit only the corresponding NMR data portion from one isotope of each NMR data portion individually to the function.

[0144] Isotope examples suitable for the at least one isotope are selected from the group consisting of the following isotopes: ¹H, ²H, 6 Li, 7 Li,¹ 0 B、¹¹B、¹³C、¹ 4 N、¹ 5 N、¹ 7 O、¹ 9 F, ²³Na, ² 7 Al、² 9 Si、³¹P、³³S、³ 5 Cl、³ 7 Cl、³ 9 K, 4 ¹K、 4 ³Ca、 47 Ti、 49 Ti、 50 V. 5 ¹V、 5 ³Cr、 55 Mn, 57 Fe、 59 Co、 6 ¹Ni、 6 ³Cu、 65 Cu、 67 Zn, 69 Ga、 7 ¹Ga、 75 As、 77 Se、 79 Br、 8 ¹Br、 8 ³Kr、 85 Rb、 87 Rb、 87 Sr、 89 Y、 9 ¹Zr、9 ³Nb、 95 Mo、 97 Mo,¹ 05 Pd,¹ 07 Ag、¹ 09 Ag, ¹¹¹Cd, ¹¹³Cd, ¹¹ 7 Sn、¹¹ 9 Sn、¹¹ 5 Sn, ¹²¹Sb, ¹³ 5 Ba、¹³ 7 Ba、¹ 77 Pb,¹ 99 Hg,² 0 ¹Hg and² 07 Pb.

[0145] The first and second isotopes, or the corresponding isotopes, may conveniently include at least two isotopes selected from the following isotope group: ¹H, ²H, 6 Li, 7 Li,¹ 0 B、¹¹B、¹³C、¹ 4 N、¹ 5 N、¹ 7 O、¹ 9 F, ²³Na, ² 7 Al、² 9 Si、³¹P、³³S、³ 5 Cl、³ 7 Cl、³ 9 K, 4 ¹K、 4 ³Ca、 47 Ti、 49 Ti、 50 V. 5 ¹V、 5 ³Cr、 55 Mn, 57 Fe、 59 Co、 6 ¹Ni、 6 ³Cu、 65 Cu、 67 Zn, 69 Ga、 7 ¹Ga、 75 As、 77 Se、 79 Br、 8 ¹Br、 8 ³Kr、 85 Rb、 87 Rb、 87 Sr、 89 Y、 9 ¹Zr、 9³Nb、 95 Mo、 97 Mo,¹ 05 Pd,¹ 07 Ag、¹ 09 Ag, ¹¹¹Cd, ¹¹³Cd, ¹¹ 7 Sn、¹¹ 9 Sn、¹¹ 5 Sn, ¹²¹Sb, ¹³ 5 Ba、¹³ 7 Ba、¹ 77 Pb,¹ 99 Hg,² 0 ¹Hg and² 07 Pb.

[0146] The one or more isotopes used are preferably selected to include one or more isotopes expected to be present in the substance, and preferably include one or more isotopes present in high abundance in the substance. Examples of preferred isotopes include ¹H, ¹³C ... 4 N、¹ 5 N、¹ 7 O、¹ 9 F, ²³Na, ³ 9 K, 4 ¹K、³ 5 Cl、³ 7 Cl and 4 ³Ca.

[0147] In one embodiment, processing of each NMR data portion includes performing multiple fittings to two or more different functions, which may be as described above.

[0148] In one embodiment, the method further includes fitting two or more determined fitting datasets to a Bloch McConnell equation, such as a Bloch McConnell equation for two-point swaps or a Bloch McConnell equation for three-point swaps. The fitting datasets applied to fit the Bloch McConnell equations each include T1, T2, and M0.

[0149] The fitted dataset is advantageously fitted to the Bloch-McConnell equation within one or more of two or more groups to obtain additional information about chemical exchange, and this information is parameterized in the form of kinetic and / or exchange constants or rates. In one embodiment, the fitted dataset is fitted to the Bloch-McConnell equation within a group of fitted datasets representing corresponding spins, and the method includes determining at least two kinetic constants representing chemical exchange between the corresponding spins.

[0150] When one or more subsets of fitted data have been solved, Beta i Factor (1 / A) i When the expected Beta factor x2 differs from the expected Beta factor, it is particularly beneficial to fit two or more of the determined fitted datasets to the Bloch McConnell equation.

[0151] The Bloch-McConnell equation is known in the art and was developed by Harden M. McConnell based on the Bloch equation, as described by Harden M. McConnell in "Reaction Rates by Nuclear Magnetic Resonance", The Journal of Chemical Physics 28, 430 (1958); doi: 10.1063 / 1.1744152.

[0152] The recording of NMR data can be performed by any method, such as those known in the art. Advantageously, the step of obtaining the one or more portions of NMR data includes performing at least one NMR readout using an NMR spectrometer, such as a low-field NMR spectrometer.

[0153] Low-field NMR spectrometers can be advantageously configured to generate static magnetic fields, such as low fields of about 0.1 to about 5 Tesla. Low-field NMR spectrometers using a maximum magnetic field of about 5 Tesla, preferably about 3 Tesla or less, such as about 3 Tesla or less, have been found to be very advantageous.

[0154] Recording NMR data advantageously includes performing at least one readout process, and preferably multiple readout processes.

[0155] In cases involving multiple reads, these reads can be performed on a common sample of the substance or on different samples of the substance.

[0156] In one embodiment, the at least one NMR readout process includes subjecting a material sample to an RF excitation (exiting) pulse sequence and reading the relaxation / dephase signal of at least one spin of at least one isotope.

[0157] RF excitation pulses can include any RF irradiation that excites the nucleus of the isotope to be read.

[0158] Advantageously, the at least one NMR readout process includes subjecting a material sample to one or more RF irradiation sequences, the sequences including frequencies and / or time periods suitable for determining T1 and T2 relaxation and / or J coupling evolution, and / or dipole-dipole coupling evolution, and / or quadrupole coupling evolution, and / or chemical exchange and / or diffusion-dependent evolution.

[0159] In one embodiment, the at least one NMR readout process includes subjecting a material sample to one or more RF excitation pulse sequences, the sequences including at least one of the following pulse sequences: DEPT (distortion-free polarization transfer enhancement), DEPTQ (DEPT with quaternary carbon retention), HSQC (heteronuclear single quantum coherence), INEPT (insensitive nuclear polarization transfer enhancement), BIRD (bilinear rotation decoupling pulse), TANGO (detection of adjacent nuclei using a gyroscope operator), or NOE (nuclear Overhauser effect).

[0160] The nuclear Overhausen effect (NOE) is the transfer of nuclear spin polarization from one nuclear spin population to another through cross-relaxation. This is a common phenomenon observed in nuclear magnetic resonance (NMR) spectroscopy. The NOE can be used, for example, to determine intramolecular (and even intermolecular) distances. The NOE effect is the change in the population of protons (or other nuclei) when a spatially neighboring magnetic nucleus is decoupled or selectively saturated with a 180-degree pulse. Preferably, the pulse used is a soft pulse.

[0161] More information about DEPT and NOE and how they are performed can be found, for example, in John Homer et al., “New Method for NMR Signal Enhancement by Polarization Transfer, and Attached Nucleus Testing”, J. Chem. Soc., Chem. Commun., 1994, and Thomas L. James, “Fundamentals of NMR”, Chapter 1, Department of Pharmaceutical Chemistry, University of California, 1998.

[0162] (http: / / qudev.ethz.ch / content / courses / phys4 / studentspresentations / nmr / James_Fundamentals_of_NMR.pdf)

[0163] The at least one NMR readout process can conveniently subject the material sample to a sequence of RF excitation pulses comprising multiple pulses, including at least one 180º RF pulse and at least one 90º RF pulse.

[0164] In one embodiment, the at least one NMR readout process includes subjecting a material sample to an RF excitation pulse sequence comprising a Hahn echo sequence or a CPMG sequence or a combination of at least one of these sequences.

[0165] Advantageously, the at least one NMR readout process includes subjecting the material sample to an RF excitation pulse sequence including a spin echo sequence, a fast spin echo sequence, a stimulated echo sequence, a gradient echo sequence, or any combination of at least one of these sequences, such as as described above.

[0166] In one embodiment, the at least one NMR readout process includes subjecting a material sample to one or more RF excitation pulse sequences, the sequences including inversion recovery pulse sequences, Carr-Purcell-Meiboom-Gill (CPMG) sequences and / or spin echo sequences, or one or more combinations of at least one of these sequences.

