Calibration of a multidimensional spectrometer

Abstract

The present invention provides a spectrometer support system, as well as related methods, computing devices, and computer-readable media. [Solution] For example, in some embodiments, the spectrometer support device may receive an array of spectrometer output intensities for each of a plurality of calibration samples of analytes at different known concentrations, wherein different spectrometer output intensities in the array are associated with data representing different deflection amounts, and a machine learning computational model may be trained using a plurality of arrays associated with multiple known concentrations of analytes in the calibration samples and spectrometer output intensities to output the concentration of analytes in the target sample based on an input array of spectrometer output intensities of the target sample, and the trained machine learning computational model may be used as a calibration model for the analyte for subsequent spectrometer operation.

Description

Background Art

[0001] Many scientific instruments require calibration, i.e., an association between the output of the scientific instrument and a known state or property. A spectrometer, for example, can output an intensity that is a function of the properties of a sample, and calibration of such a spectrometer can define the relationship between the output intensity and the properties of the sample.

Summary of the Invention

Means for Solving the Problems

[0002] Disclosed herein are a spectrometer assistance system, as well as related methods, computing devices, and computer-readable media. For example, in some embodiments, a spectrometer assistance device may generate an array of spectrometer output intensities for each of a plurality of calibration samples of an analyte at different known concentrations, where different ones of the spectrometer output intensities within the array are associated with data representing different polarization vectors, and use the associated plurality of arrays of the plurality of known concentrations of the analyte in the calibration samples and the spectrometer output intensities to train a machine learning computational model to output the concentration of the analyte in a target sample based on an input array of the spectrometer output intensities of the target sample, and the trained machine learning computational model may be used as a calibration model for the analyte for subsequent spectrometer operations.

[0003] The spectrometer-assisted embodiments disclosed herein can achieve improved performance compared to conventional approaches. In conventional spectroscopic measurements, the user is required to identify a single deflection (representing the order of diffraction and wavelength in optical spectroscopy, or the mass-to-charge ratio in mass spectrometry) and use that single deflection to determine the concentration of an analyte in a sample. For example, a user of optical spectroscopy may analyze an unknown sample using a spectrometer and obtain an output intensity signal with peaks at different orders of diffraction and wavelengths, and then be required to select a single order of diffraction and wavelength on which the output intensity has a peak for use in determining the concentration of an analyte associated with that order of diffraction and wavelength. However, an analyte (e.g., a single element) is typically associated with multiple peaks in the output intensity signal, and any information provided by these additional peaks is conventionally discarded. Rule-based algorithms for determining which individual peaks should be used to determine the analyte concentration often fail because they cannot account for all analytical conditions and samples. Several attempts have been made to determine the concentration of the relevant analyte by using the average or sum of the output intensities at different deflection values ​​(e.g., diffraction order / wavelength or mass-to-charge ratio), but these attempts have failed to achieve a significant improvement over the "single" deflection value approach.

[0004] Embodiments disclosed herein enable spectrometer-assisted devices to utilize far more information about the spectrometer's output intensity signal than previously used when determining analyte concentrations, thus providing improved accuracy to spectrometer technology (e.g., improvements in computer technology assisting such spectrometers, among other improvements). Furthermore, by reducing the need for expert users who can selectively identify specific deflection amounts that depend on determining specific concentrations, embodiments disclosed herein enable non-expert users to determine analyte concentrations in samples without difficulty, increasing accuracy, dynamic range, and throughput, and reducing costs.

[0005] Various embodiments disclosed herein can improve upon conventional approaches to achieve the technical advantage of more accurate determination of analyte concentrations in a sample by constructing calibration models that utilize more output intensity signals in concentration determination. Such technical advantages cannot be achieved with routine conventional approaches, and all users of systems including such embodiments benefit from these advantages (for example, by assisting the user in performing technical tasks, such as determining analyte concentrations in a sample, through guided human-machine interaction processes). For example, various embodiments disclosed herein can enable the performance of calibration-less semi-quantitative analysis, which differs from conventional approaches. Thus, the technical features of the embodiments disclosed herein, as well as combinations of the features of the embodiments disclosed herein, are clearly unconventional in the field of spectroscopic measurement. The computational and user interface features disclosed herein involve not only the collection and comparison of information, but also the application of new analytical and technical techniques to modify the operation of spectroscopic measurement support systems. Thus, this disclosure introduces functions that neither conventional computing devices nor humans have been able to perform.

[0006] Accordingly, embodiments of the present disclosure may serve any of several technical purposes, such as controlling the determination of analyte concentrations in a particular technical system or process (e.g., a spectroscopic measurement system or process), and determining the characteristics of a sample by processing data obtained from a spectroscopic measurement sensor.

[0007] In some embodiments, the spectrometer support device may include: first logic for generating an array of spectrometer output intensities for each of a plurality of calibration samples of analytes at different known concentrations, wherein different spectrometer output intensities in the array are associated with data representing different deflection amounts; second logic for training a machine learning computational model using a plurality of arrays associated with a plurality of known concentrations of analytes in the calibration samples and spectrometer output intensities, and outputting the concentration of analytes in the target sample based on an input array of spectrometer output intensities for the target sample; and third logic for using the machine learning computational model trained as a calibration model for the analyte for subsequent spectrometer operation.

[0008] In some embodiments, the spectrometer support device may include: a first logic for generating an array of spectrometer output intensities of a sample, wherein different spectrometer output intensities in the array are associated with data representing different deflection amounts; a second logic for providing the received array of spectrometer output intensities to a trained machine learning computational model, wherein the trained machine learning computational model is for outputting the concentration of an analyte in the sample; and a third logic for outputting the concentration of an analyte in the sample.

[0009] In some embodiments, the spectrometer support device may include: a first logic for generating an array of spectrometer output intensities of a sample, wherein different spectrometer output intensities in the array are associated with data representing different deflection amounts; a second logic for generating concentrations of analytes in the sample based on the received array of spectrometer output intensities; and a third logic for outputting the concentrations of analytes in the sample and feature association indicators associated with one or more of the spectrometer output intensities.

[0010] In some embodiments, the spectrometer output device may include: a first logic for generating an array of spectrometer output intensities of a sample, wherein different spectrometer output intensities in the array are associated with data representing different deflection amounts; a second logic for generating the concentration of an analyte in a sample based on the received array of spectrometer output intensities, without requiring the user to select one or more of the data representing deflection amounts prior to generating the concentration of an analyte in the sample; and a third logic for outputting the concentration of an analyte in the sample.

[0011] In some embodiments, the spectrometer support device may include: first logic for generating an array of spectrometer output intensities of a calibration sample for each of a plurality of calibration samples of analytes at different known concentrations, wherein different spectrometer output intensities in the array are associated with data representing different deflection amounts; second logic for generating a preprocessing method for retraining a pretrained machine learning computation mode or for use with a pretrained machine learning computation model using a plurality of arrays associated with multiple known concentrations of analytes in the calibration sample and spectrometer output intensities, and outputting the concentration of analytes in the target sample based on an input array of spectrometer output intensities of the target sample; and third logic for running the preprocessing method using the retrained machine learning computation model or with a pretrained machine learning computation model as a calibration model of the analyte for subsequent spectrometer operation.

[0012] These and other embodiments disclosed herein can solve one or more of the technical problems of conventional spectroscopic measurements, such as the inability to properly identify analyte concentrations during actual operation and the technical problems of insufficient calibration that require a specialized operator to perform, by constructing, using, and transferring calibration models that use more output intensity signals than conventional approaches in concentration determination.

[0013] The embodiments will be readily apparent from the following detailed description in conjunction with the accompanying drawings. For the sake of this description, similar reference numerals refer to similar structural elements. The embodiments are illustrated in the figures of the accompanying drawings as examples, not as limitations. [Brief explanation of the drawing]

[0014] [Figure 1] This is a block diagram of a spectroscopic measurement system configured to perform or facilitate the support operations disclosed herein, according to various embodiments. [Figure 2] This figure shows a detector array in which an image of an echelle spectrum is formed in a spectroscopic measurement system according to various embodiments. [Figure 3] This is a block diagram of an exemplary spectrometer support module for performing support operations according to various embodiments. [Figure 4] This figure shows an exemplary array of spectrometer output intensities that may be used by the spectrometer support modules disclosed herein, according to various embodiments. [Figure 5] This figure shows an exemplary array of spectrometer output intensities that may be used by the spectrometer support modules disclosed herein, according to various embodiments. [Figure 6] This figure shows an exemplary array of spectrometer output intensities that may be used by the spectrometer support modules disclosed herein, according to various embodiments. [Figure 7] This figure shows an exemplary array of spectrometer output intensities that may be used by the spectrometer support modules disclosed herein, according to various embodiments. [Figure 8] This figure shows an exemplary array of spectrometer output intensities that may be used by the spectrometer support modules disclosed herein, according to various embodiments. [Figure 9] This figure shows an exemplary array of spectrometer output intensities that may be used by the spectrometer support modules disclosed herein, according to various embodiments. [Figure 10]FIG. Illustrative arrays of spectrometer output intensities that may be used by a spectrometer assistance module disclosed herein according to various embodiments. [Figure 11] FIG. Machine learning calculation models that may be included in a spectrometer assistance module according to various embodiments. [Figure 12] FIG. Sets of training data that may be used to train machine learning calculation models included in a spectrometer assistance module according to various embodiments. [Figure 13] FIG. Flow diagram of an exemplary method of performing a spectrometer assistance operation according to various embodiments. [Figure 14] FIG. Flow diagram of an exemplary method of performing a spectrometer assistance operation according to various embodiments. [Figure 15] FIG. Flow diagram of an exemplary method of performing a spectrometer assistance operation according to various embodiments. [Figure 16] FIG. Flow diagram of an exemplary method of performing a spectrometer assistance operation according to various embodiments. [Figure 17] FIG. Flow diagram of an exemplary method of performing a spectrometer assistance operation according to various embodiments. [Figure 18] FIG. An example of a graphical user interface that may be used in some or all implementations of the assistance methods disclosed herein according to various embodiments. [Figure 19] FIG. Block diagram of an exemplary computing device that may perform some or all of the spectrometer assistance methods disclosed herein according to various embodiments. [Figure 20] FIG. Block diagram of an exemplary spectrometer assistance system that may perform some or all of the spectrometer assistance methods disclosed herein according to various embodiments.