[0167] Preferably, the at least one NMR readout process includes subjecting the material sample to an RF pulse and a delayed sequence comprising a subsequence SE1:

[0168] Apply the first 180º RF pulse

[0169] Apply the first inversion recovery delay (IRD).

[0170] Apply the first 90º RF pulse, and

[0171] The resonance intensity can be determined directly or through echo sequences (e.g., CPMG).

[0172] The SE1 subsequence can be repeated, and different IRD values ​​can be applied optionally. Furthermore, it can include determining a zero-point IRD value, at which the resonance intensity is determined passing through the zero-intensity point.

[0173] The delay time null Related to T1, i.e., Ƭ null It is 0.71 times that of T1.

[0174] In one embodiment, the at least one NMR readout process includes subjecting the material sample to an RF pulse and a delayed sequence comprising a subsequence SE1:

[0175] i. Apply a 90º RF pulse

[0176] ii. Apply echo delay

[0177] iii. Apply a 180º RF pulse

[0178] iv. Repeat steps ii and iii N times.

[0179] v. To apply a cyclic delay,

[0180] The SE2 subsequence can be repeated, for example, by applying different cyclic delays.

[0181] Advantageously, the at least one NMR readout process includes subjecting the material sample to a repeated sequence of RF excitation signals, the sequence including

[0182]

[0183] Where π and π / 2 represent radio frequency pulses, the pulse duration and the radio frequency pulse field intensity are adjusted to provide pulse flip angles of π (e.g., 180º) and π / 2 (e.g., 90º), respectively, and T INV It is the inversion recovery delay (IRD), T E It is echo delay (ED), T W It is the cyclic delay (RD), and N is the number of echoes.

[0184] In one embodiment, the SE2 pulse sequence may involve two TE-π elements, which are repeated N-1 and N'-1 times and separated by delay periods.

[0185] Therefore, highly useful input NMR data can be obtained, which is suitable for generating high-quality characteristic representations.

[0186] The pulse sequence has different components of one or more of the parameters IRD, ED, N, RD, radio frequency pulse field strength, or radio frequency pulse duration, and can be conveniently repeated multiple times to improve NMR data, thereby improving the NMR data portion that forms part of the input NMR data.

[0187] NMR data used to generate NMR portions containing different isotopes can be obtained, for example, by recording a signal from one isotope during one or more delay periods of a recording from another isotope. This is referred to herein as staggered NMR recording.

[0188] In one embodiment, the pulse sequence is repeated multiple times, each time with different components and numbers of inversion, wait, and echo periods.

[0189] Advantageously, the NMR recording includes a reading process according to at least any of the foregoing items, wherein the dataset is generated to include data from at least two different NMR readable isotopes. Preferably, one of the NMR readable isotopes is ¹H. Preferably, interleaved NMR recording is used to read the at least two different NMR readable isotopes.

[0190] In one embodiment, the NMR reading process includes reading at least two different NMR-readable isotopes, one of which is ¹³C.

[0191] In one embodiment, the NMR readout process includes parallel reading of multiple NMR readable isotopes, including initiating another pulse sequence for another NMR readable isotope during the RD, IRD, or ED period of a previous NMR readable isotope pulse sequence, i.e., applying staggered NMR recording.

[0192] NMR recordings can advantageously include performing one or more additional readouts over time, wherein additional NMR data portions can be obtained from each additional readout. The method may preferably include processing the additional NMR data portions to obtain additional representation components, and comparing and / or adding the additional representation components to the characteristic representation.

[0193] This allows us to obtain valuable information about potential changes in the sample.

[0194] This can be used, for example, to determine the stability of a substance, the development of a substance, changes in a substance in a production line, the reactivity of a substance, or similar information.

[0195] Information about changes in matter can include, for example, changes in physical conditions, such as temperature, hydration, pressure, salt content, pH, solvent type, and concentration of phase or component / element.

[0196] In one embodiment, fitting multiple data portions of a sample can improve measurement accuracy in terms of the respective standard deviation of one or more parameters, and / or reduce the data measurement time required to achieve a selected measurement accuracy.

[0197] The substance can, in principle, be any substance containing an active NMR isotope. The substance can be a solid, liquid, gas, or any mixture thereof.

[0198] Advantageously, the substance comprises at least two components, preferably each component comprising a unique NMR-readable isotope. The method and embodiments of the present invention have proven highly useful for analyzing complex substances, such as those comprising several different components, as described above.

[0199] Advantageously, the substance is a substance comprising two or more components, and preferably, for example, a food product, a sample of any stage of wastewater treatment, an animal slurry or manure, such as livestock feed, a biogas digester slurry, one or more samples of any stage of biorefining, or one or more samples of any stage of a mining or mineral refining process.

[0200] In one embodiment, the substance comprises at least organic and / or inorganic components. The substance may, for example, comprise monomers capable of polymerization, and the method may include performing one or more additional recordings over time, generating additional NMR data portions for each additional recording, processing the additional NMR data portions to obtain additional representation components, and determining the stage and / or quality of polymerization.

[0201] In one embodiment, the substance comprises at least organic and / or inorganic components. The substance may, for example, include chemical or biological reactants capable of interacting in chemical or biological synthesis, and the method may include performing one or more additional recordings over time, generating additional NMR data portions for each additional recording, processing the additional NMR data portions to obtain additional representation components, and determining the stage and / or quality of the chemical or biological reaction.

[0202] In one embodiment, the method includes preparing a property representation map of the substance. The property representation map of the substance may, for example, include probability maps of the corresponding spin types T1* and T2* as described above and / or a depiction as described above.

[0203] Advantageously, the method includes determining a mass parameter of the substance, including deriving data from a characteristic representation and determining the mass parameter, and optionally determining whether the mass parameter satisfies a mass criterion, such as a mass criterion including a selected threshold.

[0204] The mass parameters of a substance can be related to, for example, the amount of components, such as the amount of organic components, the amount of lipids, the amount of fats, the amount of carbohydrates, the amount of proteins, the amount of nucleic acids, the amount of chemical reagents, the amount of isotopes, the amount of mobile sample components, the amount of rigid / immobile sample components, the amount of species undergoing chemical exchange, or the amount of species undergoing molecular, ionic, or spin diffusion.

[0205] In one embodiment, the mass parameter of a substance is associated with its chemical properties (e.g., pH value).

[0206] The characteristic representation can include multiple identifiers, each of which represents the statistical distribution of the sample components.

[0207] The method may conveniently include comparing a property representation of a substance with at least one historical property representation to identify differences, such as chemical and / or biological and / or compositional differences between a substance and substantially similar substances.

[0208] The method may include, for example, determining quality parameters based on the identified differences.

[0209] The at least one historical characteristic representation may, for example, include the historical characteristic representation of a reference material, or a historical characteristic representation synthesized from two or more historical characteristic representations of a reference material.

[0210] The property representation of the reference material can be used, for example, to define quality criteria, such as thresholds. The generated property representation of the material can be compared with such a property representation of the reference material, for example, to determine whether the quality criteria are met.

[0211] The at least one reference substance may correspond, for example, to the substance, based on the fact that it contains at least 25% of the components present in the substance, for example, at least 50% of the components present in the substance, for example, at least 90% of the components present in the substance.

[0212] In one embodiment, the characteristic representation is a visual characteristic representation, such as a visual probability map or in the form of data representing such a probability map, such as a depiction. The method advantageously includes determining at least one quality parameter of the material by comparing the visual characteristic representation with a visual characteristic representation of a reference material or a historical visual characteristic representation, wherein the at least one quality parameter includes a similarity parameter, such as a similarity parameter including Bach distance. In one embodiment, the method includes classifying the resulting visual characteristic representation based on the determined similarity parameter.

[0213] In one embodiment, the processing may be performed by a trained computer, wherein the trained computer is trained to determine one or more quality parameters and one or more sample identifiers, for example, to determine quantitative parameters and / or to monitor differences in historical data, such as differences in chemical, physical, or biological processes, for example in a process monitoring setup. Preferably, the trained computer is trained to classify the substance based on a property representation generated on the substance by an embodiment of the method, according to one or more quality parameters and / or according to one or more sample identifiers.

[0214] The present invention also includes a method for generating a trained computer. The method includes subjecting the computer to a machine learning process, the process including feeding the computer data, wherein the data includes data packets of previously obtained NMR data associated with at least one fitted data subset containing T1*, associated T2*, and associated intensity (M0).

[0215] The method for generating the trained computer may include feeding a data packet, the data packet including data representing the properties of a substance and an NMR data portion applied to generate that property representation, for example as described above.

[0216] In one embodiment, the processing of the NMR data portion includes applying artificial intelligence, machine learning, reinforcement learning, or similar techniques in a supervised, unsupervised, or any other form.

[0217] The present invention also includes a trained computer that can be obtained by the above method.

[0218] The present invention also includes an NMR system for performing the method described above, the NMR system comprising a computer system configured to receive one or more input NMR data, wherein the input NMR data comprises a plurality of NMR datasets, each NMR dataset comprising intensity data and relaxation data, and

[0219] The computer system described therein is configured to:

[0220] - By performing multiple fits to the function, the NMR data portion is processed into a multidimensional distribution that fits the dataset, wherein the function includes at least two fitting parameters and is a function representing a model of the obtained NMR data portion, and

[0221] - Generate a property representation of the substance from the fitted dataset.

[0222] The NMR system can be advantageously configured to perform the data processing described above.

[0223] The computer system may be in the form of a single computer, or may include a separate computer component adapted for communication via wired or wireless data communication, preferably via the Internet and / or via Bluetooth or similar means. The computer system may include a personal computer, tablet, smartphone, and / or any integrated or external storage device.

[0224] Advantageously, the NMR system includes an NMR spectrometer, preferably in data communication with the computer system, and optionally includes a computer that forms part of the computer system.

[0225] In one embodiment, the computer system includes a supercomputer system, for example, capable of executing at least 10 [unclear] per second. 14 A computer capable of performing at least 10 floating-point operations per second (FLOPS) 15 A computer with FLOPS, for example, capable of executing at least 10 16FLOPS computers.

[0226] In one embodiment, the computer system includes a computer cluster, a computer system with multiple CPUs, or a computer system with multiple GPUs, such as 1,000 or more GPUs, optionally located at a distance relative to the NMR spectrometer.

[0227] Advantageously, the computer system includes a trained computer as described above. Attached Figure Description

[0228] In the following description, with reference to the accompanying drawings, the invention will be further illustrated by describing a number of illustrative and non-limiting embodiments and examples.

[0229] The accompanying drawings are schematic, not drawn to scale, and may have been simplified for clarity. The same reference numerals are used throughout the same or corresponding sections.

[0230] Figure 1a , Figure 1b and Figure 1c The visual characteristic representation obtained in Example 1 is shown.

[0231] Figure 2 This is a process diagram illustrating an embodiment of the method of the present invention, which includes generating characteristic representations of at least two NMR data portions.

[0232] Figure 3 This is a process diagram illustrating an embodiment of the method of the present invention, which includes determining whether a change has occurred in the substance.

[0233] Figure 4 This is a process diagram illustrating an embodiment of the method of the present invention, which includes determining whether a sample substance meets a quality criterion.

[0234] Figure 5a , 5a ', 5b and 5b' are visual feature representations related to Example 2.

[0235] Figure 6a , 6a ', 6b and 6b' are visual feature representations related to Example 3.

[0236] Figure 7a , 7b 7c, 7d, and 7e are visual feature representations related to Example 4.

[0237] Figure 8a , 8a ', 8b and 8b' are visual feature representations related to Example 5.

[0238] Specific implementation methods

[0239] The process includes characterization of at least two NMR data portions of the generated substance.

[0240] The substance can be as described above.

[0241] In step 20a, a first NMR data subset is obtained from a first isotope, such as one of the active isotopes described above. The first NMR data subset includes multiple NMR datasets, each including intensity data and relaxation data. In step 20b, the first NMR data subset is processed into a multidimensional distribution of the fitted dataset by performing multiple fittings to the function X1 to obtain a fitted dataset. In step 20c, a first fitted data subset is derived from the fitted dataset. In this embodiment, the first fitted data subset includes associated T1*, T2*, M0, and a Beta factor. In variations thereof, the Beta factor may be omitted and / or each fitted data subset may include additional fitting using a function and associated parameters that provide more detailed information about properties such as chemical exchange and / or diffusion, as described above. In step 20d, a first representation component is generated from the fitted data subset.

[0242] In step 21a, a second NMR data subset is obtained from a second isotope, such as one of the active isotopes described above. The second NMR data subset includes multiple NMR datasets, each containing intensity and relaxation data. In step 21b, the second NMR data subset is processed into a multidimensional distribution of the fitted dataset by performing multiple fittings to the function X2. In step 21c, a second fitted data subset is derived from the fitted dataset. In this embodiment, the first fitted data subset includes associated T1*, T2*, M0, and Beta factors. For spin types where the Beta factor differs from factor x2, fittings with other functions and parameters (e.g., Bloch McDonnell types) may be used to supplement the characterization of the chemical exchange and / or spin diffusion characteristics described above in more detail. In step 20d, a second representation component is generated from the fitted data subset.

[0243] In step 22, a feature representation is generated by combining the first representation component and the second representation component. It should be noted that each of the first and second representation components also represents a feature representation. In one embodiment, one or more additional representation components are generated in a manner similar to that used to generate the first and second representation components, and these additional representation components can be added to the feature representation.

[0244] Formulas X1 and X2 can be equal or different from each other. Formulas X1 and X2 can be independent of each other as described above.

[0245] In step 23, a feature representation is generated as a probability graph including T1* and T2*, for example as described above.

[0246] In step 24, a feature representation may be generated to include a description of at least a portion of the first and / or second fitted data subsets, such as as described above or as... Figure 1c As shown.

[0247] Figure 3 The process diagram illustrates an embodiment that includes determining whether a change has occurred in the substance.

[0248] In step 30a, one or more NMR data portions of the substance are acquired at a first time point. The NMR data portions may include NMR data from one or more active isotopes. In step 30b, a fitted dataset is obtained by performing multiple fittings to the function X3, processing each NMR data portion into a multidimensional distribution of the fitted dataset. In step 30c, a subset of fitted data is derived from the fitted dataset. In this embodiment, the first subset of fitted data includes associated T1*, T2*, M0, and Beta factors. For spin types where the Beta factor differs from factor x2, fittings with other functions and parameters may be used to supplement the model to characterize the chemical exchange and / or spin diffusion properties as described above in more detail. In step 30d, a property representation is generated from the subset of fitted data.

[0249] Repeat the process at the second time point.

[0250] In step 31a, one or more NMR data fractions of the substance are acquired at a second time point. The NMR data fractions may preferably include NMR data of one or more active isotopes that are the same as those acquired at the first time point. In step 31b, a fitted dataset is obtained by performing multiple fittings to the function X3, processing each NMR data fraction into a multidimensional distribution of the fitted dataset. In step 31c, a subset of the fitted data is derived from the fitted dataset.

[0251] In step 32, the characteristic representation from step 31d is compared with the characteristic representation from step 30d, and it can be determined whether a change has occurred during the period from the first time point to the second time point.

[0252] Figure 3The embodiment illustrated in the process diagram is well-suited for determining the stability and / or reactivity of a substance. This embodiment can be applied, for example, to determine the activity of a drug when mixed with a biological fluid, and / or to detect potential allergic reactions, wherein blood samples are acquired at a first time point, a second time point, and optionally subsequent time points, and analyzed by generating a characteristic representation. The second time point can be from one second to several months or even years after the first time point, depending on the potential changes being analyzed.

[0253] Figure 4 The process diagram illustrates one embodiment of the method, which includes determining whether a sample substance meets a quality criterion.

[0254] The process includes generating a reference property representation based on one or more reference materials, then generating a property representation of the sample material, and determining whether the sample material meets the quality criteria.

[0255] In step 40a, one or more NMR data portions of a reference material are acquired. The NMR data portions may include NMR data from one or more active isotopes. In step 40b, a fitted dataset is obtained by performing multiple fits to the function X4, processing each NMR data portion into a multidimensional distribution of the fitted dataset. In step 40c, a subset of fitted data is derived from the fitted dataset. In this embodiment, the first subset of fitted data includes associated T1*, T2*, M0, and Beta factors. This can be supplemented using additional fittings to obtain more detailed characteristics regarding chemical exchange or diffusion phenomena. In step 40d, a reference characteristic representation is generated from the subset of fitted data.

[0256] In step 41a, one or more NMR data portions of the sample material are acquired. The NMR data portions preferably include NMR data from one or more active isotopes used for the reference material. In step 41b, a fitted dataset is obtained by performing multiple fits to the function X4, processing each NMR data portion into a multidimensional distribution of the fitted dataset. In step 41c, a subset of fitted data is derived from the fitted dataset. The first subset of fitted data includes the same or similar parameters as the subset of fitted data obtained for the reference material. In step 41d, a characteristic representation is generated from the subset of fitted data.