DETAILED DESCRIPTION OF THE INVENTION

[0015] In the following detailed description, reference is made to the accompanying drawings which form a part hereof, and like numerals designate like parts throughout. By way of example, embodiments which may be implemented are shown. It is to be understood that other embodiments may be utilized and structural or logical changes may be made without departing from the scope of the present disclosure. Accordingly, the following detailed description should not be construed in a limiting sense.

[0016] Various operations may be described sequentially as a number of discrete actions or operations in a manner that is most helpful in understanding the claimed subject matter. However, the order of description should not be construed as to imply that these operations are necessarily order dependent. Specifically, these operations may not be performed in the order of presentation. The described operations may be performed in an order different than the described embodiments. Various additional operations may be performed and / or the described operations may be omitted in additional embodiments.

[0017] Figure 1 is a block diagram of a spectroscopic measurement system 10 configured to perform or facilitate the assistive operations disclosed herein, according to various embodiments. In particular, Figure 1 shows an optical spectroscopic measurement system 10, but embodiments disclosed herein may also be used in conjunction with other types of spectroscopic measurement systems, such as mass spectrometry measurement systems. The optical spectroscopic measurement system 10 in Figure 1 may include a light source 11, an optical device 12, a detector array 13, a processor 14, a memory 15, and an input / output (I / O) unit 16. The light source 11 may be a plasma source, such as an inductively coupled plasma (ICP) source. The optical device 12 may include an echelle diffraction grating and a prism (and / or further diffraction gratings) to generate an echelle spectrum of light produced by the light source 11. An image of the two-dimensional echelle spectrum may be formed on the detector array 13. Such an image is described further below with reference to Figure 2. The detector array 13 may be, for example, a charge-coupled device (CCD) array. The detector array 13 includes an array of detector elements or pixels that generate an output signal representing the detected spectral value, and in some embodiments, the detector array 13 may have at least about 1024 × 1024 pixels (1 megapixel). The rectangular detector array 13 may be square, but is not necessarily required.

[0018] The detector array 13 may be configured to generate spectral values ​​corresponding to the detected light intensity of the echelle spectrum and transfer the spectral values ​​to the processor 14. The processor 14 may include one or more commercially available processing devices, such as one or more commercially available microprocessors (e.g., one or more of the processing devices 4002 discussed below with reference to Figure 19). The memory 15 may include one or more suitable storage devices, such as one or more suitable semiconductor memory devices (e.g., any of the storage devices 4004 discussed below with reference to Figure 19), and may be used to store non-transient, computer-readable instructions that, when executed by the processor 14, cause the spectroscopic measurement system 10 to perform one or more embodiments of the methods disclosed herein. The I / O unit 16 may include any suitable circuitry (e.g., one or more of the interface device 4006, the display device 4010, or other I / O devices 4012 discussed below with reference to Figure 19) and may be configured to input data or commands to the spectroscopic measurement system 10, output data from the spectroscopic measurement system 10, and / or enable communication between the spectroscopic measurement system 10 and other instruments or computing devices. Some or all of the components of the spectroscopic measurement system 10 may implement together the spectrometer support module 1000 disclosed herein (e.g., as discussed below with reference to Figure 3), or the spectrometer support module 1000 may be implemented by another set of hardware and / or software components and communicate with the spectroscopic measurement system 10 via the I / O unit 16.

[0019] Figure 2 shows a detector array 13 in which an image of an echelle spectrum 20 is formed in a spectroscopic measurement system 10 according to various embodiments. The echelle spectrum 20 is shown in Figure 2 to include diffraction order 7 that extends individually and substantially horizontally. That is, diffraction order 7 extends substantially in a first direction of the detector array 13, which may be called the x-direction in the example of Figure 2. Thus, diffraction order 7 extends substantially perpendicular to a second direction of the detector array 13 that is perpendicular to the first direction, which may be called the y-direction. Since diffraction order 7 of the echelle spectrum 20 is usually slightly curved, the degree to which diffraction order 7 is parallel or perpendicular to the first and second directions may vary across the echelle spectrum.

[0020] In the example shown in Figure 2, the first direction (x-direction) is parallel to the long side of the rectangular detector array 13, while the second direction (y-direction) is parallel to the short side. The orientation of the detector array is selected to best fit the two-dimensional spectrum, and it will be understood that the terms first direction and second direction are interchangeable.

[0021] Diffraction order 7 is a region of high light intensity, resulting in high spectral values. Diffraction order 7 is separated by valleys or valleys 8 of low light intensity, and therefore low spectral values. The echelle spectrum 20 typically has one or more spectral value peaks that are characteristic of a particular substance. For example, when using ICP as the light source 11 to generate the echelle spectrum 20, there is typically a peak representing the presence of carbon dioxide. Figure 2 schematically shows the first peak 1 and the second peak 2. In an actual echelle spectrum 20, there will usually be three or more peaks. Each peak is located at diffraction order 7 and, at least locally, constitutes the maximum value of that diffraction order 7. It can be seen that each peak extends in both the first direction (x direction in Figure 2) and the second direction (y direction in Figure 2). Note that in a typical embodiment, the peaks may have a length and width of only a few pixels, e.g., 3-5 pixels.

[0022] Different substances will generate peaks at various positions in the optical spectrum. As described above, a single peak in the optical spectrum has conventionally been used to identify an analyte (e.g., a single element) in a sample being tested by the spectroscopic measurement system 10. However, an analyte is typically associated with multiple peaks in the output intensity signal, and any information provided by these additional peaks has conventionally been discarded. As will be discussed further below, embodiments disclosed herein enable a spectrometer support device (which may be implemented by the spectroscopic measurement system 10 or by another system communicating with the spectroscopic measurement system 10) to utilize far more information about the spectrometer's output intensity signal than previously used in determining the analyte concentration, and thus provide improved accuracy to spectrometer technology (e.g., improvements in computer technology supporting such a spectrometer, among other improvements).

[0023] Figure 3 is a block diagram of a spectrometer support module 1000 for performing support operations in various embodiments. The spectrometer support module 1000 may be implemented by a circuit (including, for example, electrical and / or optical components) such as a programmed computing device. The logic of the spectrometer support module 1000 may be contained in a single computing device or distributed across multiple computing devices communicating with each other as needed. Embodiments of computing devices that can implement the spectrometer support module 1000, either individually or in combination, are considered with reference to the computing device 4000 in Figure 19, and embodiments of a system of interconnected computing devices in which the spectrometer support module 1000 can be implemented across one or more computing devices are considered herein with reference to the spectrometer support system 5000 in Figure 20. The spectrometer support module 1000 may be implemented, for example, by the spectroscopic measurement system 10 of Figure 1 (for example, by combining some or all of the processor 14, memory 15, and I / O unit 16), by another spectroscopic measurement system 10, or by a computing system that communicates with a spectroscopic measurement system such as the spectroscopic measurement system 10 of Figure 1.

[0024] The spectrometer support module 1000 may include spectrometer intensity logic 1002, training logic 1004, analyte concentration logic 1006, and output logic 1008. As used herein, the term “logic” may include devices that perform a set of operations associated with the logic. For example, any of the logic elements included in the spectrometer support module 1000 may be implemented by one or more computing devices programmed with instructions that cause one or more processing devices of a computing device to perform a set of operations associated with that device. In certain embodiments, a logic element may include one or more non-temporary computer-readable media having instructions that cause one or more computing devices to perform a set of operations associated with that device when executed by one or more processing devices of the computing device. As used herein, the term “module” may refer to a collection of one or more logic elements that together perform a module and associated functions. Different logic elements within a module may take the same form or different forms. For example, some logic within a module may be implemented by a programmed general-purpose processing device, while other logic within a module may be implemented by an application-specific integrated circuit (ASIC). In another embodiment, different logic elements within a module may be associated with different sets of instructions executed by one or more processing devices. A module may not contain all of the logic elements shown in the relevant drawings; for example, a module may contain a subset of the logic elements shown in the relevant drawings when the module performs a subset of the operations considered herein with reference to that module.

[0025] The spectrometer intensity logic 1002 may generate an array of spectrometer output intensities based on data output by the spectrometer while the spectrometer is analyzing a sample. During analysis, the spectrometer may separate spectral components associated with analytes present in the sample (e.g., the elemental composition of the sample) by deflecting these spectral components differently on the detector, and the intensities measured by the detector for different deflections may provide a signature of one or more analytes present in the sample. For example, as described above with reference to Figures 1 and 2, an optical spectrometer may include a diffraction grating or other optical element that deflects radiation of different wavelengths differently at one or more positions on the detector such that the intensity of incident radiation at a position on the detector represents the intensity associated with a particular wavelength of radiation. In another example, a mass spectrometer may include an electric field and / or magnetic field that deflects ions of different mass-charge ratios differently at positions on the detector such that the intensity of incident ions at a position on the detector represents the intensity associated with a particular mass-charge ratio. Accordingly, for any sample analyzed by the spectrometer, the spectrometer intensity logic 1002 may generate an array of spectrometer output intensities associated with the sample, with different spectrometer output intensities in the array associated with data representing different deflection quantities (e.g., diffraction order / wavelength or mass-to-charge ratio). As used herein, “deflection quantity” includes dispersion quantity, as appropriate for a given spectroscopic measurement system. Embodiments disclosed herein may be used with any characteristic of the spectrometer output that represents multiple characteristic observables acting as proxies for the identity of the analyte in the sample, as well as deflection quantities.

[0026] Figures 4 to 10 show exemplary sequences of spectrometer output intensities that may be used by the spectrometer support module 1000 in various embodiments. In particular, any of the spectrometer output intensity sequences shown in Figures 4 to 10 may be generated by the spectrometer intensity logic 1002 based on intensity data provided to the spectrometer intensity logic 1002 by the spectrometer. The spectrometer output intensity sequences generated by the spectrometer intensity logic 1002 may be background-corrected intensities, and background correction is performed according to any preferred technique (e.g., any preferred background correction technique known in the art). Figures 4 to 10 and other accompanying drawings may illustrate the spectrometer output intensity as a function of diffraction order / wavelength (as appropriate for optical spectroscopic measurements), but this is merely for the sake of clarity, and the spectrometer output intensity may also be a function of other parameters representing the deflection amount (e.g., a mass-to-charge ratio appropriate for mass spectrometry measurements).