[0257] In step 42, the characteristic representation from step 41 is compared with the reference characteristic representation from step 40d. Based on this, it can be determined whether the sample substance meets the quality criteria.

[0258] Figure 4The embodiment shown in the process diagram is well-suited for determining whether a substance meets quality criteria, such as quality criteria associated with the content of one or more components, quality criteria associated with the interaction between two or more components, quality criteria associated with the chemical exchange between two or more components, quality criteria associated with the diffusion of components, quality criteria associated with the homogeneity of the substance, quality criteria associated with the presence of potential impurities, or any other chemical, physical, or biological criteria.

[0259] In the following examples, the Tveskaeg™ NMR instrument, sold by NanoNord A / S, was used.

[0260] Example 1

[0261] In this example, the substance is commercially salted butter based on milk.

[0262] A package of butter was provided.

[0263] A portion of the sample was analyzed by ¹H NMR to obtain partial NMR data of the isotopes.

[0264] NMR data were recorded using a series of CPMG and inversion recovery CPMG experiments. The experiments used 18 (m) logarithmically distributed inversion recovery delays between 0 and 3 seconds (i.e., in addition to the standard CPMG experiment, 18 different inversion recovery delays (IRDs) were used, with delays of 0, 0.001, 0.00164938, 0.00272045, 0.00448704, 0.00740083, 0.0122068, 0.0201336, ...). The inversion recovery CPMG experiment was performed at pulses of 0.0332078, 0.0547723, 0.0903401, 0.149005, 0.245765, 0.40536, 0.668592, 1.10276, 1.81887, and 3 seconds. Free induction decay was recorded between a series of 1000 (N) echoes spaced 200 μs (TE) apart in the CPMG sequence. 101 NMR data points were sampled at equal intervals within each 200 μs interval. The experiment used a 9 μs 90° (π / 2) RF pulse and an 18 μs 180° (π) RF refocusing pulse, with a cycle delay of 4 seconds (TW), two scans (standard phase cycling to reduce instrument defects), a static magnetic field of approximately 0.9 T, and a sample temperature of approximately 39°C.

[0265] Before recording NMR data, the NMR spectrometer is in a standard calibration state with calibrated radio frequency pulses, and the intensity of the CPMG experiment is calibrated using the calibrated state with calibrated radio frequency pulses and the intensity of a standard calibration sample. This is called a one-time factory calibration.

[0266] An NMR data portion is generated from the recorded data. The recorded NMR data portion is organized into NMR datasets, each dataset including intensity data and relaxation data (T1 and T2).

[0267] As described above, by performing multiple fits to the function FI, the NMR data is partially processed into a multidimensional distribution of the fitted dataset:

[0268]

[0269] Where A i B i C i D i and E i These are the fitted parameters, and IRD j TE k,l It is input data from the corresponding NMR dataset;

[0270] i is one of the four different spin types;

[0271] IRD j The inversion recovery delay is the inversion recovery delay applied when acquiring j corresponding NMR datasets from m different inversion recovery delays of 0-3 seconds.

[0272] TE k,l is the echo time, where k represents the k-th length of each echo from 1 to p in the echo string, and l is the number of echoes from 1 to q in the corresponding echo string.

[0273] In this example, n = 4 spin types, and 1000 random initial guesses for M0i, T1*i, and T2*i are applied during the fitting process.

[0274] Further analysis was performed on the fitted dataset, and subsets of the fitted data were derived, each subset including associated T1*, T2, intensity (M0), and Beta factor values, wherein the values ​​of the corresponding subsets are associated by representing a common spin type and derived from a common subset.

[0275] Mapping T1* and T2* values ​​to obtain, for example: Figure 1a The characteristics are represented in the form of the T1-T2 spectrum shown. Figure 1aThe graph in the image shows all possible solutions obtained from the fitted dataset, where each point (or circle) in the T1-T2 graph represents a set of T1* and T2* values ​​for a subset of data of a spin type; the axes are displayed on a linear scale in seconds. In this case, the Beta factor is not significantly different from the x2 factor, meaning that the measured T1* and T2* values ​​are not significantly affected by chemical exchange, while the effects of field inhomogeneity may influence the values. Figure 1b It shows how to accumulate Figure 1a The intensity of the solutions is used to obtain a probability plot, thus yielding the statistical probabilities of distinguishable components in the T1*-T2* spectrum. In this probability plot, the coordinate axes are scaled logarithmically to reveal more feature information. From... Figure 1b In the probability plot, two main regions, W and F, are identified by the dashed circle. Regions W and F are distinguished by the difference in their T1* and T2* values. Region W is found to represent water, and region F represents fat.

[0276] from Figure 1b Based on the probability plot and regions W and F, the water content of the butter sample was determined to be 19.2 wt%, while the water content declared on the sampled butter packaging was 15 wt%.

[0277] Figure 1c The characteristics are shown in a descriptive form. This descriptive form can be provided, for example, as described above. Figure 1c The depiction shown is based on Figure 1b The probability plot provides a clear illustration of the water content relative to the fat content. By generating a characteristic representation as a depiction, it becomes easier and relatively faster for users, such as operators or computers in a machine learning setup, to determine whether the substance generating that characteristic representation meets selected criteria.

[0278] Example 2

[0279] In this example, a number of food samples were analyzed, which were cream samples with two different fat contents.

[0280] Cream samples with declared fat contents of 8 wt% (low-fat cream) and 37 wt% (high-fat cream) were studied separately. A subset of the corresponding samples was subjected to NMR analysis. ¹H NMR data were recorded using a series of CPMG and inversion recovery CPMG experiments, which employed 18 (m) logarithmic inversion recovery delays (T0) between 0 and 3 seconds. INV(The values ​​were 0.001, 0.00164938, 0.00272045, 0.00448704, 0.00740083, 0.0122068, 0.0201336, 0.0332078, 0.0547723, 0.0903401, 0.149005, 0.245765, 0.40536, 0.668592, 1.10276, 1.81887, and 3 seconds of inversion recovery CPMG experiment), and free induction attenuation was recorded between a series of 1000 (N) echoes spaced 200 μs (TE) apart in the CPMG sequence; 101 points were sampled at equal intervals within each 200 μs time period. The experiment used 9μs 90° (π / 2) and 18μs 180° (π) pulses with a repetition delay of 4 seconds (T). W Two scans were performed, with a static magnetic field of approximately 1 T and a sample temperature of approximately 39°C.

[0281] Before recording NMR data, the NMR instrument is in a standard calibration state, with calibrated radio frequency pulses, and the intensity of the CPMG experiment is calibrated using a standard calibration sample. This is called a one-time factory calibration.

[0282] An NMR data portion is generated from the recorded NMR data, which includes multiple datasets, each of which includes inversion recovery delay (IRD) and associated echo time (ED).

[0283] The NMR data portion was analyzed using the same function as in Example 1, which uses n = 4 spin types and 1000 parameters related to M0. i T1* i and T2* i The random initial guess value. This yields the corresponding low-fat cream. Figure 5a And corresponding high-fat cream Figure 5b The visual characteristics shown are represented in the form of T1*-T2* maps.

[0284] For low-fat cream samples ( Figure 5a ) and high-fat cream samples ( Figure 5b Each corresponding sample in ) Figure 5a and 5b The graphs on the left of each graph show all possible solutions, where each point represents a T1* and T2* value for a spin type; the axes are displayed on a linear scale in seconds. Figure 5a and Figure 5b The plots on the right side of each plot show probability maps using logarithmic scale axes. For both cream samples, we marked the fat region with a box labeled F and the water region with a box labeled W, clearly showing that the low-fat sample had a significantly higher fat / water ratio than the high-fat sample. Figure 5a and 5b The graphs on the left of each graph indicate that some, but not all, of the water underwent exchange through a beta value greater than 2 (marked in gray). From the data, we extrapolated fat contents of 9.3 wt% and 36.2 wt%, which are in excellent agreement with known values ​​of 8 wt% and 37 wt%, respectively.

[0285] Figure 5a 'and 5b' are shown in color corresponding to Figure 5a and 5b The map.

[0286] Example 3

[0287] In this example, a number of food samples were analyzed, which were samples of two different vegetable oils.