[0027] Figure 4 shows the array of spectrometer output intensities 102 in the form of a plot of intensity as a function of wavelength. The array of spectrometer output intensities 102 shown in Figure 4 may include several peaks at specific wavelengths (e.g., wavelengths labeled WL1, WL2, ..., WL8). The wavelengths at which peaks occur in the array of spectrometer output intensities 102 may be a function of the analytes (e.g., elements) present in the sample under spectrometer analysis, and the magnitude of the associated intensities at these peaks may indicate the concentration of the analyte in the sample. For a single-element sample, the wavelength positions of these peaks (referred to herein as “peak wavelengths”) may be characteristic of that element (e.g., in emission spectroscopy, peak wavelengths associated with aluminum may include 185.580 nanometers, 220.462 nanometers, etc.). Figure 5 shows the same array of spectrometer output intensities 102 as shown in Figure 4, but in a "heatmap" format where the peaks in the array of spectrometer output intensities 102 in Figure 4 are represented as having "brighter" grayscale values ​​at positions associated with wavelengths WL1, WL2, ..., WL8, as shown.

[0028] The spectrometer intensity logic 1002 may generate a sequence of spectrometer output intensities 102 for each of several different samples. In particular, during calibration, the spectrometer intensity logic 1002 may generate a sequence of spectrometer output intensities 102 for each of several calibration samples, where different calibration samples have different known concentrations of the analyte of interest (e.g., the element of interest). For example, during calibration, the spectrometer may be provided with different single-element solution calibration samples having different known concentrations of molybdenum or another element of interest, and the spectrometer intensity logic 1002 may generate a different sequence of spectrometer output intensities 102 for each of these calibration samples. Figure 6 shows a set of five such sequences of spectrometer output intensities 102 (labeled 102-1, 102-2, ..., 102-5) associated with five different concentrations (labeled C1, C2, ..., C5) of the analyte of interest in the corresponding calibration samples. In the specific example shown in Figure 6, the concentration of the analyte of interest can increase from C1 to C5, and it should be noted that as the concentration of the analyte of interest increases, the magnitude of the intensity in the corresponding sequence of spectrometer output intensities 102 increases.

[0029] As shown in Figures 4 and 5, the array of spectrometer output intensities 102 may be represented in one of several ways. For example, Figure 7 shows a set of five arrays of spectrometer output intensities 102 (labeled 102-1, 102-2, ..., 102-5) associated with five different concentrations (labeled C1, C2, ..., C5) of the analyte of interest in the corresponding calibration sample, including five wavelengths (labeled W1, W2, ..., W5) at which intensity peaks occur. Figure 8 shows the same set of five arrays of spectrometer output intensities 102 (labeled 102-1, 102-2, ..., 102-5) associated with five different concentrations (labeled C1, C2, ..., C5) of the analyte of interest in the corresponding calibration sample, but here only the peak magnitude is of interest and is represented as a set of peak intensity magnitudes at each of the five wavelengths (labeled W1, W2, ..., W5). For example, the array of spectrometer output intensities 102-5 in Figure 8 shows a magnitude of approximately 50 units for the peak associated with wavelength WL1, and a magnitude of approximately 80 units for the peak associated with wavelength WL2, and so on. Thus, Figure 8 shows one way in which the array of spectrometer output intensities 102 can be specified (i.e., by the magnitude of the peak for each associated peak wavelength).

[0030] Figures 9 and 10 show another way in which the array of spectrometer output intensities 102 can be represented. Figure 9 is a replica of Figure 7 (for clarity of explanation) which shows a set of five arrays of spectrometer output intensities 102 (labeled 102-1, 102-2, ..., 102-5) associated with five different concentrations (labeled C1, C2, ..., C5) of the analyte of interest in the corresponding calibration sample, including five wavelengths (labeled W1, W2, ..., W5) at which intensity peaks occur. Figure 10 shows the same set of five sequences of spectrometer output intensities 102 (labeled 102-1, 102-2, ..., 102-5) associated with five different concentrations (labeled C1, C2, ..., C5) of the analyte of interest in the corresponding calibration sample, but here only the relative peak magnitudes are of interest and are expressed as the aggregate of the relative intensity magnitudes of the peaks at each of the five wavelengths (labeled W1, W2, ..., W5). For example, the sequence of spectrometer output intensities 102-5 in Figure 10 shows that approximately 25% of the total aggregate intensity magnitude of the peaks associated with wavelengths WL1, WL2, ..., WL5 is provided by the intensity magnitude of the peak associated with WL1, and approximately 72% of the total aggregate intensity magnitude of the peaks associated with wavelengths WL1, WL2, ..., WL5 is provided by the intensity magnitude of the peak associated with wavelength WL2. Therefore, Figure 10 shows another way in which the array of spectrometer output intensities 102 can be specified (i.e., by the relative peak magnitude for each of the relevant peak wavelengths).

[0031] The training logic 1004 may use an array of spectrometer output intensities 102 of a calibration sample, along with a known concentration of the analyte of interest in the calibration sample, to train a machine learning computational model to output the concentration of the analyte of interest in the sample based on the input array of spectrometer output intensities 102 of the sample. In particular, the training logic 1004 may adjust the parameters of an untrained or pre-trained machine learning computational model according to known training techniques so that when an array of spectrometer output intensities 102 associated with a calibration sample of known concentration is input to the machine learning computational model, the output of the machine learning computational model is equal to or close to the value of the known concentration. Thus, such a trained machine learning computational model may be used as a calibration model of the analyte of interest for subsequent spectrometer operation, relating the spectrometer intensity output to the analyte concentration.

[0032] Figure 11 shows a diagram of a machine learning computation model 110 that can be trained by training logic 1004 using calibration data. The number of nodes in the input layer of the machine learning computation model 110 may be equal to the dimension of the tensor provided to the machine learning computation model 110, and the number of exemplary tensors that may be provided to the machine learning computation model 110 is described below with reference to Figure 12. The number of hidden layers in the machine learning computation model 110, the number of nodes in each hidden layer, and the connectivity between layers may take any suitable values. For example, in some embodiments, the machine learning computation model 110 may include eight hidden layers, 12 to 128 nodes in each hidden layer, and full (also called "high-density") connectivity between layers. In some specific embodiments, the number of nodes in a particular layer increases with respect to the first layer (e.g., increases between 12 and 128 in the first six layers) and decreases with respect to the last layer (e.g., decreases from 128 to 1 in the output layer). The activation function used between layers, the error function used to train the machine learning computation model 110, and the training technique itself may be selected as suitable. For example, in some embodiments, the activation function used between layers may be normalized linear (ReLU), the error function used to train the machine learning computation model may be the mean standard error (MSE), and the training technique used to train the machine learning computation model 110 may be gradient-based. In some embodiments, the number of output nodes in the machine learning computation model 110 may be 1 (corresponding to the concentration of the analyte of interest in the sample of interest, as discussed above). As is known in the art, the training data (i.e., an array of spectrometer output intensities 102 of the calibration sample, along with the known concentrations of the analyte of interest in the calibration sample) may be normalized, encoded / decoded, or otherwise processed as part of training and using the machine learning computation model 110.

[0033] As described above, the array of spectrometer output intensities 102 input to the machine learning computation model 110 may take any of a number of forms. For example, Figure 12 shows a set of training data 112 that may be generated by spectrometer intensity logic 1002 and used by training logic 1004 to train the machine learning computation model 110, according to various embodiments. The training data 112 may include a set of arrays of spectrometer output intensities 102 (labeled as 102-1, 102-2, ..., 102-N) and their associated concentrations (labeled as C1, C2, ..., CN). Each array of spectrometer output intensities 102 may include a first sub-array 102A that includes the magnitude of the spectrometer output intensity at each of a plurality of peak wavelengths (labeled as WL1, WL2, ..., WLM) associated with the analyte of interest. In some embodiments, the individual arrays of spectrometer output intensities 102 may also include a second sub-array 102B that includes a pairwise ratio of the magnitudes of the spectrometer output intensities at different peak wavelengths (e.g., the magnitude of the intensity associated with peak wavelength WL1 divided by the magnitude of the intensity associated with peak wavelength WL2, denoted as WL1 / WL2, etc.). For analytes where the relative magnitude associated with peak wavelengths is consistent across different concentrations of the analyte of interest (as described above with reference to Figure 10), including the ratio of different peak magnitudes in the tensor input to the machine learning computational model 110 can help train the machine learning computational model 110 to recognize the relevance of this information more quickly when determining the concentration of the analyte of interest in a sample. In some embodiments, the array of spectrometer output intensities 102 input to the machine learning computational model 110 may include only the first sub-array 102A and may not include the second sub-array 102B, or may include other representations of the array of spectrometer output intensities 102 (e.g., other functions or combinations of magnitudes of intensity data at different peak wavelengths and / or non-peak wavelengths).Furthermore, the array of spectrometer output intensities 102 input to the machine learning computation model 110 does not necessarily represent the magnitude of the intensity at all peak wavelengths associated with the analyte of interest, but may represent the magnitude of the intensity at a subset of peak wavelengths.

[0034] The training logic 1004 may train different machine learning computation models 110 for different analytes of interest. For example, the training logic 1004 may use calibration data for each of a plurality of single-element samples to generate a plurality of associated machine learning computation models 110, each associated with a different specific element. In other embodiments, a single machine learning computation model 110 may be trained to generate concentrations of multiple analytes of interest based on an array of spectrometer output intensities 102, in which case the number of output nodes of the machine learning computation model 110 may be equal to the number of analytes whose concentrations can be determined by the machine learning computation model 110. Various embodiments of the embodiments disclosed herein may be described by referring to a single analyte of interest associated with a single machine learning computation model 110, but this is merely for the sake of illustration, and any of the techniques disclosed herein can use a single machine learning computation model 110 to generate concentrations of multiple analytes of interest.

[0035] In some embodiments, the training logic 1004 may use training data containing one or more saturated spectrometer output intensities to train the machine learning computational model 110. Conventionally, peak wavelengths where the spectrometer output intensity saturates (i.e., reaches the upper limit of intensity that can be separated by the detector) are discarded during subsequent analysis. However, the techniques disclosed herein allow intensity data associated with multiple peak wavelengths to be used together to determine the analyte concentration, and therefore, allowing the training logic 1004 to use some training data containing saturated intensities can help the machine learning computational model 110 contextualize such data and, when input, rely more heavily on unsaturated intensities to make an appropriate concentration determination. Since the analyte concentration determination techniques disclosed herein can appropriately determine the concentration even when saturation occurs (and when there are no low-sensitivity peaks from the intensity signal representing a low-concentration sample), the spectrometer support module 1000 disclosed herein can significantly increase the dynamic range of the spectrometer compared to conventional approaches. In some embodiments, the training data may be preprocessed by the training logic 1004 before being used to train the machine learning computational model, removing some or all of the spectrometer output intensities that are saturated or otherwise abnormal. For example, the training logic 1004 may preprocess the training data by performing an initial linearity check, during which it may compare the magnitudes of peaks associated with different samples to determine whether the ratio of peak magnitudes is approximately equal to the ratio of the concentrations of the associated analytes in the samples, as expected based on the laws of physics. If one or more peaks fail this linearity check (for example, due to saturation, insufficient intensity, or interference), those peaks may be discarded from the set of data used to train the machine learning computational model.

[0036] In some embodiments, the training logic 1004 may retrain the machine learning computation model 110. For example, if the spectrometer calibration is re-performed for a particular analyte, the training logic 1004 may retrain the pre-trained machine learning computation model 110 using the new calibration data (for example, to compensate for drift or other changes since the previous calibration and / or to improve the quality of the calibration by using more data). In another example, a machine learning computation model 110 trained to output the concentrations of one or more specific analytes in a particular spectrometer may be retrained by the training logic 1004 to output the concentrations of the analytes in a different spectrometer. When retraining the machine learning computation model 110 for a different spectrometer, the training logic 1004 may use a transfer learning technique in which only the last fully connected layer of the pre-trained machine learning computation model 110 (or another subset of the parameters of the pre-trained machine learning computation model 110) is retrained, while other parameters remain fixed. Utilizing such techniques can reduce the training burden associated with building machine learning computation models 110 for other spectrometers while still achieving customization of the machine learning computation model 110 for a specific spectrometer at hand. The transfer learning approach may also take into account the operating conditions of different instruments (e.g., acquisition settings), which may be used in combination with the aforementioned training data (which may cover all elements of interest across a concentration range that covers the instrument's dynamic range) to build a complete model of the behavior of the first instrument, which may then be readily deployed on the second instrument by performing a set of linear transformations of the complete model from the first instrument.

[0037] In some embodiments, the training logic 1004 may perform a transfer learning technique that does not require retraining of the machine learning computation model 110. The training logic 1004 may be configured to perform such a technique, for example, when the internal or operating conditions of the instrument change (which may result in a change in the measured intensity for the same concentration of a given analyte), or when deploying the machine learning computation model 110 to a second instrument with different operating conditions. In both such cases, the training logic 1004 may renormalize the intensity before the machine learning computation model 110 is used to output the concentration data. To renormalize the intensity, the training logic 1004 may utilize a new array of intensity representing at least one known concentration of the selected analyte measured by the instrument. These concentrations do not need to be predetermined and can be selected, for example, by the instrument user or service technician. In some embodiments, the concentrations may be selected to fall within a linear regime of the calibration curve for the selected analyte, but in other embodiments, the concentrations do not need to fall within such a linear regime (for example, when the linear regime changes as the machine learning computation model 110 is retrained with more data). The training logic 1004 may generate a set of normalization parameters by using an existing trained machine learning computation model 110 to generate a function of the following form: I[I1,I2,I3,...,In]=f(concentration), This returns an array of intensities (I1, I2, I3, ..., In), one for each observed spectral emission of a particular element, associated with a given concentration value in the same normalized intensity-concentration space used to train the machine learning computational model 110. This function may take any of a number of forms; for example, the function may be represented as a lookup table containing tuple values ​​of (intensity, concentration) evaluated over a concentration span (which can cover the entire dynamic range of the instrument), or, among other forms, in particular, as a machine learning-based function trained to map concentration values ​​back to an array of intensities. Using this function, the training logic 1004 may transform the measured array of intensities into the normalized intensity-concentration space of the machine learning computational model 110. The analyte concentration logic 1004 may then use the transformed intensity array together with the existing machine learning computational model 110 to predict the concentration of a given analyte in a sample.

[0038] The analyte concentration logic 1006 may use a trained machine learning computation model 110 as a calibration model during subsequent spectrometer operation. In particular, when the spectrometer is used to analyze a sample in which the analyte concentration is unknown, the spectrometer intensity logic 1002 may generate an array of spectrometer output intensities 102 associated with the sample, and may provide this array of spectrometer output intensities 102 to the trained machine learning computation model 110, which then outputs the analyte concentration in the sample. Thus, in contrast to conventional approaches, the spectrometer support module 1000 can generate the analyte concentration prior to its generation without requiring the user to select other specific data representing diffraction order / wavelength, mass-to-charge ratio, or deflection, thereby reducing the burden on the user and enabling problem-free operation of the spectrometer achievable by non-expert users.

[0039] In some embodiments, the analyte concentration logic 1006 may also generate one or more feature relevance indicators associated with the analyte concentrations output by the machine learning computation model. The feature relevance indicators may indicate which elements within the array 102 of spectrometer output intensities were more important in determining the analyte concentration than other spectrometer output intensities. In some embodiments, the feature relevance indicators may include which of the peak wavelengths (and the magnitudes of their associated intensities) best predicted the analyte concentration (e.g., when the array of spectrometer output intensities 102 includes the first sub-array 102A in Figure 12) and / or which of the peak wavelengths (and the magnitudes of their associated intensities) best predicted the analyte concentration (e.g., when the array of spectrometer output intensities 102 includes the second sub-array 102B in Figure 12 or other combinations of intensity data). For example, the analyte concentration logic 1006 may generate a list of the most relevant diffraction orders, peak wavelengths, or combinations of peak wavelengths for determining particulate analyte concentrations. The analyte concentration logic 1006 may implement any feature relevance scoring method known in the art, such as linear regression feature relevance, logistic regression feature relevance, or decision tree feature relevance.

[0040] In some embodiments, the analyte concentration logic 1006 may perform further processing on the output of the trained machine learning computation model to determine the concentration of the analyte in the sample. For example, in some embodiments, the analyte concentration logic 1006 may utilize a feature relevance indicator to identify which peak wavelengths were most important to the output of the machine learning computation model, and may use the magnitude of the intensity of these peak wavelengths to determine the analyte concentration. In some specific embodiments, the analyte concentration logic 1006 may express the analyte concentration indicated by the magnitude of the intensity of these “most important” peak wavelengths as an approximation, and may use a mode of the approximately expressed magnitude of intensity to determine the analyte concentration. In some other specific embodiments, the analyte concentration logic 1006 may calculate a weighted average of the analyte concentrations indicated by the magnitude of the intensity of the “most important” peak wavelengths, with weights assigned to each peak wavelength according to a feature relevance indicator (e.g., representing the relative influence of the output intensity associated with different peak wavelengths among the peak wavelengths). In other specific embodiments, the analyte concentration logic 1006 may approximate the analyte concentrations, which are indicated by the magnitude of the intensity of all peak wavelengths, sort the analyte concentrations by decreasing their frequency, and calculate a weighted average of the sets of analyte concentrations that appear most frequently.

[0041] The output logic 1008 may output the concentration of the analyte in the sample determined by the analyte concentration logic 1006 (using the trained machine learning computation model 110). In some embodiments, the output logic 1008 may also output one or more feature relevance indicators generated by the analyte concentration logic 1006 (for example, so that the user can verify which peak wavelengths were selected as most important for determining the analyte concentration). In some embodiments, the output logic 1008 may output the analyte concentration and / or feature relevance indicators to a display device (for example, via a graphical user interface (GUI) such as GUI 3000 in Figure 18). In some embodiments, the output logic 1008 may output the analyte concentration and / or feature relevance indicators to a storage device (for example, storage device 4004 of computing device 4000 in Figure 19). In some embodiments, the output logic 1008 may output analyte concentrations and / or feature relevance indicators to an interface device (e.g., interface device 4006 of computing device 4000 in Figure 19) for transmission to a local or remote computing device.

[0042] Figures 13 to 17 are flowcharts illustrating methods for performing spectrometer support operations according to various embodiments. While the operations of the methods in Figures 13 to 17 can be illustrated by referring to specific embodiments disclosed herein (e.g., the spectrometer support module 1000 discussed herein with reference to Figure 3, the GUI 3000 discussed herein with reference to Figure 18, the computing device 4000 discussed herein with reference to Figure 19, and / or the spectrometer support system 5000 discussed herein with reference to Figure 20), the methods in Figures 13 to 17 can be used in any preferred configuration to perform any preferred support operation. Although the operations are shown once in each of Figures 13 to 17 in a specific order, the operations can be rearranged as desired and appropriately, and / or performed repeatedly (e.g., various operations to be performed may be performed in parallel as appropriate).

[0043] Referring to method 2000 in Figure 13, in 2002, the array of spectrometer output intensities of the calibration samples may be generated for each of several calibration samples of the analyte at different known concentrations. Different spectrometer output intensities in the array may be associated with data representing different deflection amounts. For example, the spectrometer intensity logic 1002 of the spectrometer support module 1000 may perform the operation of 2002.

[0044] In 2004, the machine learning computation model may be trained to output the concentration of an analyte in a target sample based on an input sequence of spectrometer output intensities of the target sample, using multiple associated sequences of multiple known concentrations of the analyte in a calibration sample and spectrometer output intensities. For example, the training logic 1004 of the spectrometer support module 1000 may perform the operation in 2004.

[0045] In 2006, the trained machine learning computational model can be used as a calibration model for the analyte for subsequent spectrometer operation. For example, the analyte concentration logic 1006 of the spectrometer support module 1000 may perform the operation in 2006.

[0046] Referring to method 2100 in Figure 14, in 2102, an array of spectrometer output intensities of the sample may be generated. Different spectrometer output intensities in the array may be associated with data representing different deflection amounts. For example, the spectrometer intensity logic 1002 of the spectrometer support module 1000 may perform the operation of 2102.

[0047] In 2104, the received sequence of spectrometer output intensities may be provided to a trained machine learning computational model. The trained machine learning computational model is for outputting the concentration of the analyte in the sample. For example, the analyte concentration logic 1006 of the spectrometer support module 1000 may perform the operation of 2104.

[0048] In 2106, the concentration of the analyte in the sample may be output. For example, the output logic 1008 of the spectrometer support module 1000 may perform the operation of 2106.

[0049] Referring to method 2200 in Figure 15, an array of sample spectrometer output intensities may be generated in 2202. Different spectrometer output intensities within the array may be associated with data representing different deflection amounts. For example, the spectrometer intensity logic 1002 of the spectrometer support module 1000 may perform the operation of 2202.

[0050] In 2204, the concentration of the analyte in the sample may be generated based on the received sequence of spectrometer output intensities. For example, the analyte concentration logic 1006 of the spectrometer support module 1000 may perform the operation of 2204.