[0288] Coconut oil and rapeseed oil samples were subjected to NMR analysis. ¹H NMR data were recorded using a series of CPMG and inversion recovery CPMG experiments, which employed 18 (m) logarithmic inversion recovery delays between 0 and 3 seconds (i.e., in addition to the standard CPMG experiment, experiments with inversion recovery delays (T0, T ... INV (The values ​​were 0.001, 0.00164938, 0.00272045, 0.00448704, 0.00740083, 0.0122068, 0.0201336, 0.0332078, 0.0547723, 0.0903401, 0.149005, 0.245765, 0.40536, 0.668592, 1.10276, 1.81887, and 3 seconds of inversion recovery CPMG experiment), and free induction attenuation was recorded between a series of 1000 (N) echoes spaced 200 μs (TE) apart in the CPMG sequence; 101 points were sampled at equal intervals within each 200 μs time period. The experiment used 9μs 90° (π / 2) and 18μs 180° (π) pulses with a repetition delay of 4 seconds (T). W Two scans were performed, with a static magnetic field of approximately 1 T and a sample temperature of approximately 39°C.

[0289] Before recording data, the NMR instrument is in standard calibration mode, with calibrated radio frequency pulses, and the intensity of the CPMG experiment is calibrated using a standard calibration sample. This is called a one-time factory calibration.

[0290] An NMR data portion is generated from the recorded NMR data, which includes multiple datasets, each of which includes inversion recovery delay (IRD) and associated echo time (ED).

[0291] The NMR data portion was analyzed using the same function as in Example 1, which uses n=4 spin types and 1000 parameters related to M0. i T1* i and T2* i The random initial guess value. This yields the corresponding coconut oil. Figure 6a and corresponding rapeseed oil Figure 6b The visual characteristics shown are represented in the form of T1*-T2* maps.

[0292] For coconut oil ( Figure 6a ) and rapeseed oil ( Figure 6b Each corresponding sample in ) Figure 6a and Figure 6b The graphs on the right show all possible solutions, with each point representing a T1* and T2* value for a spin type; the axes are displayed on a linear scale in seconds. Figure 6a and Figure 6b The plots on the right side of each plot show probability maps with the axes scaled logarithmically. For both vegetable oils, the T1*-T2* plots reflect the lipid composition (chain length) and flowability of the samples. Therefore, these plots provide detailed information that can be used to identify and monitor changes in the production line, for example, for quality control. Both samples contain a variety of different lipids, resulting in... Figure 6a and 6b The ridge-like pattern observed in the visual characteristics representation. Coconut oil is characterized by its shorter lipid chain length compared to rapeseed oil.

[0293] Figure 6a 'and 6b' are shown in color corresponding to Figure 6a and 6b The diagrams on the left side of each diagram.

[0294] Example 4

[0295] In this example, substances containing various fluorinated compounds were analyzed, including Teflon® and environmentally challenging PFAS / PFOS molecules.

[0296] Samples of various substances were provided. A portion of each sample was subjected to NMR analysis. A series of CPMG and inversion recovery CPMG experiments were used to record¹ 9 F NMR data, the experiment used 18 (m) logarithmic inversion recovery delays between 0 and 9 seconds (i.e., in addition to the standard CPMG experiment, with inversion recovery delays (T0). INV(CPMG inversion recovery experiments with 0.001, 0.00176661, 0.0031209, 0.00551341, 0.00974004, 0.0172068, 0.0303977, 0.0537008, 0.0948683, 0.167595, 0.296075, 0.523048, 0.92402, 1.63238, 2.88378, 5.09451, and 9 seconds) were conducted, and free induction attenuation was recorded between a series of 1000 (N) echoes spaced 200 μs (TE) apart in the CPMG sequence; 101 points were sampled at equal intervals within each 200 μs time period. The experiment used 9μs 90° (π / 2) and 18μs 180° (π) pulses with a repetition delay of 12 seconds (T). W Two scans were performed, with a static magnetic field of approximately 1 T and a sample temperature of approximately 39°C.

[0297] Before recording data, the NMR instrument is in standard calibration mode, with calibrated radio frequency pulses, and the intensity of the CPMG experiment is calibrated using a standard calibration sample. This is called a one-time factory calibration.

[0298] An NMR data portion is generated from the recorded NMR data, which includes multiple datasets, each of which includes inversion recovery delay (IRD) and associated echo time (ED).

[0299] The NMR data portion was analyzed using the same function as in Example 1, which uses n = 4 spin types and 1000 parameters related to M0. i T1* i and T2* i The random initial guess value.

[0300] This yielded a visual representation of the properties in the form of T1*-T2* spectra. The figure shows a representative sample containing the PFAS compounds PFDA, PFOS, and PFTeA. Figure 7a The T1* and T2* probability plots are shown for samples containing 30 ppm Teflon (left), 2500 ppm of two PFAS molecules (middle), and 20 ppm of one PFAS molecule (right). The axes of the probability plots are displayed on a logarithmic scale in seconds for T1* and T2*. Figure 7b T1* and T2* probability plots are shown for samples containing three different PFAS molecules: PFOA, PFOS, and PFTeA. The results were obtained through NMR measurements¹. 9 The F concentration (mg / L or ppm) is 469 ppm, which is consistent with the laboratory measurement of 381 ppm, considering the uncertainty of laboratory data. Figure 7c The typical characteristics of different types of fluorine in this study¹ are presented.9 F T1*-T2* domain. Figure 7d Laboratory and NMR measurements of various fluorine-containing samples are shown. Figure 7e It shows Figure 7e The graphs of laboratory and NMR measurement data show the correlation between them.

[0301] Example 5

[0302] In this example, two animal pulp samples containing different amounts of total phosphorus (TP), total nitrogen (TN), and ammonium nitrogen (NHx-N) were provided and analyzed.

[0303] Each sample was subjected to NMR analysis, with NMR data recorded using a series of CPMG and inversion recovery CPMG experiments. These experiments used 18 (m) logarithmically distributed inversion recovery delays between 0 and 3 seconds (i.e., in addition to the standard CPMG experiment, experiments with inversion recovery delays (T0) were performed). INV (The values ​​were 0.001, 0.00164938, 0.00272045, 0.00448704, 0.00740083, 0.0122068, 0.0201336, 0.0332078, 0.0547723, 0.0903401, 0.149005, 0.245765, 0.40536, 0.668592, 1.10276, 1.81887, and 3 seconds of inversion recovery CPMG experiments), and a series of 1000 (N) intervals of 200 μs (T) in the CPMG sequence. E Free induction attenuation was recorded between echoes; 101 points were sampled at equal intervals within each 200 μs period. The experiment used 9 μs 90° (π / 2) and 18 μs 180° (π) pulses with a repetition delay of 4 seconds (T). W Two scans were performed, with a static magnetic field of approximately 1 T and a sample temperature of approximately 39°C.

[0304] Before recording data, the NMR instrument is in standard calibration mode, with calibrated radio frequency pulses, and the intensity of the CPMG experiment is calibrated using a standard calibration sample. This is called a one-time factory calibration.

[0305] An NMR data portion is generated from the recorded NMR data, which includes multiple datasets, each of which includes inversion recovery delay (IRD) and associated echo time (ED).

[0306] The NMR data portion was analyzed using the same function as in Example 1, which uses n=4 spin types and 1000 parameters related to M0. i T1* i and T2* iThe random initial guess value.

[0307] These respectively yielded the following results: Figure 8a and 8b The visual characteristics of Sample 1 (TP: 508 ppm, TN: 3389 ppm, NHx-N: 2006 ppm) and Sample 2 (TP: 771 ppm, TN: 5213 ppm, NHx-N: 3270 ppm) shown on the left are represented in the form of T1*-T2* spectra. Figure 8a and 8b The right side shows the corresponding T1* and T2* probability plots, with the axes of the probability plots displaying T1* and T2* on a logarithmic scale in seconds.

[0308] Figure 8a ' and 8b' are shown in color, corresponding to respectively Figure 8a and 8b The diagram on the left.

Claims

1. A method for generating a property representation of a substance, comprising: - Record nuclear magnetic resonance (NMR) data for at least one sample of the substance. - Generate one or more portions of nuclear magnetic resonance (NMR) data from the recorded NMR data, wherein each NMR data portion is generated as comprising one or more NMR datasets. - Each NMR data portion is processed to fit a multidimensional distribution of the dataset, the processing including selecting a function and performing multiple fits on the selected function, and - Generate a property representation of the substance from the fitted dataset. Each dataset includes associated intensity data and relaxation data. The selected function includes at least three fitting parameters. The processing of each of the NMR data portions includes selecting and applying at least one initial guess value, and Each fitted dataset includes data embedded with at least one subset of fitted data, the subset of fitted data including apparent longitudinal relaxation time (T1*), associated apparent transverse relaxation time (T2*), and associated intensity (M0), preferably the function being a function representing a model of the obtained NMR data portion.