[0051] In 2206, a feature association indicator associated with one or more of the analyte concentration in the sample and the spectrometer output intensity may be output. For example, the output logic 1008 of the spectrometer support module 1000 may perform the operation of 2206.

[0052] Referring to method 2300 in Figure 16, in 2302, an array of spectrometer output intensities of the sample may be generated. Different spectrometer output intensities in the array may be associated with data representing different deflection amounts. For example, the spectrometer intensity logic 1002 of the spectrometer support module 1000 may perform the operation of 2302.

[0053] In 2304, the concentration of the analyte in the sample may be generated based on the received sequence of spectrometer output intensities without requiring the user to select one or more of the data representing the deflection amount prior to the generation of the analyte concentration. For example, the analyte concentration logic 1006 of the spectrometer support module 1000 may perform the operation of 2304.

[0054] In step 2306, the concentration of the analyte in the sample may be output. For example, the output logic 1008 of the spectrometer support module 1000 may perform the operation of step 2306.

[0055] Referring to method 2400 in Figure 17, in 2402, an array of spectrometer output intensities of calibration samples may be generated for each of several calibration samples of the analyte at different known concentrations. Different spectrometer output intensities in the array may be associated with data representing different deflection amounts. For example, the spectrometer intensity logic 1002 of the spectrometer support module 1000 may perform the operation of 2402.

[0056] In 2404, a pre-trained machine learning computational model may be retrained using multiple known concentrations of the analyte in the calibration sample and multiple associated sequences of spectrometer output intensities, and output the concentration of the analyte in the target sample based on the input sequence of spectrometer output intensities of the target sample. Alternatively, in 2404, a preprocessing method may be generated for use with the pre-trained machine learning computational model (for example, to perform normalization as described above). For example, the training logic 1004 of the spectrometer support module 1000 may perform the operation of 2404.

[0057] In 2406, the retrained machine learning computational model may be used as a calibration model for the analyte for subsequent spectrometer operations. Alternatively, in 2406, the preprocessing method generated in 2404 may be executed together with the pretrained machine learning computational model as a calibration model for the analyte for subsequent spectrometer operations. For example, the analyte concentration logic 1006 of the spectrometer support module 1000 may perform the operation in 2406.

[0058] The spectrometer support methods disclosed herein may include interaction with a human user (for example, via a user-local computing device 5020, as discussed herein with reference to Figure 20). These interactions may include providing the user with information (e.g., information about the operation of a spectrometer, such as the spectrometer 5010 in Figure 20, information about a sample being analyzed, or information about other tests or measurements performed by the spectrometer, information retrieved from a local or remote database, or other information), or providing the user with options to input commands (e.g., to control the operation of a spectrometer, such as the spectrometer 5010 in Figure 20, or to control the analysis of data generated by the spectrometer), queries (e.g., to a local or remote database), or other information. In some embodiments, these interactions may be carried out through a graphical user interface (GUI) that includes a visual display on a display device (e.g., display device 4010, discussed herein with reference to Figure 19) which provides the user with outputs and / or instructs the user to provide inputs (e.g., via one or more input devices such as a keyboard, mouse, trackpad, or touchscreen, which are included in other I / O devices 4012 considered herein with reference to Figure 19). The spectrometer support systems disclosed herein may include any preferred GUI for user interaction. In some embodiments, the output logic 1008 may provide any of the GUIs disclosed herein.

[0059] Figure 18 shows an exemplary GUI 3000 that may be used in some or all implementations of the support methods disclosed herein according to various embodiments. As described above, the GUI 3000 may be provided on a display device (e.g., display device 4010, as discussed herein with reference to Figure 19) of a computing device (e.g., computing device 4000, as discussed herein with reference to Figure 19) of a spectrometer support system (e.g., spectrometer support system 5000, as discussed herein with reference to Figure 20), and a user may interact with the GUI 3000 using any suitable input device (e.g., any of the input devices included in other I / O devices 4012, as discussed herein with reference to Figure 19), and input techniques (e.g., cursor movement, motion capture, face recognition, gesture detection, voice recognition, button activation, etc.).

[0060] GUI3000 may include a data display area 3002, a data analysis area 3004, a spectrometer control area 3006, and a setting area 3008. The specific number and arrangement of areas shown in Figure 18 are illustrative only, and GUI3000 may include any number and arrangement of areas containing any desired features.

[0061] The data display area 3002 may display data generated by a spectrometer (for example, the spectrometer 5010 discussed herein with reference to Figure 20). For example, the data display area 3002 may display the output intensity signal from the spectrometer (which may be background-corrected by the spectrometer intensity logic 1002 or processed in other ways).

[0062] The data analysis area 3004 may display the results of data analysis (e.g., the results of analyzing data illustrated in the data display area 3002 and / or other data). For example, the data analysis area 3004 may display the concentration of the analyte of interest in the sample (e.g., determined by the analyte concentration logic 1006 according to any of the embodiments disclosed herein), one or more feature relevance indicators (e.g., determined by the analyte concentration logic 1006 according to any of the embodiments disclosed herein), or any other suitable information. In some embodiments, the data display area 3002 and the data analysis area 3004 may be combined within the GUI 3000 (e.g., to include data output from a spectrometer and some analysis of the data in a common graph or area).

[0063] The spectrometer control area 3006 may include options that allow the user to control the spectrometer (e.g., the spectrometer 5010 discussed herein with reference to Figure 20). For example, the spectrometer control area 3006 may include options to initiate or otherwise control the analysis of a sample by the spectrometer.

[0064] The configuration area 3008 may include options that allow the user to control the features and functions of GUI 3000 (and / or other GUIs) and / or perform common computing operations relating to the data display area 3002 and the data analysis area 3004 (for example, storing data such as analyte concentrations and / or feature relevance indicators on a storage device such as storage device 4004 as discussed herein, referring to Figure 19; transmitting data such as analyte concentrations and / or feature relevance indicators to another user; displaying the data, etc.).

[0065] As described above, the spectrometer support module 1000 can be implemented by one or more computing devices. Figure 19 is a block diagram of a computing device 4000 that can perform some or all of the spectrometer support methods disclosed herein in various embodiments. In some embodiments, the spectrometer support module 1000 can be implemented by a single computing device 4000 or multiple computing devices 4000. Furthermore, as discussed below, the computing device 4000 (or multiple computing devices 4000) implementing the spectrometer support module 1000 may be part of one or more of the spectrometer 5010, the user local computing device 5020, the service local computing device 5030, or the remote computing device 5040 in Figure 20. In some embodiments, the processor 14, memory 15, and I / O unit 16 may be part of the computing device 4000.

[0066] The computing device 4000 in Figure 19 is illustrated as having several components, but one or more of these components may be omitted or duplicated to suit the application and configuration. In some embodiments, some or all of the components included in the computing device 4000 may be mounted on one or more motherboards and enclosed in a housing (e.g., including plastic, metal, and / or other materials). In some embodiments, some of these components can be manufactured on a single system-on-a-chip (SoC) (for example, the SoC may include one or more processing devices 4002 and one or more storage devices 4004). Additionally, in various embodiments, the computing device 4000 may not include one or more of the components illustrated in Figure 19, but may include interface circuits (not shown) for coupling to one or more components using any suitable interface (e.g., a Universal Serial Bus (USB) interface, a High-Definition Multimedia Interface (HDMI), a Controller Area Network (CAN) interface, a Serial Peripheral Interface (SPI) interface, an Ethernet interface, a wireless interface, or any other suitable interface). For example, the computing device 4000 may not include the display device 4010, but may include display device interface circuits (e.g., connectors and driver circuits) that enable coupling to the display device 4010.

[0067] The computing device 4000 may include processing devices 4002 (e.g., one or more processing devices). As used herein, the term “processing device” may mean any device or part of a device that processes electronic data from registers and / or memory and converts that electronic data into other electronic data that can be stored in registers and / or memory. The processing devices 4002 may include one or more digital signal processors (DSPs), application-specific integrated circuits (ASICs), central processing units (CPUs), graphics processing units (GPUs), cryptographic processors (dedicated processors that execute cryptographic algorithms in hardware), server processors, or any other suitable processing devices.

[0068] The computing device 4000 may include a storage device 4004 (e.g., one or more storage devices). The storage device 4004 may include one or more memory devices such as random access memory (RAM) (e.g., static RAM (SRAM) devices, magnetic RAM (MRAM) devices, dynamic RAM (DRAM) devices, resistive RAM (RRAM) devices, or conductive-bridging RAM (CBRAM) devices), hard drive-based memory devices, solid-state memory devices, network drives, cloud drives, or any combination of memory devices. In some embodiments, the storage device 4004 may include memory that shares a die with the processing device 4002. In such embodiments, the memory may be used as cache memory and may include, for example, embedded dynamic random-access memory (eDRAM) or spin transfer torque magnetic random-access memory (STT-MRAM). In some embodiments, the storage device 4004 may include a non-temporary computer-readable medium having instructions that cause the computing device 4000 to perform any suitable method or part thereof of the methods disclosed herein when executed by one or more processing devices (e.g., processing device 4002).

[0069] The computing device 4000 may include an interface device 4006 (for example, one or more interface devices 4006). The interface device 4006 may include one or more communication chips, connectors, and / or other hardware and software to manage communication between the computing device 4000 and other computing devices. For example, the interface device 4006 may include circuitry that manages wireless communication for transferring data to and from the computing device 4000. The term “wireless” and its derivatives may be used to describe circuits, devices, systems, methods, techniques, communication channels, etc., that can communicate data through the use of modulated electromagnetic radiation over a non-solid medium. This term does not imply that the devices in question do not include any wiring, although in some embodiments they may not. The circuitry included in the interface device 4006 for managing wireless communications may implement any of several wireless standards or protocols, including, but are not limited to, Wi-Fi (IEEE 802.11 family), IEEE 802.16 standards (e.g., IEEE 802.16-2005 Amendment), Institute for Electrical and Electronic Engineers (IEEE) standards, and Long-Term Evolution (LTE) projects (e.g., Advanced LTE project, Ultra-Mobile Broadband (UMB) project (also known as "3GPP2")) with any modifications, updates, and / or revisions.In some embodiments, the circuitry included in the interface device 4006 for managing wireless communication may operate according to the Global System for Mobile Communication (GSM), General Packet Radio Service (GPRS), Universal Mobile Telecommunications System (UMTS), High Speed ​​Packet Access (HSPA), Evolved HSPA (E-HSPA), or LTE network. In some embodiments, the circuitry included in the interface device 4006 for managing wireless communication may operate according to Enhanced Data for GSM Evolution (EDGE), GSM EDGE Radio Access Network (GERAN), Universal Terrestrial Radio Access Network (UTRAN), or Evolved UTRAN (E-UTRAN). In some embodiments, the circuitry included in the interface device 4006 for managing wireless communication may operate in accordance with Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Digital Enhanced Cordless Telecommunications (DECT), Evolution-Data Optimized (EV-DO), and their derivatives, as well as any other wireless protocols designated as 3G, 4G, 5G, and beyond. In some embodiments, the interface device 4006 may include one or more antennas (e.g., one or more antenna arrays) for receiving and / or transmitting wireless communication.