2. The method of claim 1, wherein each NMR dataset is generated to include frequency data, such as at least one of resonance frequency, chemical shift, quadrupole coupling, J coupling, or dipole-dipole coupling.

3. The method according to claim 1 or 2, wherein each NMR dataset is generated to include at least one of inversion recovery data or echo NMR data, preferably each NMR dataset includes data representing at least one inversion recovery delay (IRD) and / or data representing at least one echo delay (ED).

4. The method according to claim 3, wherein the echo data includes at least one of spin echo data, fast spin echo data, stimulated echo data, or gradient echo data.

5. The method according to any one of the preceding claims, wherein each NMR dataset includes data from recorded NMR data, wherein the NMR data is recorded using a pulse sequence, the pulse sequence including at least one period of the NMR data being affected by T1 (longitudinal) relaxation and / or at least one period of the NMR data being affected by T2 (lateral) relaxation, for example, each NMR dataset includes data from recorded NMR data, the NMR data being recorded using a pulse sequence, the pulse sequence including at least two periods of T1 relaxation or at least two periods of T2 relaxation.

6. The method according to any one of the preceding claims, wherein each NMR dataset includes data from recorded NMR data, said NMR data being recorded using a pulse sequence, said pulse sequence including at least one period in which the recorded data is affected by chemical exchange.

7. The method according to any one of the preceding claims, wherein each NMR dataset includes data from recorded NMR data, said NMR data being recorded using a pulse sequence, said pulse sequence including at least one period in which the recorded data is affected by diffusion, said diffusion being, for example, molecular diffusion, ion diffusion, or spin diffusion.

8. The method according to any one of the preceding claims, wherein one or more NMR data portions include at least two NMR datasets, such as at least three NMR datasets, such as at least four NMR datasets, such as at least eight NMR datasets, such as sixteen or more NMR datasets.

9. The method according to any one of the preceding claims, wherein the data of at least one NMR dataset is data from a common or identical pulse sequence of the recorded NMR data, optionally, the NMR dataset of at least one NMR data portion includes data from a common or identical pulse sequence of the recorded NMR data.

10. The method according to any one of the preceding claims, wherein the data of at least one NMR dataset is data from a common or identical pulse sequence group of recorded NMR data, said pulse sequence group including one or more sequences among inversion recovery pulse sequences, Carr-Purcell-Meiboom-Gill (CPMG) sequences and / or spin echo sequences having one or more echoes, optionally, the NMR dataset of at least one NMR data portion includes data from a common or identical pulse sequence group.

11. The method according to any one of the preceding claims, wherein the data of at least one NMR dataset is based on data from recorded NMR data of a common isotope, preferably, the NMR dataset of at least one NMR data portion is based on data from recorded NMR data of a common NMR readable isotope.

12. The method according to any one of the preceding claims, wherein processing for each of the NMR data portions includes performing the multiple fittings of the NMR data portion individually to the formula.

13. The method according to any one of the preceding claims, wherein processing of each of the NMR data portions includes performing the multiple fittings of the one or more NMR datasets of the NMR data portions to the formula during a common fitting process.

14. The method according to any one of the preceding claims, wherein performing the multiple fitting on the NMR data portion comprises applying the at least one initial guess value to at least one of the multiple fittings, preferably, performing the multiple fitting on the NMR data portion comprises applying the at least one initial guess value to multiple of the multiple fittings, such as all of the multiple fittings, wherein the at least one initial guess value may be the same or different.

15. The method of claim 14, wherein the initial guess value includes an initial guess value for at least one of the at least three fitting parameters for at least one of the multiple fits, preferably the initial guess value includes an initial guess value for at least one of the at least three fitting parameters for multiple fits, such as for all fits in the multiple fits, wherein the initial guess value can be the same or different from one fit to subsequent fits depending on the selected pattern.

16. The method according to any one of the preceding claims, wherein the at least one initial guess value includes at least one of a random initial guess value or an initial guess value generated by applying a bootstrap method.

17. The method according to any one of the preceding claims, wherein the at least one initial guess is applied to at least one fitting parameter in multiple fittings, for example, to each fitting in the multiple fittings of the function, for example, including in at least two or more fittings in the multiple fittings of the function, the initial guess applied to the at least one fitting parameter is the same initial guess, for example, the same initial guess in each fitting in the multiple fittings of the function.

18. The method according to any one of the preceding claims, wherein processing of a corresponding NMR data portion among a plurality of NMR data portions comprises applying a randomly selected initial guess value to at least one fitting parameter for each NMR data portion.

19. The method according to any one of the preceding claims, wherein the at least one initial guess is characterized in that it is a Monte Carlo type random initial guess.

20. The method according to any one of the preceding claims, wherein the at least one initial guess is selected within a starting range, preferably the at least one initial guess is an initial guess for the fitting parameters, and the initial guess is within a starting range represented by a span of selected values, wherein the span of selected values ​​preferably includes one or more typical expected values ​​or consists of one or more typical expected values.

21. The method according to any one of the preceding claims, wherein the at least one initial guess is selected as an initial guess for a fitting parameter associated with at least one fitting parameter of the fitted dataset and / or with a subset of the fitted data, and wherein the initial guess preferably includes a parameter within a starting range represented by a span of selected values, wherein the span of selected values ​​includes or consists of typical expected values, wherein the typical expected values ​​are determined based on typical values ​​of the at least one fitting parameter, for example, previously determined values ​​of the at least one fitting parameter during the fitting process of another NMR data portion.

22. The method according to any one of the preceding claims, wherein the at least one initial guess is selected as an initial guess for a fitting parameter associated with at least one parameter of the fitted data subset, wherein the at least one parameter of the fitted data subset is selected from the apparent intensity (M0), T1* and T2*, frequency, chemical exchange or diffusion of the species in the at least one sample.

23. The method according to any one of the preceding claims, wherein the function comprises one or more exponential terms.

24. The method according to any one of the preceding claims, wherein the function comprises one or more hyperbolic terms.

25. The method according to any one of the preceding claims, wherein the function comprises one or more triangular terms.

26. The method according to any one of the preceding claims, wherein the function comprises one or more terms selected from logarithmic terms, exponential terms, sine terms, polynomial terms, or differential terms.

27. The method according to any one of the preceding claims, wherein the function comprises 4 fitting parameters, such as 5 fitting parameters, such as 6 fitting parameters, such as 7 fitting parameters, such as 8 fitting parameters or more.

28. The method according to any one of the preceding claims, wherein the function includes the function FI. Where A i B i C i D i and E i These are the fitted parameters, and IRD j TE k,l It is input data from the corresponding NMR dataset; i is one of the n different spin types; IRD j It is the inversion recovery delay applied when obtaining j corresponding NMR datasets from a total of m different inversion recovery delays; TE k,l It is the echo time, where k represents the k-th length of each echo from 1 to p in the echo string, and l is the number of echoes from 1 to q in the corresponding echo string; and Where n, m, p, and q are each an integer of 1 or higher, for example, at most 100, for example, from 1 to 50, for example, from 4 to 25.

29. The method according to claim 28, wherein the function FI is 。 30. The method according to any one of the preceding claims, wherein the function is function FII. Where A i B i C i D i and E i These are the fitted parameters, and IRD j TE k,l and RD h It is the input NMR data from the corresponding NMR dataset; i is one of the n different spin types; IRD j It is the inversion recovery delay applied when obtaining j corresponding NMR datasets from a total of m different inversion recovery delays; TE k,l is the echo time, where k represents the k-th length of each echo from 1 to p in the echo string, and l is the number of echoes from 1 to q in the corresponding echo string; RD h It is a cyclic delay, during which the magnetization recovers or partially recovers from the h-th cyclic delay of 1 to r cyclic delays from the execution of one NMR recording to the execution of a subsequent NMR recording; and Where n, m, p, q and r are each an integer of 1 or higher, for example, at most 100, for example, from 1 to 50, for example, from 4 to 25.

31. The method according to claim 30, wherein the function FII is 。 32. The method according to any one of the preceding claims, wherein the intensity data of the corresponding NMR dataset includes echo time, and wherein the relaxation data of the corresponding NMR dataset includes inversion recovery delay, preferably the corresponding NMR dataset further includes cyclic delay.

33. The method according to any one of claims 28-32, wherein the parameter C i The parameter D is associated with T1*, which is a subset of fitted data representing the echo data of spin type i out of n spin types; i The T2* is related to the subset of fitted data representing the echo data of spin type i out of n spin types; and / or the parameter A. i and B i At least one of the fitted data subsets is associated with the intensity M0 and / or chemical exchange value of the echo data representing spin type i among n spin types.