[0070] In some embodiments, the interface device 4006 may include circuitry for managing wired communications, such as electrical, optical, or any other preferred communication protocol. For example, the interface device 4006 may include circuitry to support communications according to Ethernet technology. In some embodiments, the interface device 4006 may support both wireless and wired communications and / or multiple wired communication protocols and / or multiple wireless communication protocols. For example, a first set of circuits in the interface device 4006 may be dedicated to short-range wireless communications such as Wi-Fi or Bluetooth, and a second set of circuits in the interface device 4006 may be dedicated to long-range wireless communications such as Global Positioning System (GPS), EDGE, GPRS, CDMA, WiMAX, LTE, EV-DO, or others. In some embodiments, a first set of circuits in the interface device 4006 may be dedicated to wireless communications, and a second set of circuits in the interface device 4006 may be dedicated to wired communications.

[0071] The computing device 4000 may include a battery / power circuit 4008. The battery / power circuit 4008 may include one or more energy storage devices (e.g., batteries or capacitors) and / or circuits for coupling components of the computing device 4000 to an energy source separate from the computing device 4000 (e.g., AC line power).

[0072] The computing device 4000 may include a display device 4010 (for example, multiple display devices). The display device 4010 may include any visual indicator such as a head-up display, computer monitor, projector, touchscreen display, liquid crystal display (LCD), light-emitting diode display, or flat panel display.

[0073] The computing device 4000 may include other input / output (I / O) devices 4012. These other I / O devices 4012 may include, for example, one or more audio output devices (e.g., speakers, headsets, earphones, alarms, etc.), one or more audio input devices (e.g., microphones or microphone arrays), location devices (e.g., GPS devices that communicate with satellite-based systems to receive the location of the computing device 4000, as is known in the art), audio codecs, video codecs, printers, sensors (e.g., thermocouples or other temperature sensors, humidity sensors, pressure sensors, vibration sensors, accelerometers, gyroscopes, etc.), image capture devices such as cameras, cursor control devices such as keyboards, mice, styluses, trackballs, or touchpads, barcode readers, Quick Response (QR) code readers, or radio frequency identification (RFID) readers.

[0074] The computing device 4000 may have any suitable form factor for the application and configuration, such as a handheld or mobile computing device (e.g., a cell phone, smartphone, mobile internet device, tablet computer, laptop computer, netbook computer, ultrabook computer, personal digital assistant (PDA), ultramobile personal computer, etc.), a desktop computing device, or a server computing device, or other network computing component.

[0075] One or more computing devices implementing any of the spectrometer support modules or methods disclosed herein may be part of a spectrometer support system. Figure 20 is a block diagram of an exemplary spectrometer support system 5000 in which some or all of the spectrometer support methods disclosed herein may be performed according to various embodiments. The spectrometer support modules and methods disclosed herein (e.g., the spectrometer support module 1000 in Figure 3 and the methods in Figures 13 to 17) may be implemented by one or more of the spectrometer 5010, the user local computing device 5020, the service local computing device 5030, or the remote computing device 5040 of the spectrometer support system 5000. In some embodiments, the spectrometer measurement system 10 in Figure 1 may be part of the spectrometer support system 5000.

[0076] Any of the spectrometer 5010, user local computing device 5020, service local computing device 5030, or remote computing device 5040 may include any of the embodiments of the computing device 4000 discussed herein with reference to Figure 19, and any of the spectrometer 5010, user local computing device 5020, service local computing device 5030, or remote computing device 5040 may take the form of any suitable embodiment of the computing device 4000 discussed herein with reference to Figure 19.

[0077] The spectrometer 5010, user local computing device 5020, service local computing device 5030, or remote computing device 5040 may each include a processing device 5002, a storage device 5004, and an interface device 5006. The processing device 5002 may take any preferred form, including any form of the processing device 4002 considered herein with reference to Figure 6, and the processing device 5002 included in different versions of the spectrometer 5010, user local computing device 5020, service local computing device 5030, or remote computing device 5040 may take the same or different forms. The storage device 5004 may take any preferred form, including any form of the storage device 5004 considered herein with reference to Figure 6, and the storage device 5004 included in different versions of the spectrometer 5010, user local computing device 5020, service local computing device 5030, or remote computing device 5040 may take the same or different forms. The interface device 5006 may take any preferred form, including any form of the interface device 4006 discussed herein with reference to Figure 6, and the interface device 5006 included in different versions of the spectrometer 5010, user local computing device 5020, service local computing device 5030, or remote computing device 5040 may take the same or different forms.

[0078] The spectrometer 5010, user local computing device 5020, service local computing device 5030, and remote computing device 5040 may communicate with other elements of the spectrometer support system 5000 via a communication path 5008. The communication path 5008 may be a wired or wireless communication path, and may communicatively couple the interface devices 5006 of different elements of the spectrometer support system 5000 (for example, according to any of the communication techniques considered herein with reference to the interface device 4006 of computing device 4000 in Figure 19). The particular spectrometer support system 5000 depicted in Figure 20 includes communication paths between each pair of spectrometer 5010, user local computing device 5020, service local computing device 5030, and remote computing device 5040, but this “fully coupled” implementation is merely illustrative, and various forms may not exist in different embodiments. For example, in some embodiments, the service local computing device 5030 may not have a direct communication path 5008 between its interface device 5006 and the interface device 5006 of the spectrometer 5010. Instead, it may communicate with the spectrometer 5010 via the communication path 5008 between the service local computing device 5030 and the user local computing device 5020, and the communication path 5008 between the user local computing device 5020 and the spectrometer 5010.

[0079] The spectrometer 5010 may include any suitable spectrometer, such as an inductively coupled plasma optical emission spectrometer (ICP-OES), a mass spectrometer, or any other suitable spectrometer.

[0080] The user-local computing device 5020 may be a computing device that is local to the user of the spectrometer 5010 (for example, according to any embodiment of the computing device 4000 considered herein). In some embodiments, the user-local computing device 5020 may also be, but not required to be, local to the spectrometer 5010. For example, the user-local computing device 5020 located in the user's home or office may be remote from the spectrometer 5010 but may communicate with the spectrometer 5010 so that the user can use the user-local computing device 5020 to control and / or access data from the spectrometer 5010. In some embodiments, the user-local computing device 5020 may be a laptop, smartphone, or tablet device. In some embodiments, the user-local computing device 5020 may be a portable computing device.

[0081] The service local computing device 5030 may be a computing device that is local to an entity servicing the spectrometer 5010 (for example, according to any embodiment of the computing device 4000 considered herein). For example, the service local computing device 5030 may be local to the manufacturer of the spectrometer 5010 or to a third-party service company. In some embodiments, the service local computing device 5030 may communicate with the spectrometer 5010, the user local computing device 5020, and / or the remote computing device 5040 (for example, via a direct communication path 5008 or via a plurality of “indirect” communication paths 5008, as considered above), and receive data relating to the operation of the spectrometer 5010, the user local computing device 5020, and / or the remote computing device 5040 (for example, the results of a self-test of the spectrometer 5010, the calibration coefficients used by the spectrometer 5010, the measurements of sensors associated with the spectrometer 5010, etc.). In some embodiments, the service local computing device 5030 may communicate with the spectrometer 5010, the user local computing device 5020, and / or the remote computing device 5040 (for example, via a direct communication path 5008 or via multiple “indirect” communication paths 5008, as considered above), and transmit data to the spectrometer 5010, the user local computing device 5020, and / or the remote computing device 5040 (for example, updating programmed instructions such as firmware in the spectrometer 5010, initiating the execution of a test or calibration sequence for the spectrometer 5010, updating programmed instructions such as software in the user's local computing device 5020 or the remote computing device 5040, etc.).The user of the spectrometer 5010 may communicate with the service local computing device 5030 to report problems with the spectrometer 5010 or the user local computing device 5020, request a visit from a technician to improve the operation of the spectrometer 5010, order consumables or replacement parts related to the spectrometer 5010, or use the spectrometer 5010 or the user local computing device 5020 for other purposes.

[0082] The remote computing device 5040 may be a computing device located remotely from the spectrometer 5010 and / or the user-local computing device 5020 (for example, according to any embodiment of the computing device 4000 considered herein). In some embodiments, the remote computing device 5040 may be included in a data center or other large-scale server environment. In some embodiments, the remote computing device 5040 may include network-attached storage (for example, as part of storage device 5004). The remote computing device 5040 may store data generated by the spectrometer 5010, perform analysis of the data generated by the spectrometer 5010 (for example, according to programmed instructions), facilitate communication between the user-local computing device 5020 and the spectrometer 5010, and / or facilitate communication between the service-local computing device 5030 and the spectrometer 5010.

[0083] In some embodiments, one or more of the elements of the spectrometer support system 5000 illustrated in Figure 20 may be absent. Furthermore, in some embodiments, multiple of the various elements of the spectrometer support system 5000 in Figure 20 may be present. For example, the spectrometer support system 5000 may include multiple user local computing devices 5020 (e.g., different user local computing devices 5020 associated with different users or at different locations). In another example, the spectrometer support system 5000 may include multiple spectrometers 5010 all communicating with a service local computing device 5030 and / or a remote computing device 5040, in which embodiment the service local computing device 5030 may monitor these multiple spectrometers 5010, or the service local computing device 5030 may trigger updates or other information may be simultaneously "broadcast" to the multiple spectrometers 5010. Different spectrometers 5010 within the spectrometer support system 5000 may be located close to each other (e.g., in the same room) or far apart from each other (e.g., on different floors of a building, in different buildings, in different cities, etc.). In some embodiments, the spectrometers 5010 may be coupled to an Internet of Things (IoT) stack that enables command and control of the spectrometers 5010 via web-based applications, virtual or augmented reality applications, mobile applications, and / or desktop applications. Any of these applications may be accessed by a user who communicates with the spectrometers 5010 and operates a user-local computing device 5020 via an intervening remote computing device 5040. In some embodiments, the spectrometers 5010 may be sold by the manufacturer as part of a local spectrometer computing unit 5012, together with one or more associated user-local computing devices 5020.