34. The method according to any one of claims 28-33, wherein each subset of fitted data includes the factor Beta. i Beta i =-1 / A i The factor, i, is used for spin type i among the n spin types in each fitted dataset, representing the label of chemical exchange.

35. The method according to any one of claims 28-34, wherein each subset of fitted data includes an M0 factor, where M0i = -A i *B i The factor, used for spin type i among the n spin types in each fitted dataset, represents the strength of the spin type.

36. The method according to any one of claims 28-35, wherein T1* i =-1 / C i Factor, for spin type i among n spin types for each fitted dataset, where T1* i T1* represents spin type i.

37. The method according to any one of claims 28-36, wherein the T2* i =-1 / D i Factor, for spin type i among n spin types for each fitted dataset, where T2* i T2* represents spin type i.

38. The method according to any one of claims 28-37, wherein the E i The factor is used for spin type i among the n spin types in each fitted dataset, where E i The factor represents the diffusion of molecules, fragments, or ions containing spin type i.

39. The method according to any one of claims 28-38, wherein n is at least 2, for example at most 1000, for example from 4 to 100.

40. The method according to any one of the preceding claims, wherein each subset of fitted data represents a spin-type property, preferably including at least T1*, associated T2* and associated intensity (M0), more preferably each subset of fitted data represents a spin-type property, which further includes one or more of diffusion parameters, frequency shift, chemical exchange, J coupling, quadrupole coupling or dipole-dipole coupling.

41. The method according to any one of the preceding claims, wherein generating a property representation of the substance from the fitted dataset comprises: A representation component is derived from each fitted dataset to obtain multiple representation components, and the multiple representation components are combined to produce a property representation of the substance, wherein the representation component includes at least one subset of fitted data containing T1*, associated T2*, and associated intensity (M0).

42. The method of claim 41, wherein the T1*, the associated T2*, and the associated intensity (M0) of the at least one subset of fitted data comprise data representing the spin diffusion characteristics of molecular or ionic species in a non-uniform static field or radio frequency field.

43. The method of claim 41 or 42, wherein the T1*, the associated T2*, and the associated intensity (M0) of the at least one subset of fitted data comprise data representing chemical exchange properties involving two or more spin types in the material.

44. The method according to any one of the preceding claims, wherein the characteristic representation includes probability diagrams of T1* and T2* of the corresponding spin types.

45. The method according to any one of the preceding claims, wherein if the solution of Beta for one or more subsets of fitted data is obtained... i Factor (-1 / A) i The value exceeds or is at least different from the expected value of factor x2, which is for occurrences with at least one IRD. j Indicators of chemical exchange or diffusion over similar timescales.

46. ​​The method according to any one of the preceding claims, wherein the observed T1*-T2* correlation does not represent the T1-T2 correlation of the spin type, and wherein the method includes extracting the T1-T2 correlation as part of a fitting process.

47. The method according to any one of the preceding claims, wherein at least one of the fitted datasets comprises data embedding two or more subsets of fitted data, said subsets including T1*, associated T2*, and associated intensity (M0), and wherein generating a property representation of the substance from the at least one fitted dataset comprising said two or more subsets of fitted data comprises: Derive representation components from the at least one fitted dataset, the representation components comprising the two or more subsets of fitted data of T1*, associated T2*, and associated intensity (M0).

48. The method of any one of claims 41-47, wherein at least one subset of fitted data each comprises a pair or set of characteristics of spin types containing at least two spin types, and wherein the method comprises deriving chemical exchange characteristics from the characteristics of the pair or set of spin types, and determining T1 and T2 from the characteristics of T1* and T2* and the associated chemical exchange process or multiple processes.

49. The method according to any one of claims 41-48, wherein each subset of fitted data includes at least one Beta. i Factor and / or at least one diffusion parameter, wherein the at least one Beta i The factor and / or the at least one diffusion parameter includes a property of one, a pair, or a set of spin types, and the method includes deriving chemical exchange properties from the properties of the pair or set of spin types and / or the diffusion properties of at least one spin type. Preferably, the generation of the property representation of the substance includes generating a property representation of the substance to include the chemical exchange properties from the properties of the pair or set of spin types and / or the diffusion properties of at least one spin type.

50. The method according to any one of the preceding claims, wherein the NMR data portion comprises data from at least one NMR-readable isotope, preferably the NMR data of one of the one or more data portions is monoisotope data comprising NMR data from one isotope.

51. The method according to any one of the preceding claims, wherein the NMR data portion comprises data from at least two isotopes, preferably the method comprises generating at least a plurality of NMR data portions, including generating at least a first data portion comprising NMR data from a first isotope and a second NMR data portion comprising data from a second isotope, preferably the plurality of NMR data portions comprising at least a certain number of NMR data portions comprising NMR data from corresponding isotopes, wherein the number of NMR data portions is at most 50, for example at most 10, for example from 2 to 5.

52. The method according to claim 50 or 51, wherein the at least one isotope is selected from the group consisting of: ¹H, ²H, 6 Li, 7 Li,¹ 0 B、¹¹B、¹³C、¹ 4 N、¹ 5 N、¹ 7 O、¹ 9 F, ²³Na, ² 7 Al、² 9 Si、³¹P、³³S、³ 5 Cl、³ 7 Cl、³ 9 K, 4 ¹K、 4 ³Ca、 47 Ti、 49 Ti、 50 V. 5 ¹V、 5 ³Cr、 55 Mn, 57 Fe、 59 Co、 6 ¹Ni、 6 ³Cu、 65 Cu、 67 Zn, 69 Ga、 7 ¹Ga、 75 As、 77 Se、 79 Br、 8 ¹Br、 8 ³Kr、 85 Rb、 87 Rb、 87 Sr、 89 Y、 9 ¹Zr、 9 ³Nb、 95 Mo、 97 Mo,¹ 05 Pd,¹ 07 Ag、¹ 09 Ag, ¹¹¹Cd, ¹¹³Cd, ¹¹ 7 Sn、¹¹ 9 Sn、¹¹ 5 Sn, ¹²¹Sb, ¹³ 5 Ba、¹³ 7 Ba、¹ 77 Pb,¹ 99 Hg,² 0 ¹Hg and² 07 Pb.

53. The method of claim 50 or 51, wherein the first and second isotopes or the corresponding isotopes comprise at least two isotopes selected from the group consisting of: ¹H, ²H, 6 Li, 7 Li,¹ 0 B、¹¹B、¹³C、¹ 4 N、¹ 5 N、¹ 7 O、¹ 9 F, ²³Na, ² 7 Al、² 9 Si、³¹P、³³S、³ 5 Cl、³ 7 Cl、³ 9 K, 4 ¹K、 4 ³Ca、 47 Ti、 49 Ti、 50 V. 5 ¹V、 5 ³Cr、 55 Mn, 57 Fe、 59 Co、 6 ¹Ni、 6 ³Cu、 65 Cu、 67 Zn, 69 Ga、 7 ¹Ga、 75 As、 77 Se、 79 Br、 8 ¹Br、 8 ³Kr、 85 Rb、 87 Rb、 87 Sr、 89 Y、 9 ¹Zr、 9 ³Nb、 95 Mo、 97 Mo,¹ 05 Pd,¹ 07 Ag、¹ 09 Ag, ¹¹¹Cd, ¹¹³Cd, ¹¹ 7 Sn、¹¹ 9 Sn、¹¹ 5 Sn, ¹²¹Sb, ¹³ 5 Ba、¹³ 7 Ba、¹ 77 Pb,¹ 99 Hg,² 0 ¹Hg and² 07 Pb.

54. The method according to any one of the preceding claims, wherein the step of generating the one or more NMR data portions includes recording the NMR data, including performing at least one NMR readout using an NMR spectrometer (e.g., a low-field NMR spectrometer).

55. The method of claim 54, wherein the at least one NMR readout process comprises subjecting at least one sample of the substance to a sequence of RF excitation pulses and reading the relaxation / dephase signal of at least one spin of at least one isotope.

56. The method of claim 54 or 55, wherein the at least one NMR readout process comprises subjecting at least one sample of the substance to a sequence of RF excitation pulses comprising a plurality of pulses, the plurality of pulses comprising at least one 180º RF pulse and at least one 90º RF pulse.

57. The method according to any one of claims 54-56, wherein the at least one NMR readout process comprises subjecting at least one sample of the substance to an RF excitation pulse sequence comprising a Hahn echo sequence or a CPMG sequence or a combination of at least one of these sequences.