[0084] In some embodiments, different spectrometers 5010 included in the spectrometer support system 5000 may be different types of spectrometers 5010, for example, one spectrometer 5010 may be a mass spectrometer and another spectrometer 5010 may be an optical spectrometer. In some such embodiments, the remote computing device 5040 and / or the user local computing device 5020 may combine data from different types of spectrometers 5010 included in the spectrometer support system 5000.

[0085] The following paragraphs provide various examples of embodiments disclosed herein.

[0086] Example A1 is a spectrometer support device comprising: a first logic for generating an array of spectrometer output intensities for each of a plurality of calibration samples of analytes at different known concentrations, wherein different spectrometer output intensities in the array are associated with data representing different deflection amounts; a second logic for training a machine learning computational model using a plurality of arrays associated with multiple known concentrations of analytes in the calibration samples and spectrometer output intensities, and for outputting the concentration of analytes in a target sample based on an input array of spectrometer output intensities for the target sample; and a third logic for using the machine learning computational model trained as a calibration model for the analyte for subsequent spectrometer operation.

[0087] Example A2 includes the subject matter of Example A1 and further clarifies that the spectrometer is an optical spectrometer and that the data representing different deflection amounts represent data representing different wavelengths of radiation.

[0088] Example A3 includes the subject matter of Example A1 and further clarifies that the spectrometer is a mass spectrometer and that data representing different deflection amounts represent data representing different mass-to-charge ratios.

[0089] Example A4 includes the subject matter of any of Examples A1-3 and further clarifies that the analyte is a single element.

[0090] Example A5 further clarifies that it includes the subject matter of any of Examples A1-4 and that the tensor input to the machine learning computation model includes spectrometer output intensities at different deflection levels.

[0091] Example A6 further clarifies that it includes themes from any of Examples A1-5 and that the tensor input to the machine learning computation model includes ratios of different spectrometer output intensities at different deflection levels.

[0092] Example A7 includes the subject matter of any of Examples A1 to A6 and further clarifies that the array of spectrometer output intensities of the calibration sample is the array of output intensities corrected for background.

[0093] Example A8 includes the subject matter of any of Examples A1-7, wherein the calibration sample is the first calibration sample, the analyte is the first analyte, the machine learning computational model is the first machine learning computational model, the first logic is for generating an array of spectrometer output intensities of the second calibration sample for each of several second calibration samples of the second analyte at different known concentrations, the different spectrometer output intensities in the array are associated with data representing different deflection amounts, the second analyte is different from the first analyte, the second logic trains the second machine learning computational model using multiple arrays associated with multiple known concentrations of the second analyte in the second calibration sample and spectrometer output intensities, outputs the concentration of the second analyte in the target sample based on the input array of spectrometer output intensities of the target sample, and the third logic further specifies that the trained second machine learning computational model is used as the calibration model for the second analyte for subsequent spectrometer operation.

[0094] Example A9 incorporates the themes of Examples A1-8 and further clarifies that for each of several calibration samples of analytes at different known concentrations, the array of spectrometer output intensities of the calibration sample includes three or more output intensities.

[0095] Example A10 includes the subject matter of any of Examples A1 to A9 and further clarifies that for one or more calibration samples of analytes at different known concentrations, the array of spectrometer output intensities of the calibration sample includes at least one saturated spectrometer output intensity.

[0096] Example BI1 is a spectrometer support device comprising: a first logic for generating an array of spectrometer output intensities of a sample, wherein different spectrometer output intensities in the array are associated with data representing different deflection amounts; a second logic for providing the received array of spectrometer output intensities to a trained machine learning computational model, wherein the trained machine learning computational model is for outputting the concentration of an analyte in the sample; and a third logic for outputting the concentration of an analyte in the sample.

[0097] Example BI2 includes the themes of Example BI1 and further clarifies that the spectrometer is an optical spectrometer and that the data representing different deflection amounts represent data representing different wavelengths of radiation.

[0098] Example BI3 includes the themes of Example BI1 and further clarifies that the spectrometer is a mass spectrometer and that data representing different deflection amounts represent data representing different mass-to-charge ratios.

[0099] Example BI4 includes the subject matter of any of Examples BI1-3 and further clarifies that the analyte is a single element.

[0100] Example BI5 includes themes from any of Examples BI1-4 and further clarifies that the array of spectrometer output intensities of the sample is a background-corrected array of output intensities.

[0101] Example BI6 includes the subject matter of any of Examples BI1-5 and further clarifies that the array of spectrometer output intensities of the sample includes three or more output intensities.

[0102] Example BI7 includes the subject matter of any of Examples BI1-6, and further clarifies that the array of spectrometer output intensities of the sample includes at least one saturated spectrometer output intensity.

[0103] Example BI8 includes the subject matter of any of Examples BI1-7 and further clarifies that a third logic outputs a feature relevance indicator associated with one or more of the spectrometer output intensities to a display device.

[0104] Example BI9 incorporates the themes of Example BI8 and further clarifies that a feature relevance indicator associated with spectrometer output intensity indicates that spectrometer output intensity was more important than other spectrometer output intensity in determining the concentration of the analyte in the sample.

[0105] Example BI10 incorporates the subject matter of any of Examples BI8-9 and further clarifies that a feature association indicator associated with two spectrometer output intensities indicates that a combination of two spectrometer output intensities was more important than any other combination of spectrometer output intensities for determining the concentration of the analyte in the sample.

[0106] Example BI11 incorporates the themes of Example BI10 and further clarifies that a feature correlation indicator associated with the two spectrometer output intensities indicates that the ratio of the two spectrometer output intensities was more important than other ratios of spectrometer output intensities for determining the concentration of the analyte in the sample.

[0107] Example BI12 further clarifies that it includes the subject matter of any of Examples BI8-11 and includes a list of specific data where the feature relevance indicator represents different deflection amounts associated with a particular spectrometer output intensity.

[0108] Example BI13 includes the subject matter of Example BI12 and further clarifies that the characteristic relevance indicator includes a list of wavelengths that are more important than other wavelengths for determining the concentration of the analyte in the sample.

[0109] Example BI14 includes the subject matter of Example BI12 and further clarifies that the characteristic relevance indicator includes a list of mass-to-charge ratios that are more important than other mass-to-charge ratios for determining the concentration of the analyte in the sample.

[0110] Example BI15 incorporates the subject matter of any of Examples BI1 to BI14 and further clarifies that a third logic outputs the concentration of the analyte in the sample to a display device.

[0111] Example BII1 is a spectrometer support device comprising: a first logic for generating an array of spectrometer output intensities of a sample, wherein different spectrometer output intensities in the array are associated with data representing different deflection amounts; a second logic for generating the concentration of an analyte in the sample based on the received array of spectrometer output intensities; and a third logic for outputting the concentration of an analyte in the sample and a feature association indicator associated with one or more of the spectrometer output intensities.

[0112] Example BII2 includes the themes of Example BII1 and further clarifies that the spectrometer is an optical spectrometer and that the data representing different deflection amounts represent data representing different wavelengths of radiation.

[0113] Example BII3 includes the themes of Example BII1 and further clarifies that the spectrometer is a mass spectrometer and that data representing different deflection amounts represent data representing different mass-to-charge ratios.

[0114] Example BII4 includes the subject matter of any of Examples BII1-3 and further clarifies that the analyte is a single element.

[0115] Example BII5 includes themes from any of Examples BII1-4 and further clarifies that the array of spectrometer output intensities of the sample is a background-corrected array of output intensities.

[0116] Example BII6 includes the subject matter of any of Examples BII1-5 and further clarifies that the array of spectrometer output intensities of the sample includes three or more output intensities.

[0117] Example BII7 includes the subject matter of any of Examples BII1-6 and further clarifies that the array of spectrometer output intensities of the sample includes at least one saturated spectrometer output intensity.

[0118] Example BII8 incorporates the subject matter of any of Examples BII1-7 and further clarifies that a feature relevance indicator associated with spectrometer output intensity indicates that the spectrometer output intensity was more important than other spectrometer output in determining the concentration of the analyte in the sample.

[0119] Example BII9 incorporates themes from any of Examples BII1-8 and further clarifies that a feature association indicator associated with two spectrometer output intensities indicates that a combination of two spectrometer output intensities was more important than any other combination of spectrometer output intensities for determining the concentration of the analyte in the sample.

[0120] Example BII10 incorporates the subject matter of Example BII9 and further clarifies that a feature association indicator associated with two spectrometer output intensities indicates that the ratio of two spectrometer output intensities was more important than other ratios of spectrometer output intensities for determining the concentration of the analyte in the sample.

[0121] Example BII11 further clarifies that it includes the subject matter of any of Examples BII1-10 and includes a list of specific data in which the feature relevance indicator represents different deflection amounts associated with a particular spectrometer output intensity.

[0122] Example BII12 includes the subject matter of Example BII11 and further clarifies that the characteristic relevance indicator includes a list of wavelengths that are more important than other wavelengths for determining the concentration of the analyte in the sample.

[0123] Example BII13 includes the subject matter of Example BII11 and further clarifies that the characteristic relevance indicator includes a list of mass-to-charge ratios that are more important than other mass-to-charge ratios for determining the concentration of the analyte in the sample.

[0124] Example BII14 incorporates the subject matter of any of Examples BII1 to BII13 and further clarifies that a third logic outputs the concentration of the analyte in the sample to a display device.

[0125] Example BIII1 is a spectrometer output device comprising: a first logic for generating an array of spectrometer output intensities of a sample, wherein different spectrometer output intensities in the array are associated with data representing different deflection amounts; a second logic for generating the concentration of an analyte in a sample based on the received array of spectrometer output intensities, without requiring the user to select one or more of the data representing deflection amounts prior to generating the concentration of an analyte in the sample; and a third logic for outputting the concentration of an analyte in the sample.

[0126] Example BIII2 includes the subject matter of Example BIII1 and further clarifies that the spectrometer is an optical spectrometer and that the data representing different deflection amounts represent data representing different wavelengths of radiation.

[0127] Example BIII3 includes the subject matter of Example BIII1 and further clarifies that the spectrometer is a mass spectrometer and that data representing different deflection amounts represent data representing different mass-to-charge ratios.

[0128] Example BIII4 includes the subject matter of any of Examples BIII1-3 and further clarifies that the analyte is a single element.

[0129] Example BIII5 includes the subject matter of any of Examples BIII1 to BIII4 and further clarifies that the sequence of spectrometer output intensities of the sample is a background-corrected sequence of output intensities.

[0130] Example BIII6 includes the subject matter of any of Examples BIII1 to BIII5 and further clarifies that the array of spectrometer output intensities of the sample includes three or more output intensities.

[0131] Example BIII7 includes the subject matter of any of Examples BIII1 to BIII6, and further clarifies that the array of spectrometer output intensities of the sample includes at least one saturated spectrometer output intensity.

[0132] Example BIII8 includes the subject matter of any of Examples BIII1 to BIII7 and further clarifies that a third logic outputs a feature relevance indicator associated with one or more of the spectrometer output intensities to a display device.

[0133] Example BIII9 incorporates the subject matter of Example BIII8 and further clarifies that a feature relevance indicator associated with spectrometer output intensity indicates that spectrometer output intensity was more important than other spectrometer output intensity in determining the concentration of the analyte in the sample.

[0134] Example BIII10 incorporates the subject matter of any of Examples BIII8-9 and further clarifies that a feature association indicator associated with two spectrometer output intensities indicates that a combination of two spectrometer output intensities was more important than any other combination of spectrometer output intensities for determining the concentration of the analyte in the sample.

[0135] Example BIII11 incorporates the subject matter of Example BIII10 and further clarifies that a feature association indicator associated with two spectrometer output intensities indicates that the ratio of two spectrometer output intensities was more important than other ratios of spectrometer output intensities for determining the concentration of the analyte in the sample.

[0136] Example BIII12 further clarifies that it includes the subject matter of any of Examples BIII8-11 and includes a list of specific data in which the feature relevance indicator represents different deflection amounts associated with a particular spectrometer output intensity.

[0137] Example BIII13 incorporates the subject matter of Example BIII12 and further clarifies that the feature relevance indicator includes a list of wavelengths that are more important than other wavelengths for determining the concentration of the analyte in the sample.

[0138] Example BIII14 includes the subject matter of Example BIII12 and further clarifies that the characteristic relevance indicator includes a list of mass-to-charge ratios that are more important than other mass-to-charge ratios for determining the concentration of the analyte in the sample.

[0139] Example BIII15 includes the subject matter of any of Examples BIII1 to BIII14 and further clarifies that the third logic outputs the concentration of the analyte in the sample to the display device.

[0140] Example C1 is a spectrometer support device comprising: a first logic for generating an array of spectrometer output intensities of a calibration sample for each of a plurality of calibration samples of analytes at different known concentrations, wherein different spectrometer output intensities in the array are associated with data representing different deflection amounts; a second logic for generating a pre-processing method for retraining a pre-trained machine learning computational model or for use with a pre-trained machine learning computational model using a plurality of associated arrays of multiple known concentrations of analytes in the calibration samples, and outputting the concentration of analytes in a target sample based on an input array of spectrometer output intensities of the target sample; and a third logic for running the pre-processing method using the retrained machine learning computational model or with a pre-trained machine learning computational model as a calibration model of the analyte for subsequent spectrometer operation.

[0141] Example C2 includes the subject matter of Example C1 and further clarifies that the spectrometer is an optical spectrometer and that the data representing different deflection amounts represent data representing different wavelengths of radiation.

[0142] Example C3 includes the subject matter of Example C1 and further clarifies that the spectrometer is a mass spectrometer and that data representing different deflection amounts represent data representing different mass-to-charge ratios.

[0143] Example C4 includes the subject matter of any of Examples C1-C3 and further clarifies that the analyte is a single element.

[0144] Example C5 further clarifies that it includes the subject matter of any of Examples C1-4 and that the tensor input to the machine learning computation model includes spectrometer output intensities at different deflection levels.

[0145] Example C6 further clarifies that it includes the subject matter of any of Examples C1-C5 and that the tensor input to the machine learning computation model includes the ratio of different spectrometer output intensities at different deflection amounts.

[0146] Example C7 includes the subject matter of any of Examples C1 to C6 and further clarifies that the array of spectrometer output intensities of the calibration sample is the array of output intensities corrected for background.

[0147] Example C8 includes the subject matter of any of Examples C1 to C7, further clarifying that the spectrometer is the first spectrometer and the pre-trained machine learning computation model was trained using data generated by a second spectrometer, which is different from the first spectrometer.

[0148] Example C9 includes the subject matter of Example C8 and further clarifies that the amount of data from a first spectrometer used to retrain a pre-trained machine learning computational model is less than the amount of data from a second spectrometer used to pre-train the machine learning computational model.

[0149] Example C10 includes the subject matter of any of Examples C1 to C9 and further clarifies that for each of several calibration samples of analytes at different known concentrations, the array of spectrometer output intensities of the calibration sample includes three or more output intensities.

[0150] Example C11 includes the subject matter of any of Examples C1 to C10 and further clarifies that for one or more calibration samples of analytes at different known concentrations, the array of spectrometer output intensities of the calibration sample includes at least one saturated spectrometer output intensity.

[0151] Example C12 includes the subject matter of any of Examples C1-C11 and further clarifies that the second logic is for retraining a subset of the parameters of a pre-trained machine learning computation model.

[0152] Example C13 includes the subject matter of any of Examples C1-C12 and further clarifies that the second logic is for retraining only the final layer of a pre-trained machine learning computation model.

[0153] Example D includes any of the spectrometer support modules disclosed herein.

[0154] Example E includes any of the spectrometer support methods disclosed herein.

[0155] Example F includes any of the GUIs disclosed herein.

[0156] Example G includes any of the spectrometer-assisted computing devices and systems disclosed herein.

[0157] Example H includes a spectrometer system that includes any of the spectrometer support modules or devices disclosed herein.

[0158] Example I includes a spectrometer system configured to perform any of the spectrometer support methods disclosed herein.

[0159] For the purposes of this disclosure, the phrases "A and / or B" and "A or B" mean (A), (B), or (A and B). For the purposes of this disclosure, the phrases "A, B, and / or C" and "A, B, or C" mean (A), (B), (C), (A and B), (A and C), (B and C), or (A, B, and C). Some elements may be referred to in the singular form (e.g., "processing device"), but any suitable element may be represented by multiple instances of that element, and vice versa. For example, a set of operations described as being performed by a processing device may be implemented using different operations performed by different processing devices.

[0160] This description uses the phrases “one embodiment,” “various embodiments,” and “several embodiments,” each of which may refer to one or more identical or different embodiments. Furthermore, terms such as “comprising,” “including,” and “having” as used in reference to embodiments of this disclosure are synonymous. When used to describe a range of dimensions, the phrase “between X and Y” refers to a range including X and Y. As used herein, “apparatus” may refer to any individual device, a collection of devices, a part of a device, or a collection of parts of a device. Drawings are not necessarily to scale.

Claims

1. A spectrometer support device, A first logic for generating an array of spectrometer output intensities of a sample, wherein different spectrometer output intensities in the array are associated with data representing different deflection amounts. A second logic for providing a trained machine learning computational model with a received array of spectrometer output intensities and at least one ratio between the spectrometer output intensities associated with one deflection and the spectrometer output intensities associated with different deflections, wherein the trained machine learning computational model is for outputting the concentration of an analyte in the sample. A spectrometer support device comprising a third logic for outputting the concentration of the analyte in the sample.

2. The spectrometer is an optical spectrometer, and the data representing different deflection amounts are data representing different wavelengths of radiation, as described in claim 1.

3. The spectrometer is a mass spectrometer, and the data representing different deflection amounts are data representing different mass-to-charge ratios, as described in claim 1.

4. The spectrometer support apparatus according to any one of claims 1 to 3, wherein the analyte is a single element.

5. The spectrometer support device according to any one of claims 1 to 3, wherein the array of spectrometer output intensities of the sample includes three or more output intensities.

6. A method for determining the concentration of an analyte from the output intensity of a spectrometer, To generate an array of spectrometer output intensities of a sample, wherein different spectrometer output intensities within the array are associated with data representing different deflection amounts. To provide a trained machine learning computational model with data representing at least some of the received sequences of spectrometer output intensities, wherein the trained machine learning computational model is for outputting the concentration of an analyte in the sample, and the data provided to the trained machine learning computational model includes at least one ratio between the spectrometer output intensities associated with one deflection and the spectrometer output intensities associated with different deflection amounts. A method comprising outputting the concentration of the analyte in the sample.

7. The method according to claim 6, further comprising generating a feature association indicator associated with one or more of the spectrometer output intensities based on the output of the trained machine learning computation model.

8. The method according to claim 7, wherein the feature relevance indicator associated with the spectrometer output intensity indicates that the spectrometer output intensity was more important than other spectrometer output intensity in determining the concentration of the analyte in the sample.

9. The method according to claim 7, wherein the feature association indicator associated with the two spectrometer output intensities indicates that the combination of the two spectrometer output intensities was more important for determining the concentration of the analyte in the sample than any other combination of spectrometer output intensities.

10. The method according to claim 9, wherein the feature correlation indicator associated with the two spectrometer output intensities indicates that the ratio of the two spectrometer output intensities was more important than any other ratio of the spectrometer output intensities for determining the concentration of the analyte in the sample.

11. The method according to any one of claims 7 to 10, wherein generating the feature relevance indicator includes performing linear regression feature relevance, logistic regression feature relevance, or decision tree feature relevance.

12. The method according to any one of claims 7 to 10, wherein the feature relevance indicator includes a list of wavelengths that are more important than other wavelengths in determining the concentration of the analyte in the sample.

13. The method according to any one of claims 7 to 10, wherein the feature relevance indicator includes a list of mass-to-charge ratios that are more important than other mass-to-charge ratios in determining the concentration of the analyte in the sample.

14. A spectrometer system, It is equipped with a spectrometer support module, and the spectrometer support module is A first logic for generating an array of spectrometer output intensities of a sample, wherein different spectrometer output intensities in the array are associated with data representing different deflection amounts. A second logic for providing a trained machine learning computational model with at least a ratio of spectrometer output intensities associated with different deflection amounts, wherein the trained machine learning computational model is for outputting the concentration of an analyte in the sample; A spectrometer system comprising a third logic for outputting the concentration of the analyte in the sample.

15. The spectrometer system according to claim 14, wherein the third logic is for outputting the concentration of the analyte in the sample to a display device.

Images