58. The method according to any one of claims 54-57, wherein the at least one NMR readout process comprises subjecting at least one sample of the substance to an RF excitation pulse sequence comprising a spin echo sequence, a fast spin echo sequence, a stimulated echo sequence, a gradient echo sequence, or any combination thereof comprising at least one of these sequences.

59. The method according to any one of claims 54-58, wherein the at least one NMR readout process comprises subjecting at least one sample of the substance to one or more RF excitation pulse sequences comprising an inversion recovery pulse sequence, a Carr-Purcell-Meiboom-Gill (CPMG) sequence and / or a spin echo sequence, or any combination thereof comprising at least one of these sequences.

60. The method according to any one of claims 54-59, wherein the at least one NMR readout process comprises subjecting at least one sample of the substance to an RF pulse and a delay sequence comprising a subsequence SE1: - Apply the first 180º RF pulse - Apply the first inversion recovery delay (IRD). - Apply the first 90º RF pulse, and - Determine the resonance intensity directly or via echo sequences (e.g., CPMG). Preferably, the method includes repeating the SE1 subsequence with different IRD values ​​and optionally determining a zero-point IRD value, wherein the resonance intensity determined at the zero-point IRD value passes through the zero-intensity point.

61. The method according to any one of claims 54-59, wherein the at least one NMR readout process comprises subjecting at least one sample of the substance to an RF pulse and a delay sequence comprising a subsequence SE1: vi. Apply a 90º RF pulse vii. Apply echo delay viii. Apply a 180º RF pulse ix. Repeat steps ii and iii N times. x. Apply a loop delay, Preferably, the method includes repeating the SE2 subsequence, optionally applying different cyclic delays.

62. The method according to any one of claims 54-60, wherein the at least one NMR readout process comprises subjecting at least one sample of the substance to a repeated sequence of RF excitation signals, the sequence comprising Where π and π / 2 represent radio frequency pulses, the pulse duration and the radio frequency pulse field intensity are adjusted to provide pulse flip angles of π (e.g., 180º) and π / 2 (e.g., 90º), respectively, and T INV It is the inversion recovery delay (IRD), T E It is echo delay (ED), T W It is the cyclic delay (RD), and N is the number of echoes.

63. The method of claim 62, wherein the NMR data is recorded during one or more of the said delay periods.

64. The method according to any one of claims 54-63, wherein the pulse sequence is repeated multiple times, each time having a different composition of one or more parameters IRD, ED, N, RD, radio frequency pulse field strength or radio frequency pulse duration and / or flip angle.

65. The method according to any one of claims 54-64, wherein the pulse sequence is repeated multiple times, each time having a different composition and number of elements of inversion, wait, and echo periods.

66. The method according to any one of claims 54-65, wherein the at least one NMR readout process comprises a readout process according to at least any one of the preceding claims, wherein the NMR dataset comprises data from at least two different NMR readable isotopes, one of which is ¹H.

67. The method according to any one of claims 54-66, wherein the at least one NMR readout process comprises reading at least two different NMR-readable isotopes, one of which is ¹³C.

68. The method according to any one of claims 54-67, wherein the at least one NMR readout process includes reading multiple NMR readable isotopes in parallel, including initiating another pulse sequence for another NMR readable isotope during the RD, IRD, or ED period of a previous NMR readable isotope pulse sequence.

69. The method according to any one of claims 54-68, wherein the at least one NMR readout process includes performing an additional readout process over time, including obtaining an additional NMR data portion from each readout process, processing the additional NMR data portion, obtaining an additional representation component, and comparing the additional representation component with a characteristic representation to obtain information about changes in the sample.

70. The method of claim 69, wherein the information regarding the change of the substance includes changes in physical conditions, such as temperature, hydration, pressure, salt content, pH, solvent type, or concentration of the component / element in the substance.

71. The method according to any one of the preceding claims, wherein fitting of multiple NMR data portions of the sample improves measurement accuracy in terms of the standard deviation of one or more parameters, and / or reduces the data measurement time required to achieve the selected measurement accuracy.

72. The method according to any one of the preceding claims, wherein the substance comprises at least two components, preferably each component comprising a unique NMR-readable isotope.

73. The method according to any one of the preceding claims, wherein the substance is a complex substance, such as a food product, one or more samples of any stage of wastewater treatment, such as animal slurry or manure, such as biogas digester slurry, such as livestock feed, one or more samples of any stage of biorefining, such as one or more samples of any stage of mining or mineral refining processes.

74. The method according to any one of the preceding claims, wherein the substance comprises at least organic and / or inorganic components.

75. The method according to any one of the preceding claims, wherein the method includes preparing a property representation diagram of the substance.

76. The method according to any one of the preceding claims, wherein the method includes determining a mass parameter of the substance, including deriving data from the characteristic representation and determining the mass parameter.

77. The method of claim 76, wherein the mass parameter of the substance is associated with the amount of the component, such as the amount of organic component, such as the amount of lipid, such as the amount of fat, such as the amount of carbohydrate, such as the amount of protein, such as the amount of nucleic acid, such as the amount of chemical reagent, such as the amount of isotope, such as the amount of mobile sample component, such as the amount of rigid / immobile sample component, such as the amount of species undergoing chemical exchange, such as the amount of species undergoing molecular, ionic or spin diffusion.

78. The method according to claim 76 or 77, wherein the mass parameter of the substance is associated with the chemical properties of the substance, such as pH value.

79. The method according to any one of the preceding claims, wherein the characteristic representation includes a plurality of identifiers, each identifier representing a statistical distribution of the sample components.

80. The method according to any one of the preceding claims, wherein the method comprises comparing a characteristic representation of the substance with at least one historical characteristic representation to identify differences, such as chemical and / or biological and / or compositional differences between the substance and substantially similar substances.

81. The method of claim 80, wherein the method includes determining a quality parameter based on the identified differences.

82. The method according to claim 80 or 81, wherein the at least one historical characteristic representation comprises at least one historical characteristic representation of at least one reference material, or at least one historical characteristic representation synthesized from two or more historical characteristic representations of at least one reference material.

83. The method of claim 82, wherein the reference substance comprises at least 25% of the component present in the substance, for example, at least 50% of the component present in the substance, for example, at least 90% of the component present in the substance, and the at least one reference substance corresponds to the substance.

84. The method according to any one of the preceding claims, wherein the processing is performed by a trained computer, wherein the trained computer is trained to determine one or more quality parameters and one or more sample identifiers, preferably to determine quantitative parameters and / or to monitor differences in historical data, such as differences in chemical, physical or biological processes, preferably the trained computer is trained to classify based on one or more quality parameters and / or based on one or more sample identifiers.

85. The method according to any one of the preceding claims, wherein the processing relates to supervised, unsupervised or any other form of artificial intelligence, machine learning, reinforcement learning or similar techniques.

86. A method for generating a trained computer, comprising subjecting the computer to a machine learning process, the process including feeding the computer data, wherein the data includes packets of previously obtained NMR data, the packets being associated with at least one subset of fitted data comprising T1*, associated T2*, and associated intensity (M0).

87. The method for generating a trained computer according to claim 86, wherein the data packet includes data representing a property representation of a substance and an NMR data portion applied to generate the property representation.

88. A trained computer that can be obtained by the method according to claim 86 or 87.

89. An NMR system for performing the method according to any one of the preceding claims, wherein the NMR system includes a computer system configured to receive one or more nuclear magnetic resonance (NMR) data portions of the substance, wherein the NMR data portions include a plurality of NMR datasets, each NMR dataset including intensity data and relaxation data, and The computer system described therein is configured to: - The NMR data portion is processed into a multidimensional distribution of the dataset by performing multiple fittings to at least one function, wherein the function includes at least two fitting parameters and is a function representing a model of the obtained NMR data portion, and - Generate a property representation of the substance from the fitted dataset.

90. The NMR system of claim 89, wherein the NMR system is configured to perform the method of any one of claims 1-87.

91. The NMR system of claim 89 or 90, wherein the computer system includes a separate computer portion adapted for wireless data communication, preferably via the Internet.

92. The NMR system according to any one of claims 89-91, wherein the NMR system includes an NMR spectrometer, preferably the NMR spectrometer is in data communication with the computer system.

93. The NMR system according to any one of claims 89-92, wherein the computer system comprises a supercomputer system, a computer cluster, a computer system having multiple CPUs, a computer system having multiple GPUs, and is preferably located at a distance relative to the NMR spectrometer.

94. The NMR system according to any one of claims 89-93, wherein the computer system comprises a computer trained according to any one of claims 84-88.

Images